Optimal Dose Reduction in Computed Tomography Methodologies Predicted from Real-Time Dosimetry

MISSING IMAGE

Material Information

Title:
Optimal Dose Reduction in Computed Tomography Methodologies Predicted from Real-Time Dosimetry
Physical Description:
1 online resource (296 p.)
Language:
english
Creator:
Tien,Christopher J
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Biomedical Engineering
Committee Chair:
Hintenlang, David E
Committee Members:
Rill, Lynn
Bolch, Wesley E
Welt, Bruce A

Subjects

Subjects / Keywords:
ccd -- ct -- dose -- dosimeter -- imaging -- kilovoltage -- medical -- mppc -- ophthalmology -- overranging -- pmt -- psd -- radiation -- radiology -- radiosurgery -- startingangle -- tomography -- xrays
Biomedical Engineering -- Dissertations, Academic -- UF
Genre:
Biomedical Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Over the past two decades, computed tomography (CT) has become an increasingly common and useful medical imaging technique. CT is a noninvasive imaging modality with three-dimensional volumetric viewing abilities, all in sub-millimeter resolution. Recent national scrutiny on radiation dose from medical exams has spearheaded an initiative to reduce dose in CT. This work concentrates on dose reduction of individual exams through two recently-innovated dose reduction techniques: organ dose modulation (ODM) and tube current modulation (TCM). ODM and TCM tailor the phase and amplitude of x-ray current, respectively, used by the CT scanner during the scan. These techniques are unique because they can be used to achieve patient dose reduction without any appreciable loss in image quality. This work details the development of the tools and methods featuring real-time dosimetry which were used to provide pioneering measurements of ODM or TCM in dose reduction for CT.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Christopher J Tien.
Thesis:
Thesis (Ph.D.)--University of Florida, 2011.
Local:
Adviser: Hintenlang, David E.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0043183:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 OPTIMAL DOSE REDUCTION IN COMPUTED TOMOGRAPHY METHODOLOGIES PREDICTED FROM REAL TIME DOSIMETRY By CHRISTOPHER JASON TIEN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Christopher Jason Tien

PAGE 3

3 To my loving parents and sister for all the support they have given me

PAGE 4

4 ACKNOWLEDGMENTS First and foremost, I would like to express my sincere gratitude to my research advisor and the chair of my supervisory committee, Dr. David Hintenlang, for the opportunity to study at the University of Florida I cannot imagine my graduate studies without his balance of counsel, humor, tolerance, insight and guidance I would also like to extend additional thanks to each of the faculty members serving on my supervisory committee : Drs. Wesley Bolch, Lynn Rill and Bruce Welt. I am appreciative to my entire committee for their constructive comments and advice which has proved invaluable in preparation of this manuscript. I would like to thank Diana Dampier, Donna Seifert, and the rest of the departmental staff for handling all my everyday concerns. I am proud to be co authors with past students : Dan Hyer, Ryan Fisher, and James Winslow I also thank Erik Chell Mario Firpo, and Justin Cantley for their invaluable support in my research with Oraya Therapeutics, which has graciously pr ovided financial support of my research I will always be indebted to my original classmates Alan Cebula and Amir Bahadori for helping shape both my professional and personal life However, a ll my graduate student colleagues i n particular Matt Studenski, P erry Johnson and Matt Hoerner have made my time here undeniably enjoyable. I am forever indebted to my m om for her choice to sacrifice her career in order to p atien tly instill me with the work ethic and values which have made me successful and will propel me into the next stage of my life; and, of course, I thank my dad for working all those long hours to make that financially possible

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 12 ABS TRACT ................................ ................................ ................................ ................... 16 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 17 Computed Tomography ................................ ................................ .......................... 17 Motivation ................................ ................................ ................................ ............... 18 Dose Me asurement Methods ................................ ................................ .................. 19 Anthropomorphic Physical Phantom ................................ ................................ 19 Fiber Optic Coupled Plastic Scintillation Dosimeter System ............................. 21 Dose Reduction Methods ................................ ................................ ........................ 23 Organ Dose Modulation ................................ ................................ ................... 23 Tube Current Modulation ................................ ................................ .................. 24 Objectives of this Research ................................ ................................ .................... 26 2 ANTHROPOMORPHIC PHANTOM C ONSTRUCTION ................................ .......... 32 Background ................................ ................................ ................................ ............. 32 Materials and Methods ................................ ................................ ............................ 34 Phantom Construction Materials ................................ ................................ ...... 35 Bone tissue equivalent substitute (BTES) ................................ .................. 35 Lung tissue equivalent substitute (LTES) ................................ .................. 36 Soft tissue equivalent substitute (STES) ................................ .................... 37 Previous Phantom Construction Methodology ................................ .................. 38 Current Phantom Construction Methodology ................................ .................... 38 Engraving map creation ................................ ................................ ............. 38 Ins ertion of soft tissue equivalent material ................................ ................. 40 Insertion of bone tissue equivalent material ................................ ............... 40 Phantom assembly ................................ ................................ .................... 40 Insertion of lung tissue equivalent material ................................ ................ 41 Results ................................ ................................ ................................ .................... 41 Analysis ................................ ................................ ................................ .................. 42 Discussion ................................ ................................ ................................ .............. 42 3 REAL TIME POINT PLASTIC SCINTILLATION DOSIM ETER SYSTEM FOR USE IN MONITORING AND MEASURING DOSE IN AGE RELATED MACULAR DEGENERATION STEREOTACTIC RADIOSURGERY ...................... 51

PAGE 6

6 Background ................................ ................................ ................................ ............. 51 Materials and Methods ................................ ................................ ............................ 53 IRay TM Stereotactic Radiosurgery Device ................................ ........................ 53 Plastic Scintillation Dosimeter System ................................ ............................. 56 Physical System Benchmarking ................................ ................................ ....... 57 Depth dose curves ................................ ................................ ..................... 57 X ray spectra filtration ................................ ................................ ................ 58 High dose rate and high dose delivered calibration ................................ ... 58 Com putational benchmarking ................................ ................................ .... 59 IRay TM measurements ................................ ................................ ............... 60 Theoretical Optimization of PSD System ................................ .......................... 61 Results ................................ ................................ ................................ .................... 63 Depth Dose Curve in Liquid Water ................................ ................................ ... 63 Depth Dose Curve In Solid Water ................................ ................................ .... 64 Beam Quality ................................ ................................ ................................ .... 65 Ion Chamber Benchmarking ................................ ................................ ............. 65 Computational Benchmarking ................................ ................................ .......... 66 High Dose Rate and High Tot al Dose Performance ................................ ......... 66 IRay TM Measurements ................................ ................................ ...................... 67 PSD current linearity ................................ ................................ .................. 67 PSD saturation ................................ ................................ ........................... 68 PSD depth dose curves ................................ ................................ ............. 68 PSD ramping ................................ ................................ .............................. 69 C omet x ray tube beam stability ................................ ................................ 70 Theoretical Signal to Noise Optimization ................................ ......................... 71 Typical Expos ure ................................ ................................ .............................. 72 Analysis ................................ ................................ ................................ .................. 72 Appropriateness of PSD system ................................ ................................ ....... 72 Ophthalmic Considerations ................................ ................................ .............. 73 Selection of photomultiplier tube ................................ ................................ 75 Discussion ................................ ................................ ................................ .............. 75 4 A METHODOLOGY FOR DIRECT QUANTIFICATION OF OVERRANGING IN HELICAL COMPUTED TOMOGRAPHY ................................ ................................ 94 Background ................................ ................................ ................................ ............. 94 Materials and Methods ................................ ................................ ............................ 96 CT Scanner and Real Time Dosime try System ................................ ................ 96 Dosimeter Positioning ................................ ................................ ...................... 97 Exposure Pattern ................................ ................................ .............................. 99 Clinical Impact ................................ ................................ ................................ .. 99 Results ................................ ................................ ................................ .................. 101 Exposure Pattern ................................ ................................ ............................ 101 Overrangi ng Dependence upon Protocol Parameters ................................ .... 101 Analysis ................................ ................................ ................................ ................ 104 Discussion ................................ ................................ ................................ ............ 110

PAGE 7

7 5 QUANTIFICATION OF STARTING ANGLE DOSE BIASI NG IN HELICAL COMPUTED TOMOGRAPHY ................................ ................................ .............. 117 Background ................................ ................................ ................................ ........... 117 Materials and Methods ................................ ................................ .......................... 118 CT Scanner and CTDI Body Phantom ................................ ........................... 119 Measurement Phantoms ................................ ................................ ................ 119 Original Helical Dose Profile Exp ression ................................ ........................ 121 Extended Helical Dose Profile Expression ................................ ..................... 122 Axial Dose Profile Characterization ................................ ................................ 12 4 Cylindrical Phant om Dose ................................ ................................ .............. 125 Need for Starting Angle Dose Biasing Metric ................................ ................. 126 Anthropomorphic Phantom Dose ................................ ................................ ... 127 Results ................................ ................................ ................................ .................. 129 SADB Measured in Cylindrical Phantom ................................ ........................ 129 Validation of Dose Expression ................................ ................................ ........ 130 SADB Measured in Anthropomorphic Phantom ................................ ............. 131 Analysis ................................ ................................ ................................ ................ 132 Cylindrical Measurements ................................ ................................ .............. 132 Cylindrical Phantom ................................ ................................ ....................... 132 Anthropomorphic Phantom ................................ ................................ ............. 135 Overranging Considerations ................................ ................................ ........... 135 Discussion ................................ ................................ ................................ ............ 136 6 ORGAN DOSE AND INHERENT UNCERTAINTY IN HELI CAL CT DOSIMETRY DUE TO QUASI PERIODIC DOSE DISTRIBUTIONS .................... 147 Background ................................ ................................ ................................ ........... 147 Materials and Methods ................................ ................................ .......................... 149 CT Scanner and Measurement Devices ................................ ......................... 149 Determination of Cumulative Dose Distribution ................................ .............. 149 Axial Point Dose Rate ................................ ................................ .................... 150 Helical Point Dose Rate ................................ ................................ ................. 151 Cumulative Point Dose ................................ ................................ ................... 151 Total Tissue Dose ................................ ................................ .......................... 152 Results ................................ ................................ ................................ .................. 153 Axial Point Dose Rate ................................ ................................ .................... 153 Helical Point Dose Rate ................................ ................................ ................. 153 Cumulative Point Dose ................................ ................................ ................... 154 Total Tissue Dose ................................ ................................ .......................... 156 Analysis ................................ ................................ ................................ ................ 157 D iscussion ................................ ................................ ................................ ............ 160 7 A PRELIMINARY STUDY OF TUBE CURRENT MODULATION USING PHYSICAL MEASUREMENTS AND DICOM HEADER EXTRACTION ................ 173 Background ................................ ................................ ................................ ........... 173

PAGE 8

8 Materials and Methods ................................ ................................ .......................... 176 CTDI Phantom ................................ ................................ ................................ 177 CTDI Phantom with Elliptical Add Ons ................................ ........................... 177 STES Phantom ................................ ................................ ............................... 179 DICOM Information Extraction ................................ ................................ ........ 180 Management of DICOM files ................................ ................................ .... 180 Tag identifier associations ................................ ................................ ........ 181 Z axis TCM Method ................................ ................................ ........................ 182 Experimental Physical Measurements ................................ ........................... 183 Protocol selection ................................ ................................ ..................... 183 Symmetrical scan validation ................................ ................................ ..... 185 Reconstruction method effects ................................ ................................ 185 Results ................................ ................................ ................................ .................. 185 CTDI Phantom ................................ ................................ ................................ 185 CTDI and STES Phantoms with Elliptical Add ons ................................ ......... 186 Reconstruction Methods ................................ ................................ ................. 186 Z Axis Modulation Characterization Using Routine Head Protocol ................. 187 Angular TCM Characterization ................................ ................................ ....... 188 Results Using Both Z Axis and Angular TCM ................................ ................. 188 Reverse Abdominal ................................ ................................ ........................ 188 Analysis ................................ ................................ ................................ ................ 189 Angular TCM Method ................................ ................................ ..................... 190 Machine Reproducibility ................................ ................................ ................. 192 Discussion ................................ ................................ ................................ ............ 192 8 FEASIBILITY AND CONSTRUCTION OF PROTOTYPE CCD BASED DOSIMETRY SYSTEM FOR USE AT DIAGNOSTIC ENERGIES ........................ 217 Background ................................ ................................ ................................ ........... 217 Materials and Methods ................................ ................................ .......................... 219 CCD Theoretical Signal Level ................................ ................................ ........ 219 CCD Physical System Design ................................ ................................ ........ 223 PSD and waveguide ................................ ................................ ................ 223 CCD camera ................................ ................................ ............................ 224 CCD housing ................................ ................................ ............................ 225 Optical c oupling ................................ ................................ ....................... 226 Signal improvement techniques ................................ ............................... 227 Waveguide replacement ................................ ................................ .......... 228 Calibration ................................ ................................ ................................ ...... 228 Baseline Performance Testing Using Portable X ray Unit .............................. 229 Results ................................ ................................ ................................ .................. 229 Waveguide Replacement ................................ ................................ ............... 230 Hou sing Performance ................................ ................................ ..................... 230 Gain Performance ................................ ................................ .......................... 231 Lens Effect ................................ ................................ ................................ ..... 231 Initial Testing ................................ ................................ ................................ .. 231 CCD Radiation Shielding ................................ ................................ ................ 232

PAGE 9

9 Groundbreaking Resultant Images ................................ ................................ 233 MATLAB Automated ROI Analysis ................................ .............................. 233 Analysis ................................ ................................ ................................ ................ 235 Prototype Development ................................ ................................ .................. 237 Computational optimization of SNR ................................ ......................... 237 CCD lens system ................................ ................................ ..................... 237 Novel target adjustment device ................................ ................................ 239 Discussion ................................ ................................ ................................ ............ 240 9 MULTIPIXEL PHOTON COUNTERS ................................ ................................ .... 254 Background ................................ ................................ ................................ ........... 254 Preliminary Results ................................ ................................ ............................... 256 Analysis and Discussion ................................ ................................ ....................... 258 10 CONCLUSIONS ................................ ................................ ................................ ... 263 Results of This Work ................................ ................................ ............................. 263 Opportunities for Future Work and Development ................................ .................. 264 Anthropomorphic Phantom Developments ................................ ..................... 264 Commercial PSD Us age for Radiation Monitoring ................................ .......... 265 Overranging ................................ ................................ ................................ .... 266 Starting Angle and Organ Dose Modulation ................................ ................... 267 Predictive Algorithm for Tube Current Modulation ................................ .......... 268 PSD CCD Prototype Dosimetry System ................................ ......................... 269 Multi Pixel Photon Counters ................................ ................................ ........... 270 Final Thoug hts ................................ ................................ ................................ ...... 271 APPENDIX A FIBER CONSTRUCTION METHODOLOGY ................................ ........................ 272 B TUBE CURRENT MODULATION DICOM SORTER CODE (SORTERV1_2.m) .. 276 C TUBE CURRENT MODULATION DICOM READER CODE (READOUTV1_7.m) 277 D AUTOMATED ROI ANALYSIS FOR CCD ACQUISITIONS CODE (ROI_V6.M) .. 280 LIST OF REFERENCES ................................ ................................ ............................. 284 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 296

PAGE 10

10 LIST OF TABLES Table page 2 1 ICRP weighting factors and number of measurement points .............................. 44 3 1 Calibration factor (CF) of PSD PDD to Ion chamber PDD ................................ .. 77 4 1 O verranging values for various techniques ................................ ...................... 113 5 1 Protocol scan a natom ical landmarks ................................ ................................ 138 5 2 Default beam parameters for different protocols ................................ ............... 138 5 3 Number of measurement points used for each protocol ................................ ... 139 5 4 Total dose as a function of orientation ................................ .............................. 140 5 5 Root mean square dose as a function of orientation ................................ ........ 140 5 6 Average measured organ dose ................................ ................................ ........ 141 5 7 Standard deviation in measured organ ................................ ............................. 142 5 8 Effective doses for different protocols ................................ ............................... 143 6 1 C umulative point dose distribution : lens of eye ................................ ................. 162 6 2 C umulative point dose distributions : thyroid ................................ ..................... 162 6 3 Total tissue dose : lens of the eye ................................ ................................ ..... 163 6 4 Total tissue dose : thyroid ................................ ................................ .................. 163 7 1 Reconstruction mode selection ................................ ................................ ......... 195 7 2 C urrent in c ylinder with no elliptical add ons ................................ .................... 195 7 3 Point dose in axial mode using CTDI head phantom with step phantom .......... 196 7 4 Point dose in helical mode using CTDI head phantom with step phantom with no TCM ................................ ................................ ................................ ............. 196 7 5 Point dose in helical mode using CTDI head phantom with step phantom with TCM ................................ ................................ ................................ .................. 197 7 6 Console estimated helical scan dose for various parameters .......................... 197 7 7 Current in h ead base scan with thin slices ................................ ....................... 198

PAGE 11

11 7 8 Current in head cerebrum scan ................................ ................................ ........ 200 7 9 Current in r outine head base scan ................................ ................................ .... 201 7 10 Default beam parameters ................................ ................................ ................. 203 7 11 Default effective mAs values when using no TCM ................................ ............ 203 7 12 Current in r outine neck scan ................................ ................................ ............. 204 7 13 Current in r outine abdominal scan ................................ ................................ .... 205 7 14 Current in r outine chest thorax scan ................................ ................................ 207 7 15 Current in r outine pelvic scan ................................ ................................ ........... 208 7 16 Current in routine abdominal scan using caudal cranial orientation ................. 210 8 1 Counts as a function of housing and lighting ................................ .................... 242 8 2 Effects of special techniques upo n counts ................................ ........................ 242 8 3 Linearity of pixel intensity as a function of current ................................ ............ 242

PAGE 12

12 LIST OF FIGURES Figure page 1 1 T ube current as a function in a typical study using TCM ................................ .... 31 2 1 ATOM phantom ................................ ................................ ................................ .. 45 2 2 RANDO phantom ................................ ................................ ................................ 45 2 3 All tissue equivalent materials in sample axial slice ................................ ............ 46 2 4 Completed male 50 percentile anthropomorphic physical phantom ................... 46 2 5 ALRADS computational twin ................................ ................................ ............... 47 2 6 CT topogram of completed phantom ................................ ................................ .. 48 2 7 Segmented data imported for sample axial image ................................ .............. 48 2 8 Engraver software input for sample axial image ................................ ................. 49 2 9 Engraved foam for sample axial slice ................................ ................................ 49 2 10 Conformity of sample axial slice around dosimeter ................................ ............ 50 3 1 IRay TM stereotactic radiosurgery device ................................ ............................. 78 3 2 IGuide TM robotic positioning system ................................ ................................ ... 78 3 3 ILens TM eye stabilization system ................................ ................................ ........ 79 3 4 Calibration cyclops ................................ ................................ ............................. 79 3 5 Film used to ensure position of active PSD element ................................ .......... 80 3 6 Solid water eye with PSD taped to back ................................ ............................. 80 3 7 ILens TM attached to metal tip ................................ ................................ .............. 81 3 8 Oraya Jade software screen shot ................................ ................................ ....... 81 3 9 Schematic of FOC PSD ................................ ................................ ...................... 82 3 10 Completed PSD fiber ................................ ................................ .......................... 82 3 11 Completed dosimetry unit picture ................................ ................................ ....... 83 3 12 Schematics of completed dosimetry unit ................................ ............................ 83

PAGE 13

13 3 13 Laser positioning unit mounted in aperture ................................ ......................... 84 3 14 Set up for measuring percent depth dose of PSD in liquid water ....................... 84 3 15 Percent depth dose of PSD in liqu id water ................................ ......................... 85 3 16 Percent depth dose of PSD in solid water ................................ .......................... 85 3 17 Beam quality of SR 115 portable x ray unit measured in aluminum ................... 86 3 18 Percent depth dose of PSD compared with ion chamber ................................ ... 86 3 19 Percent depth dose of MCNPX results compared with ion chamber .................. 87 3 20 High dose performance: calibration factor (CF) as a function of exposure ......... 87 3 21 High dose performance: normalized version ................................ ...................... 88 3 22 High dose rate linearity: counts as a function of exposure ................................ 88 3 23 IRay TM collimator ................................ ................................ ................................ 89 4 1 Schematic of PSD placement ................................ ................................ ........... 114 4 2 Sample responses for bore PSD and dosimeter PSDs ................................ .... 114 4 3 Overranging length as a function of pitch ................................ ......................... 115 4 4 Overranging length as a function of reconstruction slice width ......................... 115 4 5 Overranging length as a function of x ray tube rotation time ............................ 116 5 1 Two views of the STES cylindrical phantom ................................ ..................... 144 5 2 Sample temporal axial response of PSD in cylindrical phantom ....................... 144 5 3 Sample longitudinal response of PSD in cylindrical phantom ........................... 145 5 4 RMS dose as a function o f depth in phantom ................................ ................... 145 5 5 Single slice total dose variation between PSD orientations .............................. 146 5 6 SADB as a function of different effective widths ................................ ............... 1 46 6 1 Normalized axial dose rate ................................ ................................ ............... 164 6 2 Normalized helical dose rate ................................ ................................ ............ 164 6 3. Lens of the eye, pitch of 1, cumulative point dose ................................ ............ 165

PAGE 14

14 6 4 Thyroid, pitch of 1, cumulative point dose ................................ ........................ 166 6 5 Pitch of 1.5, lens of the eye, cumulative point dose ................................ .......... 167 6 6 Pitch of 1.5, thyroid, cumulative point dose ................................ ...................... 168 6 7 Normalized total tissue dose, lens of the eye, pitch of 1 ................................ ... 169 6 8. Normalized total tissue dose, thyroid, pitch of 1 ................................ ............... 170 6 9 Pitch of 1.5, Lens of the eye, normalized total tissue dose ............................... 171 6 10 Pitch of 1.5, Thyroid, normalized total tissue dose ................................ ........... 172 7 1 CTDI phantom with and without Lucite PSD adapter ................................ ........ 211 7 2 Cylindrical elliptical add on inserts family ................................ ......................... 211 7 3 Construction using plywood mold ................................ ................................ ..... 212 7 4 CTDI phantom within step phantom in CT scanner ................................ .......... 212 7 5 Topogram of CTDI phantom within step phantom ................................ ............ 213 7 6 STES CTDI head phantom ................................ ................................ ............... 213 7 7 A verage slice current at 160 mAs R ................................ ................................ 214 7 8 A verage slice current at 200 mAs R ................................ ................................ 214 7 9 Average slice current at 260 mAs R ................................ ................................ 215 7 10 Typical phantom orientation, 10 measurements with abdominal protocol ........ 215 7 11 Inverted phantom orientation, 10 measurements with abdominal protocol ....... 216 8 1 Gain linearity ................................ ................................ ................................ ..... 243 8 2 Size of fiber entry ho le, coin used for reference ................................ ............... 243 8 3 Completed CCD system housing and schematic ................................ .............. 244 8 4 Different types of lenses ................................ ................................ ................... 245 8 5 Effect on edges by aspheric lens ................................ ................................ ...... 245 8 6 Lens mounted in optical adapter threaded into lens tube ................................ 245 8 7 Waveguides mounted on v clamp ................................ ................................ .... 246

PAGE 15

15 8 8 Silica waveguide transmission vs wavelength ................................ .................. 246 8 9 Resultant images with different fiber to lens distances ................................ ..... 247 8 10 Completed CCD housing ready for x ray source benchmarking ....................... 247 8 11 Lens effect on focusing ................................ ................................ ..................... 248 8 12 Streaking artifact ................................ ................................ ............................... 248 8 13 Final CCD image ................................ ................................ .............................. 249 8 14 Pixel intensity linearity as a function of mAs ................................ ..................... 249 8 15 Stability of pixel intensity from multiple trials ................................ .................... 250 8 16 Pixel intensity as a function of mA s ................................ ................................ .. 250 8 17 Calibration between mAs and pixel intensity ................................ .................... 251 8 18 Mean pixel intensity vs ROI sampling area ................................ ....................... 251 8 19 CCD double lens set up and a schematic ................................ ......................... 252 8 20 Novel target adjustment device ................................ ................................ ........ 253 9 1 MPPC with fiber with SMA and firewire connections ................................ ........ 260 9 2 Typical MPPC response ................................ ................................ ................... 261 9 3 MPPC counts vs mAs ................................ ................................ ....................... 261 9 4 PMT and MPPC response ................................ ................................ ................ 262 A 1 E limination of jagged edges of fiber ................................ ................................ .. 275

PAGE 16

16 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy OPTIMAL DOSE REDUCTION IN COMPUTED TOMOGRAPHY METHODOLOGIES PREDICTED FROM REAL T IME DOSIMETRY By Christopher Jason Tien August 2011 Chair: David Hintenlang Major: Biomedical Engineering Over the past two decades, computed tomography (CT) has become an increasingly common and useful medical imaging technique. CT is a noninvasive imaging modality with three dimensional volumetric viewing abilities, all in sub millimeter resolution. Recent n ational scrutiny on radiation dose from medical exams has spearheaded a n initiative to reduce dose in CT This work concentrates on dos e reduction of individual exams through t wo recently innovated d ose reduction techniques : organ dose modulation (ODM) and tube current modulation (TCM). ODM and TCM tailor the phase and amplitude of x ray current, respectively, used by the CT scanner during the scan. These techniques are unique because they can be used to achieve patient dose reduction without any appreciabl e loss in image quality. This work details the development of the tools and methods featuring r eal time dosimetry which were used to provide pioneering measurements of ODM or TCM in dose reduction for CT.

PAGE 17

17 CHAPTER 1 INTRODUCTION C omputed T omography X ray c omputed t omography ( CT ) is a medical imaging technology which uses computer processing in order to obtain three dimensional volumetric viewing abilities, all reconstructed with sub millimeter resolution of the insides of an object from a large number of pr ojections of two dimensional images taken around an axis of rotation. Tomography is the use of imaging sections through the use of any type of penetrating wave. In the case of CT these penetrating waves are x rays. While CT technology has been available since the 1970 s, it has enjoyed a remarkable exponential climb in advancements over the last two decades due primarily to the advances in newer and faster computer systems and software Historically, the reconstructe d images were created normal to the axis of rotation because scans were performed in an axial manner. Now, with the advance to helical scanning, volumetric images can be reconstructed in any plane desired. Previously, CT scanners used only a single row of detectors, recently innovated CT scanners use as many as 320 detector rows for imaging. While the original prototype CT scanner developed by Hounsfield required five minutes for full rotation s current x ray tube s can perform sub second full rotation s arou nd the gantry. The 300 ms rotation time imparts the ability to patient anatomical information in real time For example, the x ray tube rotation time is so fast that the heart beat can be imaged in its different phases of expansion and contraction In addi tion to fast acquisitions C T offers extremely high contrast in its images In fact, its primary unit is the Hounsfield Unit (HU) which describes change in linear attenuation coefficients as small as 0.1%.

PAGE 18

18 Motivation While CT has rapidly been incorporated into clinical diagnostic radiolog y due to its high resolution and fast scan times, it does have the disadvantage of delivering moderate to high doses, with respect to diagnostic radiology. Recently, the increased radiation dosages have been scrutinized by the media and brought to the attention of the public For example, in 2009, the N ational C ouncil on R adiation P rotection and Measurements specified computed tomography (CT) scans as the largest contribut or of manmade radiation exposure in the United States. 1 The NCRP stat ed that medical procedures in 2006, Americans were exposed to more than seven times as much ionizing radiation compared with 1 980 due primarily to the increase in utilization of CT and nuclear medicine. CT comprised 15% of total exposure in 1982 and 48% of total exposure in 2006. The annual number of CT examinations has increased from 3.6 million in 1980 to 72 million in 2007. 2 T he New England Journal of Medicine attributed 2% of all cancers in the United States to CT and concluded that too many scans were being performed. 3 Investigators estimated each chest or abdominal CT scan on the order of 15 25 mGy but each of these scans wa s performed as a part of a series This in turn is usually compounded because multiple series are performed over treatment. While the NCRP did not recommend any specific actions, the investigators from the New England Journal of Medicine article recommende d a three part approach: reducing the CT dose in individual exams, replacing CT use with alternate options such as ultrasound or MRI, and reducing the number of CT exams that are prescribed. For more specific directives, McCollough et al. and Kalra et al. have published articles which review methods for dose reduction and management. 4, 5

PAGE 19

19 Today, while more CT scans are being performed, the average dose per scan has fallen in each of the past decades. The A merican A ssociation of P hysicists in M edicine has acknowledged risks posed by CT, but its viewpoint is that the benefits of an appropriately ordered CT exam far outweigh those risks 6, 7 Quantification of radiation exposure from CT scans should be accurate in order to accurately assess the collective risk a nd to provide patient specific organ doses for use in either retrospective epidemiologic or prospective risk estimation studies. Dose Measurement Methods Anthropomorphic Physical P hantom Properly quantifying ra diation dose from CT has always been a challe nge and several methods have been developed for this purpose Presently, t he most prevalent method for clinical dose measurement and image quality measurements is the Computed T omography Dose Index (CTDI) which has undergone various refinements in order to maintain its relevance as CT has developed. 6 Currently volume CTDI ( CTDI vol ) is required to be displayed both before and after scan s in CT scanners built after 2002. 8 At the time, CTDI vol was the newest version of CTDI; CTDI vol represent s the average absorbed dose, along the z axis, from a series of contiguous irradiations D ose length product (D LP ) wa s also developed at that time, and is calculated as the product of CTDI vol and the length of scan Recently, the AAPM Task Group 23 released a new methodology for evaluating the radiation dose which is based upon equilibrium dose (D eq ), a new metric which moves beyond CTDI vol and DLP. 9 The concept of CTDI was developed to provide a standardized method to compare radiation output levels between different CT scanners using reference phantom s While the CTDI metric has evolved to adapt from a single detector axial scan to a multi

PAGE 20

20 detector helical scan, CTDI has always been measured in reference phantoms. These cylindrical phantoms are constructed to serve as generalized approximations of the human body and head with diameters of 32 cm and 16 cm, respectively. C ylindrical phantom s used in order to represent the human body lead to many problems in realistic measurement of dose. Namely, a cylinder does not approximate the shape of a human body Additionally, the uniform Lucite composition d oes not mimic the varying attenuation of a hu man body Furthermore, CTDI has become outdated because its measurement is done using a detector which is only 100 mm long as mentioned previously, some modern scanners have scan lengths which are 320 mm. Even with smaller scan lengths, 100 mm has not pr oven to be adequate to account for scatter tails. Despite those concerns, CTDI still remains popular due to i ts application as more of an image quality test metric rather than a dose measurement 6 Remember, CTDI was designed to provide standardized method to compare radiation output levels between different CT scanners using reference phantoms. Today, i n order to measure realistic organ dose and subsequently calculate effective dose from CT procedures, researchers have relied upon segmented patient models or an thropomorphic physical phantoms instead of CTDI phantoms. Measurements should be conducted in anthropomorphic physical phantoms which possess realistic radiological attenuation properties. One such phantom developed at UF is an anthropomorphic 50 percentile reference male segmented from the corresponding CT data used in the computational model used by the University of Florida Adv anced Laboratory for Radiation Dosimetry Studies (ALRADS) laboratory,

PAGE 21

21 which uses three different materials: soft tissue equivalent substitute (STES), bone tissue equivalent substitute (BTES), and lung tissue equivalent (LTES). 10 The physical phantom has a high axial resolution and readily accessible point dosimetry measurements sites. Fiber O ptic C oupled P lastic S cintillation D osimeter S ystem Plastic scintillation dosimeters (PSDs) have become a popular and economical measurement tool because of their unique combination of fast response, temperature independence, small volume, water equivalence along with energy dose and dose rate linearity. Fast scintillating materials typically emit photons in 2 4 nanoseconds while slow scintillating materials emit photons in times greater than 200 nanoseconds. 11 In other words, in most applications, the PSD response is not a limiting factor. The small volume is important in order to allow physical accommodation of the measurement device within the phantom. Additionally, by only occupying a small volume, this gives the measurement device high spatial resolution. The water equivalence is important because it mi mics tissue and does not perturb measurements. F iber optic coupling (FOC) can be used for connecting the PSD to the waveguide. This design allow s for a large number of scintillation photons to reach the photodetector while maintaining a small volume. Whil e PSDs do suffer from stem effect from Cerenkov radiation, Beddar et al. has measured Cerenkov threshold to be around 178 keV, much lower than the 100 keV peak expected. 12, 13 For this application, PSDs are specifically better than current diagnostic radiology dosimeters such as ion chambers becau se PSDs impart a higher spatial resolution due to their small volume in comparison with an ion chamber this is about an order of magnitude. PSDs are also preferred over thermoluminescent

PAGE 22

22 dosimeters (TLDs), or optically stimulated luminescent (OSLs) dosim eters because recent analyses have uncovered problems with other detectors: TLDs have an angular dependence of up to 20%; OSLs have shown degradation with repeated exposures. 14, 15 Another alternative, metal oxide field effect transistors (MOSFETs) were also considered but have shown an incredibly larg e variation in measurements of reproducibility, with the differences in some cases reaching 15% 30%. 16 Furthermore, the metallic high atomic number used for MOSFETs will lead to image artifacts. id Two distinct photodetectors will be used in this investigati on each with its own advantages and disadvantages. The majority of this work utilizes photomultiplier tubes (PMTs). Using this type of photodetector allows measurements to be taken in real time, with a high sampling rate. This design was innovated by Hyer et al. who fully characterized a separate system at diagnostic energies. 17 The other type of photodetector which has been used successfully is based upon a charge coupled device (CCD). A large source of noise in PSD measurements come from the stem effect in this case, scintillating of the waveguide itself. Furthermore, for applications of radiation therapy, Cerenkov radiation is also generated. A CCD PSD system can use simple filtration or chromatic removal to subtract a Cerenkov spectrum or other wavelength dependent features from the measured light. CCDs a lso hold a distinct advantage in cost. Essentially each additional PSD channel for a PMT based design requires another PMT (~$ 40 00), while a single CCD (~$8000) could conceivably handle 50 PSD channels. A CCD PSD system for diagnostic radiology applicatio ns will suffer from much lower signal. While CCD PSD systems are currently used in radiotherapy systems,

PAGE 23

23 exposure times are on the order of minutes and energies are on the order of megavoltages (MV). On the other hand, diagnostic radiology applications use kilovoltage (kV) energies w ith millisecond exposure times. Specifically, CT uses energies around 80 140 keV which is below the Cerenkov threshold of about 178 keV. 18 Dose R eduction M ethods The approach of this researc h is to investigate the maximum radiation dose reduction for CT examination through two recently innovated techniques: organ dose modulation (ODM) and tube current modulation (TCM). This research uses exclusively results obtained from a Siemens SOMATOM Sen sation 16 CT scanner. Note that TCM is also known as automatic tube current modulation (ATCM). B oth ODM and TCM offer refreshingly innovativ e approaches to dose reductions. T he dose reductions from ODM will be most clinically significant in small periphera l organs such as the eye lens. However, ODM is still extremely relevant because it may represent a previously overlooked source of uncertain ty in many measurements, notably computational models. On the other hand, dose reductions from TCM techniques will b e most clinically significant for low attenuation regions such as the lungs. ODM and TCM both offer the unique advantage of dose reduction with no distinguishable loss in image quality. Organ Dose Modulation Dose distribution at peripheral locations such as the lens of the eye or the breast has been demonstrated to be periodic as a function of beam collimation and pitch. 19, 20 The phase of th is dose distribution is determined primarily by the starting angle at which the x ray beam turns on and off. Organ dose modulation (ODM) involves manipulation of the starting angle of the x ray tube in o rder to cause a shift in the dose distribution.

PAGE 24

24 This could be used to match minimums in dose distribution with radiosensitive organs such as the lens of the eye or the breast. 4, 19 25 In order to accomplish ac curate phase shifting, it is important to account for overranging. 26 Overr anging volumes are the volumes adjacent to the clinical volume of interest (VOI) which must be collected in a helical CT for proper reconstruction and the phase shift must also account for these extra volumes because the starting angle for the scan is not necessarily the same as that for the VOI. ODM does not change any parameters of the scan, thus different scans should produce the same reconstruction. In fact, van Straten et al. discovered that radiologists could not distinguish between two scans using tw o different x ray tube starting angle. 27 T ube Current Modulation T ube current modulation (TCM) was developed by manufacturers and is now in use clinically in order to reduce CT patient dose in individual exams. TCM adjust s the x ray tube current during a scan in order to maintain const a nt photon fluence at the detector. If all other variables are held constant, a reduction in tube current leads to a reduction in patient dose, but an increase in quantum noise or mottle in the reconstructed image. Previously, tube current and voltage were chosen based upon patient size patient weight and image quality requirements. The tube current and voltage were fixed for the entire scan However, due to varying body circumference and tissue attenuation, the fixed techniques lead to variable attenuation though the body which lead to variable image quality Images with too much noise obscure low contrast lesions or tumors that would normally be visible in less noisy images and could lead to misdiagnoses or the need to rescan the patient, exposing them to unnecessary radiation. Furthermore, fixed

PAGE 25

25 techniques frequently suffer from photon starvation and subsequently artifacts are manifested during reconstruction. Previously, in order to avoid photon starvation artifacts, the tube current for the entire scan area was increased as opposed to simply increasing tube current in a small region as done with TCM In other words, TCM attempts to tailor the current to the region of the body. For example, the tube current is increased when scanning along areas of high attenuation such as the shoulder or hips. Simi l arly the tube current is decreased in areas of low attenuation such as the lungs or extremities. An example of tube current and attenuation plotted as a function of longitudinal position is shown in Figure 1 1 The adjustment is highly manufacturer dependent, but detector feedback is usually within one fifth of a second. TCM help s to reduce patient doses while ensuring uniform image quality which is based upon the number of photons incident upon the detector. Current CTDI phantoms do not challenge the TCM systems in modern scanners. As a result, the same CTDIvol value is currently used to predict patient doses for both fixed tube current and TCM scans, though the two exams can deliver dramatically different doses. There are currently two major TCM strategi es employed by manufacturers : angular and z axis modulation This research investigates the use of Siemens technology. Recently Siemens using angular modulating tube current based up on attenuation va lue measured from the previous 180 degree rotation. Other vendors such as GE and Philips, were not investigated in this research. However, each has their own methodology for TCM. For example, GE Medical Systems introduced its SmartScan system in 1994, whic h used

PAGE 26

26 a ngular modulation to modulate tube current in the x y plane within a single rotation of the tube. The attenuation value is determined from an AP and lateral radiographic projection images. Philips Medical Systems introduced its version of angular TCM called Dose Right Dose Modulation which modulates tube current within a single tube rotation according to the square root of the attenuation measured during the previous rotation. This modulation technique is based on image noise being inversely relate d to the square root of the number of photons captured. Z axis modulation is t he second approach to TCM In this method, the tube current is modulated for each slice as determined from a scout radiograph. Therefore, unlike angular modulation, the tube curr ent is maintained within a single rotation Z axis modulation keeps image quality uniform in each slice of the exam. Objectives of this Research Both ODM and TCM are especially promising because they reduce patient dose without any appreciable loss in image quality. 4, 5 Real time dosimetry using plastic scintillation dosimeters (PSDs) will be used to characterize the performance and behavior of both ODM and TCM in a realistic environment, namely an anthropomorphic phantom. The results can be applied in order to select t he optimal combination of parameters to produce maximal dose reduction. The main goal of this research was to develop the tools and methods necessary to accurately measure the effects of ODM and TCM upon dose reduction in helical CT. This research was brok en down into a series of smaller projects, each of which is described below. The first project was to fabricate a male 50 percentile reference anthropomorphic phantom in order to provide a realistic measurement medium. This phantom was based

PAGE 27

27 upon the Inter representing the 50 th percentile adult male. 28 This phantom was first constructed as a computational model by the University of Florida Advanced Laboratory for Radiation Studies (ALRADS) laboratory. Anatomical data for construction of the physical phantom are directly drawn from the computational phantom. Thus the physical phantom has a computational twin for benchmarking and direct comparisons. In order t o provide realistic attributes from a dosimetric standpoint, three different tissue equivalent materials were const ructed: bone, lung, and soft tissue. This phantom was constructed in 5 mm axial slices from the gonads to the top of the head. The arms and legs were constructed in 10 mm axial slices which are detachable The second project was to construct a dosimetry sy stem which utilizes a small volume dosimeter which is preferably water equivalent in order to mimic tissue. This dosimeter must be capable of measurements of absorbed dose in the anthropomorphic phantom described below at diagnostic level energies. Plastic scintillation dosimeters (PSDs) were chosen due to their unique combination of low cost, temperature independence, small volume, water equivalence along with energy dose and dose rate linearity. A fiber optic coupling (FOC) was used for connecting th e PSD to the waveguide. This was done by hand for each measurement fiber and is detailed in Appendix A of this work. The PSD was 0.5 mm in diameter and 2 mm in length. By using photomultiplier tubes (PMTs), the measurements were taken in real time with bin s as small as 10 ms. Dosimeters with a similar design by Hyer et al. were fully chara cterized at diagnostic energies. 17 This dosimetry system was used as a device for real IRay TM stereotactic

PAGE 28

28 radiosurgery (SRS) device which uses diagnostic energy photons for age related macular degeneration. Th e PMT PSD system was fully characterized for application to SRS. For example, depth dose curves and calibration factors were measured. High dose rate and high dose performance were tested. A theoretical optimization of t he PMT PSD s ystem and compared with the current design. The third project was to measure overranging Proper r econstruction requires some degree of overranging, thus this study introduce d an innovative and accurate method to directly characterize overrangi ng length using real time dosimetry. This method is more accurate and faster than other methods currently used. The overranging length is specifically measured as a function of different pitches, rotation time, beam collimation, and reconstruction slice wi dth I ncluding overranging lengths in the prediction of dose length products, or in Monte Carlo calculations of effective dose will reduce underestimates of dose. Additionally, overranging is an important aspect in starting angle calculations ( discussed in specific aim 4). The fourth project was to measure the effect of x ray tube starting angle upon measurement uncertainty and its impact upon clinical dosimetry. Direct physical measurements of this uncertainty are not possible given the uncontrollable x ra y tube starting angle. In this project a mathematical dose expression on helical CT scan dose by Dixon et al. was further developed to calculate the helical dose for any given starting angle. 23 Th is method can be used to decompose the helical response into axial and longitudinal responses which can reconstruct the uncertainty for any arbitrary starting angle. Starting angle dose bias (SADB) is a new metric which is introduced in order to express the observed difference in dose meas urements as a direct result of different x

PAGE 29

29 ray tube starting angles. The SADB calculation is first performed with a cylindrical phantom and applied to an anthropomorphic phantom. The fifth project was to determine dose savings possible from manipulation of x ray tube starting angle. This method of dose savings is also known as organ dose modulation (ODM). S ignificant variation of organ and effective doses at surfaces is due to overlaps and gaps in the helical exposure. As mentioned in Specific Aim 4, t hese potential dose reductions have not been realized because the x ray tube starting angle is both unpredictable and uncontrollable in popular CT scanners. Specific Aim 4 focused upon the uncertainty imparted. T his project continued analysis of x ray tube star ting angle and calculates potential dose savings for two important peripheral organs, namely the lens of the eye and the thyroid. This method incorporated pitch, actual beam width, beam divergence, distance from isocenter and anatomical volume distributio n. The sixth project was to determine the dose savings from tube current modulation techniques (TCM) A custom MATLAB program was created in order to extract DICOM data of tube current in each slice for reconstructed images produced by the scanner. M easu rements were calibrated to a CTDI phantom with a PSD embedded in a Lucite plug. Next, measurements were taken with a CTDI phantom surrounded by two adjacent uniform tissue equivalent add ons with three segments possessing different lateral diameters The v arying attenuation in the lateral directions challenged the TCM to appropriately match current The TCM was measured in real time for a variety of standard manufacturer protocols. This information was subsequently used to plot tube current as a function of position.

PAGE 30

30 The seventh and eighth project s were to design new prototype PSD system s. This was directly largely at a CCD PSD system. The seventh project was targeted at using a CCD PSD to i ncrease parallel capabilities of a dosimetry system. While the CCD w as being manufactured, more than 3 0 scintillation fibers were prepared. CCD PSD systems are currently used in megavoltage (MV) radiotherapy systems where exposure times are on the order of minutes. On the other hand, t he prototype is designed to demonstrat e the feasibility of a PSD system for kilovoltage (kV) energies with millisecond exposure times. The fabrication process of the system is outlined. A reasonable response is established. Also, s ince the signal and response is expected to be orders of magnit ude smaller, a calculation of system parameters is performed in a ma nner similar to Specific Aim 3. Overall, the new prototype system was aimed at a CCD PSD system, however the eighth project examines a novel multipixel photon counter PSD system. In additi on to the physical measurements originally proposed, Chapter 3 also includes an analysis of the optimization of the PMT based PSD system Chapter 6 discusses direct dose reduction, which is a nother aspect of ODM. Chapter 8 addresses in great detail the fab rication and feasibility of the CCD based PSD system along with preliminary results. Chapter 9 examines a multipixel photon counter PSD system. Chapter 7 will be dramatically expanded in the next few years by ALRADS The eventual goal is to design a predictive algorithm which will computationally simulate a scout scan and predict the current used by the scanner. The group plans to use the physical TCM data obtained in this work in order to provide input into an attenuation length based progra m

PAGE 31

31 Figure 1 1 The tube current and attenuation as a function of longitudinal position in a typical study using TCM ( Photo courtesy of 32 : 729 734 ( 2002 ) Page 733, Figure 5 )

PAGE 32

32 CHAPTER 2 ANTHROPOMORPHIC PHAN TOM CONSTRUCTION Background radiological properties in order to measure a realistic biological dose delivered during diagnostic imaging and radiotherapy procedures. Recently, a push towards precise measurements of dose delivered during scans has researchers measuring average computed tomography (CT) organ doses as well as effective doses. 29 Computational approaches have also been undertaken, but require knowledge of the exact photon energy spectrum or irradiation geometry which is often propri etary information to each specific vendor and changes among models. Advanced techniques which employ real time electronic feedback such as automatic tube current modulation in CT and automatic brightness control (A B C) in fluoroscopy are particularly diffic ult to model in computational simulations. The most popular anthropomorphic phantom s are the RANDO (The Phantom Laboratory, Salem, NY) and ATOM phantoms (Computerized Imaging Reference Systems, Inc, Norfolk, VA). 30, 31 Both the RANDO and ATOM phantoms have axial resolution of 25 mm. 30, 31 The human anatomy is sorted using attenuation properties into three different regimes : soft tissue, bone, and lung. 30, 31 These p hantoms are shown in Figure 2 1 and 2 2 In order to overcome the often prohibitively high cost of the R A NDO and ATOM phantoms, in house s oft tissue equivalent substitute (STES), lung tissue equivalent substitute (LTES), and bone tissue equivalent substitute (BTES) have been developed at the University of Florida (UF) The UF materials are the next generation of materials

PAGE 33

33 originally developed by White et al. and Jones et al . 32 35 These UF materials are low cost tissue radiological equivalent materials, prepared in a standard labora tory which have been incorporated into a family of adult anthropomorphic phantoms. While the RANDO and ATOM phantom utilize a 25 mm slice thickness, the UF phantoms utilize a 5 mm slice thickness, which provides better resolution in the cranial caudal di rection. 30, 31 While t he full body data set includes over 353 axial slices the UF phantoms were constr ucted without legs because they lack radiosensitive organs This simplified fabrication to 193 axial slices which ranged from the crown of the head to mid thigh. All int ernal organs in the phantoms were included in the phantom as STES. In order to measure organ dose the dosimeter placement is thus based directly upon position. The organ s from the segmented CT data have been translated onto each slice. From the previous generation, the BTES remains the same and is based upon an epoxy resin which forms a har d thermoset polymer, as described by Jones et al. 32, 33 The STES and LTES are based up on a new urethane mixture which forms a pliable rubber material The new STES and LTES are an improvement in terms of construct ion ability and phantom durability Furthermore, the rubber composition of STES allows much better accomm odation of real time dosimeters than the hard polymer material used previously. A completed slice showing BTES, LTES, and STES is shown in Figure 2 3 phantom created from a CT anthropomo r phic reference library developed by Lee et al. 36 The physical phantom can be used to direc tly benchmark the results from the computational twin. The physical phantom is shown in Figure 2 4 while its

PAGE 34

34 computatio nal twin is shown in Figure 2 5 T he idea has been proposed by Winslow et al. in order to determine point to organ dose scaling factors a llowing the calculation of average organ doses from simple point organ dose measurements m ade in the physical phantom. 10 Figure 2 6 shows a CT topogram of the physical phantom. A Vac Fix reusable patient po sitioning system for radiation therapy (S&S Par Scientific, Houston, TX) was used to secure the phantom in reproducible position. Both t he Vac Fix bag and medical tape were utilized to to minimize the air gaps, which are characterized by the horizontal dark lines located within the phantom in this figure. Measurements have shown less than 2% difference when these gaps are present. The phantom is shown with the individual point dose measurement locations, marked with metal pellets. Table 2 1 shows the ICRP weighting and number of available measurement points for each organ. Materials and Methods This section of the dissertation reports on the materials and methods used to construct anthropomorphic phantoms for use in dosimetry studies, in an improv ement on methods and materials previously described by Jones et al. 32, 33 This methodology was previously published as part of the PhD dissertation by Winslow. 37 Since then, this has been published as an article in the Journal of Clinical and Applied Medical Physics by Winslow et al. 10 The fabrication was streamlined with a n engraving machine which is used to create molds for the phantom slices from bitmap images computationally generated from the original reference models produced by Lee et al. 36, 38 41 Lee et al. used n onuniform rational B spline (NURBS) surfaces in order to create a reference phantom. 36, 38 41 Information regarding the new tissue equivalent materials as well as a detailed summary of the construction process and completed phantom are included.

PAGE 35

35 Phantom Construction Materials B TES LTES, and S TES were all designed in order to mimic the radiological properties of humans. Previous iterations by Jones et al. were designed for pediatric cases. 32, 33 In order to construct this adult male phantoms 253 axial slices were required per phantom so it was also crucial that the process become streamline the process. The metric used to determine the accuracy of STES, LTES, and BTES was density and attenuation. Final results are discussed in the Results section below. Attenuation was evaluated by measuring multiple samples of varying thickness in a narrow beam geometry generated by clinical radiographic unit. Additionally, Hounsfield Unit (HU) values were measure d using a Siemens SOMATOM Sensation 16 helical MDCT scanner operated at a tube voltage of 120 kVp. Average HU was determined from the selected regions of interest (ROI) usi ng areas of approximately 10 cm 2 Density was determined by first finding t he dry m ass m dry of samples of B TES, LTES, and S TES were each measured on a scale to a precision of 0.001 gram, then the wet mass m wet of each w as found in a beaker of de ionized water. Using the density of the de ionized water, water the density was measure d. Equation 2 1 uses principle was used in order to determine density. (2 1) Bone tissue equivalent s ubstitute (BTES) For the UF series of anthropomorphic phantoms, bone tissue equivalent substitute (BTES) was used to represent a homogenous mixture of cortical and trabeulcar

PAGE 36

36 spongiosa bone trabeculae and bone marrow. The BTES composition were matched to those defined by Cristy et al. in the Oak Ridge National La series 42 The m ass density, mass attenuation coefficient, and mass energy absorption coefficient were matched to the diagnostic photon en ergy range of 80 120 kVp 42 Jones et al. showed that the final effective atomic number for the BTES is 8.80 in comparison with the ORNL reference tissue effective atomic number of 8 .59 32 F urthermore, t he BTES created by Jones et al. had a maximum deviation from ORNL reference values by only a few percent for both mass attenuation coefficient and mass energy absorption coefficient. 32 Jones et al. prescribes a BTES mixture of 36.4% Araldite GY6010 and 14.6% Jeffamine T 403 (Huntsman Co rp., Woodlands, TX), 25.5% s ilicon dioxide and 23.5% c alcium carbonate (Fisher Scientific, Hanover Park, IL ). 32 Final results are discussed in the Results section below. Lung tissue equivalent s ubstitute (LTES) For th e UF series of anthropomorphic phantoms, the lung tissue equivalent substitute (LTES) was used to represent the lung and its constituent structures. A density of 0.33 g/cm 3 was chosen for this generation of anthropomorphic phantoms in order to represent th e density of a full inspiration, as defined by the ICRU. 43 Full inspiration was the most realistic choice because patients are typically hold ing their breath during the exposure. The previous generation of UF phantoms, by Jones et al. continued the fabrication structure specified by White et a l. which relied upon a foaming agent for the LTES. 32 35 The reliance up on foam created problems with uniformity and reproducibility. Hence, this next generation of UF phantoms relied upon an epoxy resin which was

PAGE 37

37 combined with polystyrene micro beads (P oly fil Fairfield Processing, Danbury, CT) with an electric mixer in order to ensure homogeneity. Specifically, the LTES was constructed in a mixture of STES as described in the next section, mixed with polystyrene micro beads in a 1:1 0 ratio by weight. T he small deviations in micro bead distribu tion actually impart a range of densities which spans various levels of inspiration. Soft tissue equivalent s ubstitute ( STES ) For the UF series of anthropomorphic phantoms, soft tissue equivalent substitute (STES) was used to represent homogenous soft tis sue as well as all organs, connective tissue, and adipose tissue. In order to measure organ dose, the dosimeters will be placed based upon position. The organs from the segmented CT data set were eventually demarcated upon the final phantom using templates made from the axial slices of the CT data set. Note only the ICRP 103 organs used in the calculation of effective dose were eventually drawn on the physical phantoms. 44 Human soft tissue in the diagnostic energy range of 80 120 kVp has a density of 1.04 g/cm 3 11 The x ray attenuation co efficient was based upon soft tissue composition as defined by the ICRU reference mentioned for the LTES fabrication 43 A commercially available, durable, readily available two part urethane compound (PMC 121/30 Dry, Smooth On, Easton, PA) was found to be easy to work with and did not suffer from problems f requently encountered by Jones et al. with epoxy resin based STES. 32, 33 The two part urethane compound was combined with 2.8 percent by weight of calcium carbonate powder (Calcium Carbonate, Fisher Scientific, Hanover Park, IL) and mixed with an electric mixer to ensure homogeneity.

PAGE 38

38 Previous Phantom Construction Methodology The previous generation of UF phantoms, as described by Jones et al. involved preparing epoxy based soft tissue material in a vacuum chamber to eliminate air bubbles, pouring the material into a square mold, milling out the outer slice contour as well as appropriate voids for bone and lung tissue equivalent material, and finally filling these voids with bone or lung tissue equivalent material as required. 32, 33 This method was quite tedious, but only a few number of slices needed to be done because the phantom used thicker slices and modeled a pediatric human. In this generation of UF phantoms, the BTES, LTES, and STES materials were specifically fabricated in order to avoid problems stemming from air bubbles. Therefore, a vacuum chamber was not required. Secondly, i nstead of manually milling the outer slice contours and appropriate voids, an automated engraving system ( 1624 Vision Engraving and Routing Systems, Phoenix, AZ) was employed. The engraving map was imported from Lee et al. based male anthropomorphic r eference phantom. 36, 38 Current Phantom Construction Methodology Engraving map creation P hantoms constructed were based on hybrid NURBS based computational phantoms of a 50th percentile adult male and female devel oped at the University of Florida by Lee et al. 36, 38 These phantoms the y are originally obtained from human tomographic data, but were subsequently modified to match anthropometric dimensions and organ masses as defined by the International Commission on Radiological Protection (ICRP) publication 8960 reference data for a 50t h percentile as described by Lee et al. 36, 38 40 The original human

PAGE 39

39 tomographic data for the male hybrid phantom came from a 36 year old Korean adult male (176 cm height, 73 kg weight) 39, 40 Information about the female hybrid phantom was omitted because only the male hybrid phantom was utilized for the measurements tak en in Chapters 3 through 8 ; more information can be found in Lee et al. 39, 40 All scans were performed at full inspiration with an in plane matrix size of 512x512 pixels with o rgan segmentation performed manually with t he guidance of a radiologist. 39, 40 An axial image showing segmented organs from the computational phantom is shown in Figure 2 7. S egmented data were passed through a speckle filter and converted to a vectorized bitmap format in o rder to be recognized b y the automatic engraver. A sample imported file is shown in Figure 2 8 All parts of the anatomy were sorted into BTES, LTES, and STES single pixel value s. Smaller diameter engraving bits were used for the head, while larger diameter engraving bit s w ere used for the body The engraving map was made in foam blanks which were fastened to the engraving table and with depths resulting in 4. 5 mm thick axial slices 0.5 mm was allocated for the adhesive between slices, which will be discussed later. In order to smooth the edges of each contour of the engraving map, an engraving path was first area to be cut in comparison with the faster rate fill the interior of the engraving map. E ach body axial slice mold took around ten minutes to complete, while head axial slice molds, in general, required slightly less time. Final j ob

PAGE 40

40 time for axial slice mold was recorded in order to determine the appropriate amount of material to prepare to fill the mold. A finished tis sue mold is shown in Figure 2 9 Insertion of s oft tissue equivalent material STES was the first material to be poured into the axial slice molds. After f ill ing, the molds were covered with wax paper air pockets were popped, and covered with smooth, weighted boards in order to force excess STES out of the molds A pproximately three hours was require d for enough curing to take place to remove the wax paper. After another 24 hours, the STES was removed from its mold Insertion of bone tissue equivalent material BTES was the next material to be incorporated into the axial slice molds. The bottom of each axial slice mold was sealed using contact paper and BTES was used to fill the voids designated for BTES in the engraving map. Again, air pockets were popped. Excess bone was removed with polishing using a belt sander. After 48 hours, the BTES was removed from the contact paper a belt sander with an 80 grit belt. Phantom assembly Before the LTES was inserted, phantom slices were glued together in groups of 5 20 slices using simple wood glue adhesive. Fortunately, wood glue has been shown by Jones et al. t o possess density and mass attenuation properties similar to STES at diagnostic energies. 32 This was done in order to facilitate rapid disassembly of the phantom and to decrease air gaps due simply to the task of stacking 200 slices together. The number of slices in each section was selected by determining where it would be valuable to have a separation for a future dosimetric point. This was done by dividing each organ required for ICRP effective dose into sections by volume.

PAGE 41

41 Insertion of lung ti ssue equivalent material LTES was the last material to be incorporated into the axial slice molds. The current methodology of LTES is different than that used by Winslow et al. wh ich was done in each slice individually Instead, the LTES was poured into t he phantom sections. Again, excess material removed by polishing with a belt sander. About 24 hours was required for the LTES to harden. 10 A completed axi al slice is shown in Figure 2 10 with a dosimeter R ecall that an axial slice with all three different materials was previously shown in Figure 2 3. Results The BTES was been previously characterized by Jones et al. to have a half value layer (HVL) of 9.8 mm at 80 kVp, and 13.3 mm at 120 kVp. 32 The aver age HU was found to be 622 HU which is consistent HU values of bone ranging from 400 to 1000 HU 45 The density of the LTES was measured to be 0.33 g cm 3 which agree s with the density for full inspiration. The average HU was found to be is 678 HU which is consistent with widely accepted HU values for lung ranging from 1000 to 500. 45 The STES had an HVL of 25 mm at 80 kVp, and 29 mm at 120 kVp. The measured density was 1.04 g cm 3 The average HU w as found to be 9.8 which agrees with a mixture of homogenous soft tissue as well as all organs, connective tissue, and adipose tissue soft. Muscle has 10 40 HU while both soft tissue and adipose tissue has a 50 to 100 HU. 45 T able 2 1 shows the ICRP weighting and number of available measurement points for each organ.

PAGE 42

42 Analysis The most significant development in this generation of UF phantoms is the new STES and LTES formulations. These new materials are easier to handle and provide more precise anatomical radiological representation while using commercially available material s. This generation also models adult anthropomorphic phantoms rather than pediatric phantoms. As both axial precision were increased and the overall size of the phantom were increased, the feasibility of construction was made possible only through streamli ning of the fabrication process, notably the incorporation of the engraver and elimination for need for vacuum chambers. The new generation of STES i s made of a durable pliable rubber compound while the epoxy resin wa s brittle and will break when dropped. Because of its pliability, insertion of real time dosimeters only required a thin slit to be cut into the material for the dosimetry system which will be discussed in Cha pter 3 without concerns about radiation streaming On the other hand, the epoxy resin required a dremel and creation of cha nnels. Small variations were observed in axial slice thickness which were initially attributed to the engraver. However, the actual cause were the accumulation of many tiny deformation of the foam molds as the attachme nt to the engraving table was broken. T he phantom slices created by Jones et al. were not attached together, therefore they did not have to account for the thickness of the actual wood glue. 32 This problem was discovered and solved by adjusting the eng raving depth to 4 .5 mm. Discussion This next generation of UF phantoms is an improvement upon the original techniques and phantoms created by White et al. and refined by Jones et al. 32 35 A

PAGE 43

43 male anthropomorphic 50 percentile reference phantom was created using this new methodology but could be used to develop a family of phan toms which would be developed from the series of computational twins generated by Lee et al. 36 Fisher et al. has include d the investigation of an adipose tissue equivalent substitute which was added to th e existing phantom s in order to represent the 90 percentile reference phantom 46 By foll owing the methodology detailed above, other institutions can feasibly fabricate their own anthropomorphic phantoms for clinical use with materials costing around $3,500 in addition to an initial equipment investment of approximately $10,000 and labor of several hundred hours per phantom

PAGE 44

44 Table 2 1 Organ/Tissue ICRP weig hting factors and number of measurement points

PAGE 45

45 Figure 2 1 ATOM phantom ( Photo courtesy of Technical Specifications: ATOM d osimetry p hantoms: A family of dosimetry phantoms Figure 2, p. 6) Figure 2 2 Head of RANDO phantom ( Photo courtesy of RSD Radiation Support ion therapy phantom, Unpublished ( 2008 ) Figure 4, p. 1)

PAGE 46

46 Figure 2 3. All tissue equivalent materials in sample axial slice showing a) STES, b) LTES, and c) BTES ( Photo courtesy of D Hyer Imaging doses in radiation therapy from kilovoltage cone beam computed tomography PhD Dissertation, University of Florida (20 10 ), Figure 2 5 p. 52) Figure 2 4 Co mpleted male 50 percentile anthropomorphic physical phantom

PAGE 47

47 Figure 2 5. ALRADS computational twin ( Photo courtesy of D Hyer Imaging doses in radiation therapy from kilovoltage cone beam computed tomography PhD Dissertation, University of Florida (20 10 ), Figure 2 1 p. 50 )

PAGE 48

48 Figure 2 6 CT t opogram of completed phantom ( Photo courtesy of R Fisher ose assessment and prediction in tube current modulated computed tomography PhD Dissertation, University of Florida (20 10 ), Figure 2 6 p. 63 ) Figure 2 7 Segmented data imported for s ample axial image ( Photo courtesy of J. 10 (3): 195 204 (2009), Figure 1 A p.199 )

PAGE 49

49 Figure 2 8 Engraver software input for s ample axial image ( Photo cou rtesy of J Winslow, 10 (3):195 204 (2009), Figure 1 C p.199) Figure 2 9 Engraved foam for sample axial slice ( Photo courtesy of D Hyer doses in radi ation therapy from kilovoltage cone beam computed tomography PhD Dissertation, University of Florida (20 10 ), Figure 2 4 p. 52)

PAGE 50

50 Figure 2 10 Conformity of sample axial slice around dosimeter ( Photo courtesy of D Hyer ose assessment and prediction in tube current modulated computed tomography PhD Dissertation, University of Florida (20 10 ), Figure 5 3 p. 115)

PAGE 51

51 CH APTER 3 REAL TIME POINT PLASTIC S CINTILLATION DOSIMET ER SYSTEM FOR USE IN MONITORING AND MEASU RING DOSE IN AGE RELATED MACULAR DEGENERATION STEREOT ACTIC RADIOSURGERY Background Age related macular degeneration (AMD) is a chronic, progressive disease of the macula, the central part of the retina and is the leading cause of severe vision loss and blindness among those over age 65 i n the developed world. 47, 48 AMD has dry and wet forms, which account for 20% and 80% of cases involving severe vision loss, respectively. 47, 48 The dry form of AMD is characterized by the loss of rods and cones in the central portion of the eye which results from atrophy of the retinal pigment epithelial layer below the retina. On the other hand, the wet form of AMD is characterized by leakage and bleedi ng from vessels which lead to increased tension at the macular site. The wet form of AMD begins with the formation of fibrovascular tissue from the choroids 48 This choroidal neovascularization grows beneath the pigment epithelium or into the sensory retina. 48 As a consequence, there is loss of rods and cones in the central portion of the eye that ultimately leads to vision loss. 48 Currently, the dry form of AMD is untreatable and monitored for signs that it might be progressing to the wet form of AMD. The wet form of AMD has many therapeutic options such as vitamin supplements, laser photocoagulation, photodynamic therapy (PDT) and intraocular drug therapy 49 52 Laser photocoagulation is the oldest form of treatment, and dates back to 1982. 50 While this method reduces chorodial neovascular membrane formation, it scars the mac ula with permanent loss of central vision. 50 PDT uses a photosensitizer which is subsequently exposed to light of a specific wavelength ; this method has had limited su ccess. 52 Popular intraocular therapy drugs are

PAGE 52

52 ranibizumab (Lucentis, Genentech, San Francisco, CA) or pegaptanib sodium ( Macugen OSI Eyetech, New York, NY). 49, 51 Both ranibizumab and pegaptanib sodium are examples of vascular endothelial growth factor inhibitors (VGEF 1). 53 These drugs work by inhibiting the growth of vascular tissue. 53 Other than comfort, an other drawback of intraocular drug treatment is that patients must return for periodic direct injections into the vitreous humor This becomes costly due to injection fees and pharmaceutical costs. T he biological tolerances of the human eye to ionizing radiation have been measured and published by the National Counc il of Radiation protection and Measurement (NCRP). 54 R adiation therapy is a novel alternative therapy option to current treatments of wet AMD. The goal of radiotherapy treatment is to destroy the leaky vasculature while minimizing radiation injury to surrounding tissues. Radiation treatments currently include both intraocular, epiretinal delivery of beta radiation and external radiations includ ing proton therapy, Gamma Knife radiosurgery and high energy external beam radiotherapy. 55 58 Oraya Therapeutics, Inc has developed the IRay TM a noninvasive, low voltage, stereotactic radiosurgery (SRS) device for treatment of the wet form of AMD The IRay TM addresses many of the inherent limitations of other radiation therapy options used in the past to treat AMD such as linear accelerators, brachytherapy sources, and Gamma Knife. 59 63 Radiation therapy has been proposed in conjunction with pharmaceutical treatments because the radiation would soothe the dehydrating effects of VGEF drugs. 64 This has been successfully administered with ranibzumab. 64 T he specifics of the IRay TM will be discussed in much more detail in the Materials and Methods section. The IRay TM is shown in Figure 3 1

PAGE 53

53 In order to ensure patient safety and accuracy in treatment, Oraya Therapeutics has graciously provided funding in order to develop a real time dosimetry system which will be capable of monitoring the constancy of the dose delivery device and also to provide a measurement of the dose delivered This dosimetry system must be easily integrated into their current machine. Plastic scintillation dosimeters (PSDs) were chosen because of their unique combination of low cost, temperature independence, water equivalence, small volume, dose rate and ene rgy independent response, and real time capabilities. 12, 17, 65 72 This system is based upon the system designed and built by Hyer et al. 17 Materials and Methods IRay TM Stereotactic Radiosurgery Device The peak kilovoltage energy used by IRay TM is 100 kVp, equivalent to the peak ene rgy used for common x ray and computed tomography (CT) procedures delivered at a source to target distance of 15.0 cm 59 61 The x rays are generated by a commercially available x ray tube ( (MXR 1 60HP/11, Comet AG, Switz erland). This tube has a tungsten anode and 0.8 mm Be filtration and is used at 100 kV between 1 18 mA. W hile the procedure is considered radiotherapy, the 100 keV energy used by the IRay TM places it squarely in the lower, diagnostic level regime. 11, 45 For comparison, typical external beam radiation therapy devices for cancer treatment generate energy spectra in the megavoltage regime 6 10, 18 MeV linear accelerator s are common; most b rachytherapy isotopes also emit a combination of charged and uncharged particles in the megavoltage regime; Gamma Knife uses Co 60, which als o emits in the megavoltage regime, specifically two photons with an average energy of 1.25 MeV 45 The use of photon beams in the kilovoltage energy range results in significantly less

PAGE 54

54 scattered radiation and thus significantly limit s the radiation dose to non targeted tissues. 11, 45 The IRay TM system is designed for office based ophthalmic use and is currently undergoing clinical trials for the treatment of wet AMD. The IRay TM is coup led to a robotic tracking system known as the IGuide TM which actively tracks using a proprietary localizing algorithm and corrects the gaze angle using a suction cup lens, ILen s TM with 25 mm Hg pressure. The IGuide TM and ILens TM are shown in Figures 3 2 a nd 3 3 respectively. The ILens TM defines the geometrical axis of treatment. The robotic positioning system uses two cameras to detect reflective fiducials attached to the lens a nd stabilizer bar of the IGuide TM which is a robotic positioning unit. This is attached to the ILens TM which is a contact lens with a negative pressure of 30 mm Hg used to stabilize the eye. The IGuide TM system tracks the position of three reflective fiducials which are dictated by the ILens TM position. The entire delivery is mon itored by the physician in real time. If the fiducials exceed a given position, a gating event is triggered and treatment is immediately stopped. Delivery resumes after the eye is repositioned and the physician orders it again. The positioning system autom atically aligns the IRay TM to the geometric axis. The intersection of the geometric axis with the retina is used as the first order alignment with the posterior pole. The proprietary localizing of the robotic positioning system also uses the fiducials to d etermine if a ny movement is beyond preset threshold le vels in which case it will automatically stop treatment. Extraneous e xposure to the lower lid of the eye is preventing by using a custom lid retractor. The axial length is determined by

PAGE 55

55 ultrasound and used for adjustment and fine tuning of the alignment with the posterior pole. A total of 24 Gy macular dose is split equally into three sequential beam deliveries, each entering through the inferior pars plana mainly in order to avoid the lens and optic nerve. 61 The IRay TM delivers a beam which is 3.5 mm wide at the entrance spot and is 4 mm wide at the macular target which is 150 mm from the x ray source The resultant focal spot has been measured to be 6 mm after accounting for movement and geometrical considerations of overlap Th e movement is attributed to microsaccades. 53, 73 Microsaccades are small, jerk like, involuntary eye movements which occur during prolonged visual fixation l onger than several seconds in any animals with foveal vision, including humans. 73 Microsaccade amplitudes vary anywhere from 2 to 120 arcminutes. 73 in Figure 3 4 This black box is typically used to position an ion chamber which is used to measure delivered dose. However, the PSD system fiber is small enough to be used as well. A close up of the radiochromic film which was used in order to ensure pro per positioning of the active element within the beam is shown in Figure 3 5 A PSD was attached to a solid water eye in order to measure dose delivered to the macular target. The solid water eye is shown on its stand with the PSD in Figure 3 6 The front of the solid water eye has a metal tip in order to attach to the tracking mechanism of the IGuide TM In a patient, this metal tip is attached to the outside of the ILens TM as shown in Figure 3 7 On the back of the solid water eye, a metal bracket was

PAGE 56

56 orig inally used hold radiochromic film which was previously used to measure dose. This bracket provided an accurate and convenient flat surface for PSD placement. Figure 3 8 shows a screenshot of the Oraya Jade software which is used to control the IRay TM The real time view from the IGuide TM is shown in top middle of the screen while the beam parameters are displayed on the right and the dose heat map is shown overlaid upon the macular target. The control unit is on the other side of the 1.6 mm lead shielding. Plastic Scintillation Dosimeter System A typical fiber optic coupled (FOC) scintillation dosimetry system is usually composed of three main components: (1) a PSD element which emits scintillation light which is captured by (2) a waveguide which subsequent ly transports the light to (3) a photodetector which transforms the light to an electrical signal. Figure 3 9 shows a schematic of a typical FOC system. Note that because of its lower energy, Cerenkov emission is not as a dominating factor, which is one of the foremost issues with higher energy, radiotherapy applications. 66, 74 Separation of the signal from Cerenkov emission has become a significant process critical to results for well funded endeavor s ; recent approaches have include d spectral filtering, scintillators which emit at orange wavelengths, and temporal gating. 18, 66, 74 76 Each fiber of the PSD system uses a 2 mm length of 0.5 mm diameter water equivalent scintillating fiber (BCF 12, Saint Gobain Crystals, Nemours, France) in the shape of a cylinder, optically coupled to a plastic waveguide, also with a 0.5 mm diameter (ESKA CK 20, Mitsubishi Rayon Co, Tokyo, Japan). BCF 12 scintillating fibers emits in the blue region with a peak at 43 5 nm with a polystyrene core, a polymethyl methacrylate (PMMA) clad ding, and a refractive index of 1.59; ESKA waveguides are

PAGE 57

57 plastic optical fibers with a PMMA core, a refractive index of 1.49, with loss of 375 dB/km at 450 nm. 77, 78 Beddar et al. predicts the Cerenkov threshold for this material to be 178 keV. 12, 13 Note the water equivalence (1.05 g/cm 3 ) of both the fiber and the waveguide. The distal portion of the scintillator was coated with reflective paint (EJ 510, Eljen Technology, Sweetwater, TX) in order to capture more scintillation photons. 79 Figure 3 10 shows a completed fiber More detailed information about fiber construction can be found in Appendix A T he uncoupled end of the waveguide was terminated in a female SMA 905 connector (SMA 505, Fiber Optic Center Inc, New Bedford, MA). A male SMA optical fiber adapter (E5776 51, Hamamatsu Corporation, Bridgewater, NJ) was used to enable connection with each of the two PMTs (H7467, Hamamatsu Corporation, Bridgewater, NJ). 80, 81 The counting data was buffered and transferred through a hub t o a computer for readout. In the original design by Hyer et al. the hub (UPORT 1610 8, Moxa Inc, Brea, CA) and five PMTs are powered by 12V and 5V DC power supplies, respectively. 17 For the Oraya project, there were only two PMTs, which allowed for a smaller hub (UPORT 1250, Moxa Inc, Brea, CA) and only one 5V DC po wer supply as the hub was powered by the USB connection itself. Figure 3 11 below shows an image of the actual system and Figure 3 12 shows the corresponding schematics. Physical System Benchmarking Depth dose c urves One of the overall objectives of this s ystem is to create a dosimetry system to measure the output of the IRay TM system. In order to characterize the IRay TM system, a depth dose curve will be created for water. Since access to the IRay TM will not be available until a trip to the Newark faciliti es, the ability to create a depth dose curve was

PAGE 58

58 tested with another x ray source at the same energy. A portable x ray source (SR 115, SRI, Bohemia, NY) was used with a 100 kVp and 1.50 mAs technique. The SR 115 uses a stationary anode, with 2.7 mm inheren t aluminum filtration and a tungsten target. 81 The generator uses a high frequency resonant inverter at 1.5 kW; the target has a focal spot of 1.0 mm and the target angle is 15 degrees. 81 In comparison, the IRay TM was modeled by Hanlon et al. as a 100 kVp x ray beam with an HVL of 2 mm Al, with an 1 mm focal spot on an anode target that is 15 cm from the macula target. 60 The depth of water was varied from 0.1 cm to around 7 cm. A constant source to surface distance was maintained as is standard procedure in radiotherapy applications. 11 A robust test can be done using solid water blocks. A laser mounted on a ring stand will be used in order to establish and confirm a precise surface definition. X ray spectra filtration Varying thicknesses of aluminum were placed in front of the x ray tube window from 0 to 4 .5 cm in 0.5 increments in order to determine the actual HVL. The IRay TM spectra is presumably not the same as the SR Prior to testing with the IRay TM yet, the filtration was adjusted for the same peak voltage of 100 kVp to match the beam qualities. High dose rate and high dose delivered calibration One of the other overall objectives of this system wa s to monitor the dose being delivered by the IRay TM system in real time. While t he PSD system ha d been well tested in CT scans which last less than 20 seconds t he IRay TM deliver ed a treatment dose of 24 Gy over the course of 5 minutes. The performance of the system at high dose rates was e xplicitly tested using a mobile c arm fluoroscopy unit (OEC 9800, GE, United Kingdom). In order to achieve the highes optional high

PAGE 59

59 ccording to Florida Department of He alth codes should not exceed 20 R oentgens per min ute at the entrance level 30 cm from the image intensifier assembly. 82 The performance of this type of system over long dose deliveries ha d been demonstrated Hyer et al. who used it for cone beam CT, which lasts 1 2 minutes, with no buffering problems. 83 This system ha d also been used for measuring computed radiography dose, in which the exposures are delivered in a very short time, thus resulting in a high dose rate. In order to test the performance, the dos e was measured using an ion chamber as well as the PSD system; the change in calibration factor between counts to exposure was examined. Computational benchmarking Hanlon et al. and Lee et al. have previously modeled the anatomy of the eye and surrounding area using a nonuniform rational B spline (NURBS) surface which includes critical nontrageted structures such as the brain, cranial bone marrow, cranial endosteum, thyroid, and salivary glands 28, 41, 59 61 A reference adult male and reference female head model was constructed consistent with anatomical data of the International 28, 60 These studies used MCNPX, a general purpose Monte Carlo radiation transport code d eveloped by Los Alamos National Labs which tracks a variety of radiation particles over broad energy ranges, to simulate ocular radiotherapy of AMD. 84 The x ray energy spectrum was generat ed using the computer software developed by the Institute of Physics and Engineering in Medicine. 85 For this study, only the depth dose measurements will be used to benchmark the computational results produced by Hanlon et al. and Lee et al. This will be done by

PAGE 60

60 using the x ray spectrum generated using the Institute of Physics and Engineering in Medicine as the source term and recreating the depth dose geometry used for the physical studies. Future studies might address the overall agreement by using multiple measurement sites for the PSDs and creating an anthropomorphic reference adult head model. Additional stu dies may investigate the agreement between different Monte Carlo radiation transport codes. IRay TM measurements Adding additional filtration to the portable x ray tube was only an approximation to the IRay TM Measurements of the performance on the actual I Ray TM were made on site using the Comet x ray tube. The main purposes of the PSD system was to measure: (1) current linearity, (2) saturation (3) depth dose curves, (4) ramping, and (5) beam stability. After warming up the tube and allowing the water coole r to settle, the positioning of the ion chamber with respect to the aperture of the x ray tube was calibrated. The x ray tube was used for 10 minutes at 100 kVp and 10 mA. The exposure positioning test was done in three steps: (1) at 200 mm distance with t he aperture on, (2) at 200 mm distance with the aperture off, and (3) at 150 mm distance with the aperture off. This was used to extrapolate the exposure spot at 150 mm distance with the aperture on. 150 mm was the distance from anode to macular target. Th is distance was controlled with stepping motors and checked with computer controls and verified using calipers. Next, the laser positioning unit was compared with the exposure spot using radiochromic film. The laser position ing unit is shown in Figure 3 13 and wa s built into the aperture. The aperture also contained a tungsten filter and a thin lead collimator.

PAGE 61

61 The PSD wa s fastened behind the top hat of the x ray tube, out of the direct path of the beam. The current linearity was tested, saturation levels were determined, and stability of beam out was measured. Depth dose curves were used to measure calibration factors between counts measured by the PSD versus the ion chamber measured dose. Additionally, the real time capabilities and high time resolution ( 100 Hz) of the PSD system allow for visualization and analysis of a few interesting features previously undiscovered. Theoretical O ptimization of PSD S ystem In general, one of the primary reasons for the l ack of commercial applications wa s the low signal t o noise ( SNR ) ratio of PSD systems. It is possible to examine each of the controllable parameters of a PSD system in order to improve the SNR to an appropriate level. For Poisson statistics, which are an accurate description of radiation detection measurem ents SNR of each PSD waveguide PMT is defined as (3 1) The variance in the noise can be defined as the quadrature sum of the main sources of PMT noise: photon shot noise, dark noise, and radiation induced PMT currents. In other words, (3 2) The signal can be defined as the product of quantum efficiency, integration time, and photon fluence per unit time. Radiation induced current in PSD applications are almost exclusively due to the Cerenkov effect, but can be ignored because it has a threshold energy of about 170 keV in plastic. 13, 74 The Cerenkov effect will be ignored for

PAGE 62

62 now; it will be discussed in C hapter 8 in conjunction with a discussion of filtering methods. SNR can thus be written as (3 3) The fluence, can be calculated as the product of the number of photons, n ; coupling efficiency of the scintillator to waveguide, S W which includes geometric considerations as well as optical transmission; waveguide transmission efficiency, W ; and co upling efficiency between waveguide and PMT, W PMT SNR is now written as (3 4) Notice that in this equation, the number of photons can be derived exactly. The couplings and dark current ca specifications. In the PSD, the dose absorbed, D for mass energy absorption coefficient at energy, E and fluence, is equivalent to 11 (3 5) The kinetic energy, T imparted upon a PSD of mass m is 11 (3 6) The exposure, X imparted upon m in free space is (3 7) where e is the charge of an electron and is the mean energy expended in a gas per ion formed. 11 Rewriting this expression in terms of fluence yields

PAGE 63

63 (3 8) Substituting Equation 3 8 into Equation 3 7 removes energy, (3 9) The number of photons which are actually produced is the kinetic energy imparted divided by the average energy requi red from an electron in order to produce a light photon, ( w PSD ). This is written as (3 10) This can be better grasped as the number of light photons per unit exposure, (3 11) Results Depth D ose C urve in L iquid W ater The depth of water was varied from 0.5 cm to 5 cm, in 0.5 cm increments. The exposure was measured with a dosimeter (35050A, Keithley, Cleveland, OH) and a pancake ionization cham ber attachment (96035B, Keithley, Cleveland, OH). The minimum depth of water was 0.5 cm because the PSD was not placed free in air, but in a soft tissue equivalent substitute (STES) which has been characterized by Winslow et al. 10

PAGE 64

64 After adding water, the x ray head was readjusted to maintain at an arbitrarily determined 30 cm window to surface distance. The only restriction on the window to surface distance is the guard rail along the treatment head, which prohibits any distance smaller than 13 cm. The distance from the entrance window to the x ray source is unknown. In any case, obtaining a depth dose curve provides a good proof of concept. The first iteration of this was done using a plastic container and manually filling it with water to a given depth. The set up is shown below in Figure 3 14 Note an STES mold was made in order to accommodate a PSD dosimeter as well as an ionization chamber. The depth dose curve measured with the PSD is displayed using markers in Figure 3 15 below. Dose measurements were repeated three more times at each depth for a total of four measurements. Figure 3 15 shows the SR cm of liquid water. Depth Dose Curve In Solid Water A more robust test was d one using standard grade solid water slabs (457, Gammex, Middleton, WI), which are advertised to provide calibrations within 1% of true water dose. Again, four measurements were taken at each depth. A laser mounted on a ring stand was used to verify the precision of the source to surface distance. Solid allows the source to surface to be maintained within the tolerance of the machined size of the solid water slabs. The depth dose curve measured with the PSD is displayed using markers in Figure 3 16 Dose measurements were repeated three more times at each depth for a total of four measurements. Both Figure 3 15 and Figure 3 16 show the SR to be around 4.0 cm of liqu id water. The solid water tests offer much more precise

PAGE 65

65 measurements of water depth, with better resolution. More importantly, it does not display the outlier seen at 4.0 cm of depth in Figure 3 15 Beam Q uality Hyer et al. used a CT x ray source, which has an HVL of 6 8 mm Al. 45 The portable x ray source was expected to ha ve an HVL around 2 mm. 45 It is known that the inherent filtration is 2.7 mm Al for the SR 115. 81 This includes the collimator, which is typically 2.0mm Al at 100kV, and the tube filtration which is an additional 0 .7 mm. 81 The anode has 0.7 mm Al inherent fil tration while the window is plexiglass but its filtration is equivalent to about 2 mm Al. 81 The focal spot to window di stance is 7.5 inches (around 19 cm ). This is long er than the Oraya source to target dis tance of 15 cm, but should not alter the depth dose measurements. Figure 3 17 shows the result of increasing the thickness of type 1100 aluminum sheets (JRT Associates, Elmsford, NY ). The HVL of the x ray beam was found to be 4 .0 mm Al by direct measurement and was verified by an exponential curve fit of the other data The goodness of fit (R 2 value) was 0.9934. Ion Chamber Benchmarking There is an increasing divergence between the PSD measured normalized intensity when compared with th e ion chamber. If the depth at which the dose is being delivered remains constant and is well known, then this simply serves as a calibration curve. In this Oraya application, this is the case, with a robotically controlled source to target distance of 15. 0 cm. The actual water equivalent depth is not currently known and depends upon the eye geometry. ICRP Publication 89 gives reference mass and density for the lens and total eye structure, with additional details provided by

PAGE 66

66 dimensions given in NCRP Report 130. 28, 5 4 Using these references, it is expected that the water equivalent depth will be around 1.5 cm. 28, 54 Figure 3 18 shows the PDDs of both the PSD and the ion chamber as measured with the port able x ray generator in a luminum Figure 3 18 shows an increasingly i mportant depth correction factor as depth increases. For example, it is around 6% at 1.5 cm, while it is around 20% at 6 cm. The correction factor for depths ranging from 0 to 2.0 cm is given explicitly in Table 3 1 for future reference. The substantial de viation between the PSD and the ion chamber depth dose is due to differing beam qualities and addressed in detail in the Analysis section. Computational Benchmarking The x ray spectrum was generated using the Institute of Physics and Engineering in Medici ne as the source term and recreating the depth dose geometry used in the physical studies. MCNPX was used as the radiation transport code, and the photon flux was calculated at different depths of water. The simulations showed excellent agreement wi th the ion chamber measurements as seen in Figure 3 19 High Dose Rate and High Total Dose Performance In order to test the performance of the PSD system for large doses, the calibration factor between counts given by the PSD vs total dose measured by the ion ch amber was measured free in air for different exposures ranging from 0 R to about 16.0 R delivered by a c arm used in boost mode. The PSD system was set up to take measurements four times per second (250 ms bins). The calibration factor is shown as number o f counts per R and is shown against exposure in Figure 3 20 The same data are shown normalized to an average value and are shown in Figure 3 21 There is no

PAGE 67

67 more than a 1% deviation for all exposures, which ranged from 0.862 R to 15.26 R. Each exposure was delivered at 100 kV, which is the energy used by the IRay TM In order to test the performance of the PSD system for large dose rates, the counts given by the PSD versus the total dose measured by the ion chamber was measured free in air for different dose rates as shown in Figure 3 22 This was accomplished by changing the current of the c arm. Like the previous tests, these measurements were done at 100 kVp, which is the energy used by the IRay TM The slope of Figure 3 22 shows that calibration factor between counts and exposure remains linearly proportional to the exposure from 0.862 R to 15.26 R, with a goodness of fit (R 2 ) greater than 0.99. IR ay TM Measurements Direct measurement of the IRay TM were made. This machine uses a commercially available x ray tube previously described. The PSD was housed outside the collimator of the IRay TM Th e collimator is shown in Figure 3 23 and demonstrates the need for a small volume dosimeter. The main purposes of the PSD system was to measure: (1) current linearity (2) saturation (3) depth dose curves, (4) ramping, and (5) beam stability. PSD current l inearity The PSD was taped behind the top hat of the x ray tube, out of the direct path of the beam as shown in Figure 3 24 T he PSD and ion chamber response was measured as a function of the x ray tube current as shown in Figures 3 25 and 3 26, respectively The current was varied between 3 and 18 mA. In addition to investigating proportional response, Figures 3 25 and 3 26 were used to check for n onl inearities which could indicate saturation. The correction factor (CF) between PSD counts and ion chamber

PAGE 68

68 charge is shown in Figure 3 27 The CF was 4.47 +/ 1.5% with the values ranging from 4.33 at 18 mA to 4.63 at 3 mA. Figure 3 25 is similar to Figure 3 22, however the latter wa s operated at a h igher dose rate PSD s aturation Saturation can be observed in two ways: scintillator overflow and PMT overflow. Specifically, scintillators can be overloaded with such high current that theoretically they could not produce enough photons. This would primarily be a function of decay time which is 3 ns for this scintillator. However, PMTs can be overloaded with a large number of signal carriers. This number was determined experimentally. The fibers were moved a fe w millimeters such that the PSD was placed directly in the beam path. The position of the PSD was verified to be in the beam using film. The average PSD counts in a three second bin within the plateau region was plotted as a function of varying current fro m 3 to 18 mA, in 3 mA increments. The counts were taken in 10 ms increments. This is shown in Figure 3 28 which shows a slight nonlinearity at 15 mA and a definite plateau at 18 mA. A safe region of operation would seem to be around 1E+5 counts per bin. Th is corresponds to 10 million counts per second. The binning was increased from 10 ms to 200 ms in order to determine the maximum number of counts per bin. The maximum number of counts was 1E+6 counts, which was measured at a bin size of 100 ms. This again corresponds to 1E+7 counts per second. PSD d epth d ose c urves Solid water blocks were used to create percent depth dose (PDD) curves. There are two depth dose techniques: an attenuation scheme and a scatter scheme. The he solid water attenuator close to the radiation

PAGE 69

69 source. With the narrow beam geometry of the beam and the lack of a backscattering large amount of scatter. The PDD curves are shown in Figure 3 29 while the calibration factor (CF) between ion chamber PDD and PSD PDD is shown in Figure 3 30 The increasing thickness of aluminum means a higher effective energy and more filtered beam. There is not a concern for secondary scatter or backscatter because the rall PDD response with respect to the ion chamber can be attributed to sensitivity to higher energy photons. Note that the plane parallel small volume (.0053 cm3) ion chamber (34013, PTW, Germany) was confirmed to be energy independent. PSD r amping The PSD system was the first system which able to visualize the tube current in real time. This is shown in Figure 3 31, which will be discussed. First, there was a small dip in the x ray tube output around 1.25 seconds which was attributed to underdamping of the x ray tube circuitry. Note that Comet uses underdamping in its circuits in order to avoid overheating the anode. The overvoltage chosen for this x ray tube was 87%. This corresponds to the voltage overshooting allowed in the circuitry before the x ray tube would turn off. Secondly, the actual ramping is clearly visible in this figure and can be discriminated using the PSD system. The end effect of the beam was already documented and measured to be around 2.1 seconds. In this case, this meant that 0 .9 seconds at full power represented 3 seconds of ramping. Therefore, 2.1 seconds of full

PAGE 70

70 power were added to the end of each measurement to artificially compensate for the lost time. The mechanism of x ray generation can explain both the small increase in counts at 0.5 seconds and the overall rise to the plateau value. Before the tube can fire x rays, the filament is pre heated. Once any voltage is applied, the electrons rush towards the anode. Next the tube voltage is ramped to its maximum value of 100 k V. Finally, the filament current is ramped to 3.8 A. The tube current was three orders of magnitude smaller and varied from 1 18 mA. The small increase seen in the tube output at 0.5 seconds was due to the initial application of the voltage. The increasin g rise from 1 to 1.4 seconds can be attributed to ramping of tube voltage. At 1.4 seconds, the small drop in output occurs as filament ramping takes over, and the final rise towards the plateau region is due to filament ramping. Comet x ray t ube b eam s tabi lity was seen in the PSD measurements using 250 ms bins. The 250 ms bins were previously chosen for the real time monitoring of the Oraya IRay TM This was due to the PSD sys time capabilities. The x effect which ripple had upon the tube current. There was an almost linear rise in tube current as the filament current was increased. However, electron self shielding created a maximum plateau region.

PAGE 71

71 Theoretical Signal to Noise O ptimization Testing the dark current of the PMT yielded negligible response. About 0.1% dark noise was also meas ured by Beddar et al. 67 Thus, Equation 3 4 can be simplified to (3 12) The lower event threshold on the control program was permanently set at 10 counts in order to eliminate spurious background dark noise. T he coupling coefficie nts are similar to Ayotte et al. and quantum efficiency have been calculated. 86 The number of photons will now also be calculated. The IRay TM which this dosimetry system was designed to characterize, uses 100 keV photons to treat AMD. From Attix, the mass energy absorption coefficient ( en ) at 100 keV for air and polystyrene are 0.0234 cm 2 /g and 0.0231 cm 2 /g, respectively. 11 The average energy spent by an electron per light photon produced ( w PSD ) is 60 eV for plastic scintillators; the average energy spent mean energy expended in a gas per ion formed divided by the charge of an electron ( w/e ) is 33.97 J/C in air. 11 The mass of the fiber is given by (3 13) where density ( ) is 1.05 g/cm 3 diameter ( d ) is 0.500 mm, length ( L ) = 2 mm. 77 Putting these values into Equation 3 11 gives (3 14)

PAGE 72

72 Typical E xposure Hanlon et al. and Lee et al. specify an absorbed dose of around 24 Gy for an IRay TM treatment. 59 61 The number of photons produced in the scintillator for 24 Gy is (3 15) Using previously calcu lated geometry factor, coupling, and transmission factors, 0 the number which make it out of the scintillator and through the waveguide is (3 16) Using previously obtained coupling and quantum efficiency factors, the number which make it through and are converted to electrical signal are (3 17) Using the previously obtai ned PMT multiplication factor, t he amount of total charge generated can be calculated by (3 18) Analysis Appropriateness of PSD system A PSD system has many unique features, notably low cost, temperature independence, water equivalence, small volume, dose rate and energy independent response, and real time capabilitie s. 12, 17, 65 72 The main purposes of the PSD system was to measure: (1) current linearity, (2) saturation (3) depth dose curves, (4) ramping, and (5 ) beam stability. For this application to SRS, the fast response, small volume, linear dose response, linear dose rate response, and real time characteristics are the most important. The depth dose curves were established in comparison with both an ion

PAGE 73

73 cha mber and computational simulations. These results showed good agreement, with the CF, as shown in Table 1, due to beam quality. Lastly, the PSD system was shown to theoretically generate an extremely measurable amount of current. This SRS system delivers 2 4 Gy over the course of five minutes in three beams. Thus the system must respond linearly to both high dose and high dose rate. This requirement becomes even more stringent when the geometrical distance factor is included. Specifically, the target is loca ted 150 mm from the source, while the PSD is located only 2 3 centimeters from the target Thus it is expected to receive more dose due strictly to geometrical factors However, it was confirmed that the PSD system does respond linearly to high doses and h igh dose rates. PSDs have response times on the order of nanoseconds, and in this application the r eal time capabilities will be used to trigger gating events. These measurements will be done in conjunction with movement registered by the IGuide TM which a lso triggers gating events. With incorporation of the PSD system, both patient macular target position and beam constancy will be ensured. Ophthalmic C onsiderations An important consideration with regards to ophthalmic radiosurgery is the development of radiation retinopathy. 53, 87, 88 Radiation retinopathy is defined as non inflammatory damage to the retina which is characterized by preferential da mage to small diameter vessels. 87, 88 The threshold dose of radiation retinopathy is estimated to be around 45 Gy, occurring 12 24 months or longer following therapy 53, 87, 88 Note that this is only an approximation and is not based upon direct studies of retinal radiosurgery, but upon studies of patients receiving radiation for head and neck cancers which also exposed the retina 87, 88 59 63

PAGE 74

74 A recent study by Evans et al. has shown that radiation retinopathy is not a concern in external beam radiotherapy with zero cases of retinopathy seen out of 1078 cases. 89 Similarly, no evidence of radiation retinopathy was seen by Avila et al. in his studies which used Sr 90 applicatiors. 55 Because r adiation retinopathy is dependent upon the size exposed, the small spot size of the IRay TM suggests that retinopathy will not be a large concern. Basic radiobiology dictates the importance of area irradiated when considering the lethal effects of radiation to normal tissue. 90 For example, a whole body dose of 5 Gy will result in mortality of 50% of in dividuals within days to months. 90 This assertion agrees with Hanlon et al. who claims that the main difference in the treatments which resulted in radiation retinopathy and those which did not appears to be the volume of retina irradiated. 60 Specifically, t he 10 th 90 th isodose curves correspond to a volume of 30 mm 3 and 300 mm 3 for external beam radiation therapy and proton beam thera py, respectively. 60 In comparison, the IRay TM has a volume of only 3.14 mm 3 within the 10 th 90 th isodose curves. 60 The dose monitoring capabili ty of the PSD system will ensure that doses which reach a threshold will automatically trigger a gating event and shut down the machine. Another safeguard against radiation overdose is the IGuide TM which has been shown to provide stable, robotically contr olled targeting Studies by Moshfeghi et al. have shown the complete three beam overlap in the foveal target using radiochromic film in a human cadaver eye. 53 Therefore, the targeting has been demonstrated by Monte Carlo simulations, human cadaver eyes, and animal pig work. 53, 59, 60 T he automated tracking and gating algorithm of the IRay TM ensures tha t the irradiated area on the macula is

PAGE 75

75 constrained to a spot (4 mm) at the plane of the macula with the 10th 90th percentile isodose volume equivalent to only 1% 10% the volumes from proton beam therapy and external beam radiation therapy, respectively. 60 Selection of photomultiplier t ube PMTs provide actual real time capabilities with resolution realistically limited only by electronic readout, with typical binning around 10 ms. 17 While it is true that other systems notably, CCD based systems have the ability to measure dose as function of time, only PMT time systems. To explain, PMT based systems measure dose directly through counts and subtract ion, providing an almost instantaneous readout limited only by buffer and transfer speed. 91 CCD based systems, on the other hand, require post processing to measure dose through brightness and chromatic filtering. 66 PMTs usually operate with a gain on the order of 1x10 6 ; in other words, PMTs provide better sensitivity: requiring only 10 ms bins, while CCDs require long measurements on the order of 5 seconds in order to acqu ire a very precise signal. 17, 91, 92 Therefore, PMT based dosimetry can be used in a monitoring situation analogous to an ion chamber in a linear accelerator treatment head. In comparison with CCDs, PMTs are also very resilient and easy to set up without the hassle of optical coupling: finding focal lengths, adjusting lenses, and positioning within the field of view. In other words, PMT based systems are an excellent option if: a small number of measurements are required, the signal is extr emely small, or if real time monitoring is required. Discussion As the leading cause of severe vision loss and bl indness among those over age 65, AMD is a disease which should cause concern. The IRay TM SRS device uses a

PAGE 76

76 quick, convenient and painless one t ime SRS as an effective alternative to painful monthly pharmaceutical injections. This device was developed by Oraya Therapeutics, Inc. and has been validated in Monte Carlo simulations, human cadaver eye studies, pre clinical animal studies and is current ly in a phase 1 clinical trial. It is important that this machine delivers dose precisely and safely. T he stability was controlled solely by the IGuide TM a robotic system which relied upon feudu cial positions. However, the PSD system was incorporated to provide a completely independent measure of (1) current linearity, (2) saturation (3) depth dose curves, (4) ramping, and (5) beam stability. Discrepancies in any of these areas trigger ed gating events PSDs offer a unique combination of low cost, temperat ure independence, water equivalence, small volume, dose rate and energy independent response, and real time capabilities. 12, 17, 65 72 These attributes make this PSD system an extremely viable choice for monitoring and measuring of dose delivered by the IRay TM SRS device. Specifically, the small volume allows placement within the existing IRay TM machine without the need for modifi cations; the linear high dose and high dose rate attributes allow accurate measurement; and the real time capabilities are used to monitor dose as it is being delivered. This work was supported by Oraya Therapeutics, Inc.

PAGE 77

77 Table 3 1 Calibration factor (CF) of PSD PDD to Ion chamber PDD

PAGE 78

78 Figure 3 1 IRay TM stereotactic radiosurgery device Figure 3 2 IGuide TM robotic positioning system

PAGE 79

79 Figure 3 3 ILens TM eye stabilization system Figure 3 4. Calibration cyclops

PAGE 80

80 Figure 3 5. Film used to ensure position of active PSD element Figure 3 6. Solid water eye with PSD taped to back

PAGE 81

81 Figure 3 7. ILens TM attached to metal tip Figure 3 8. Oraya Jade software screen shot

PAGE 82

82 Figure 3 9 Schematic of FOC PSD showing A) reflective paint, B) PSD, C) waveguide, and D) SMA connector Figure 3 10 C ompleted PSD fiber

PAGE 83

83 Figure 3 11 Complete d dosimetry unit picture Figure 3 12 Schematics of completed dosimetry unit showing 1) PMT, 2) PMT, 3) Hub, 4) Panel mount, 5) Power supply, 6) Filtered power supply

PAGE 84

84 Figure 3 13. Laser positioning unit mounted in aperture Figure 3 14 Set up for measuring p ercent depth dose of PSD in liquid water

PAGE 85

85 Figure 3 15. Percent depth dose of PSD in liquid water Figure 3 16 Percent depth dose of PSD in solid water

PAGE 86

86 Figure 3 17 Beam quality of SR 115 portable x ray unit measured in a luminum Figure 3 18 Percent depth dose of PSD compared with ion chamber measured in alu minum

PAGE 87

87 Figure 3 19 Percent depth dose of MCNPX results compared with ion chamber Figure 3 20 High dose performance: c alibration factor (CF) as a function of exposure

PAGE 88

88 Figure 3 21 High dose performance: n ormalized version of calibration factor (CF) as a function of exposure Figure 3 22 High dose rate linearity: c ounts as a function of exposure

PAGE 89

89 Figure 3 23. IRay TM collimator Figure 3 24. PSD on collimator

PAGE 90

90 Figure 3 25. PSD counts as a function of x ray tube current Figure 3 26. Ion chamber charge as a function of x ray tube current

PAGE 91

91 Figure 3 27. Calibration factor between PSD and ion chamber for x ray tube current Figure 3 28. Count saturation

PAGE 92

92 Figure 3 29. PSD and ion chamber measurements as function of solid water depth Figure 3 30. Calibration factor between PSD and ion chamber for depth dose

PAGE 93

93 Figure 3 31. Real time measurement of tube output showing tube ramping

PAGE 94

94 CH APTER 4 A METHODOLOGY FOR DI RECT QUANTIFICATION OF OVERRANGING IN HELICAL COMPUTED TOM OGRAPHY Background In helical computed tomography (CT), reconstruction information from volumes adjacent to the clinical volume of interest (VOI) must be collected in order to properly reconstruct the VOI. These additional reconstruction volumes have been various ly referred to as overranging lengths, overscanning volumes, or overranging lengths by previous investigators, who have made indirect inferences to dose contributions from these effects. 93 96 In many dosimetry evaluations of CT exams, dose contributions of the primary beam are traditionally only considered within the clinical VOI. Reconstruction algorithms will almost always require some degree of overranging, thus this study aims to introduce an innovative and accurate method to directly characterize overranging length with real time dosimetry. Typically, f or a given scan length (ie, imaged volume), at least an additional one half of a rotation is necessary at the beginning and at the end of the scan to ensure that complete data sets are obtained for the reconstruction of the first and the last sections. Including overranging lengths in the prediction of dose length products, or in Monte Carlo calculations of effective dose will red uce underestimates of dose. Overranging lengths are especially important considerations if there are radiosensitive organs in close proximity to the clinical VOI which could have been avoided. As increasingly larger longitudinal coverage is provided by mo dern multisection CT scanners, overranging effects increase and have to be taken into account properly. One method that addresses overranging is adaptive section collimation, where parts of the x ray beam exposing tissue outside of the volume to be imaged are blocked in the z

PAGE 95

95 direction by dynamically adjusted collimators at the beginning and a t the end of the CT scan 97 Studies by Deak et al. have shown dose reduction as large as 38%, but averaging 10% when dynamic collimators are used f or a typical chest scan of 30 cm and pitch factor of 1.0. 97 Previous studies have quantified overranging lengths but they rely upon either cumbersome computational studies or indirect extrapolation methods. In the study by Tzedakis et al. the effective dose is calculated with the Monte Carlo N particle (MCNP) radiation transport code. 93 95 Th eir source uses a previously generated x ray spectrum emitted from line sources along a helical path which was benchmarked with CTDI dosimetric data as well as z axis dose profiles obtained with thermoluminescence dosimeters. 93 95 These console 93 95 The proposed methodology avoids this reliance upon console measurements by determining the total length scanned using methods strictly based upon timing. Overranging lengths are particularly important considerations for deterministic Monte Carlo calculations of effective dose As noted by Tzedakis et al. typical CT simulations assume that the exposed length is equal to the VOI length. 93 95 In other words, while a simulation can accurately predict VOI dose, it will always underestimate DLP if it disregards the overranging lengths, which are always a part of a clinical scan. 45 An experimental physical study by van der Molen et al. derived the overranging length with a novel extrapolation method wh method. In this method, the DLP was derived from measurements of a varying number of rotations. 96 These represented the DLP due to overranging and a varying size of VOI.

PAGE 96

96 The number of rotations which represented the VOI were then line arly extrapolated to zero rotations to represent a DLP corresponding solely to the overranging. 96 This method assumes that the overranging behavior stays constant at a low number of rotations. If, for example, the machine were to switch to a different reconstruction scheme the dose slope method would not necessarily be accurate. This new technique has the advantage of requiring only one measurement, instead of multiple measurements which are necessary to create the extrapolation data. Furthermore, this technique is the fi rst direct physical measurement of overranging and unlike the dose slope method does not require extrapolations. This method is completely independent from the console readings. In addition to comparisons with the two other overranging methodologies, a com puted radiography (CR) plate will be used to independently verify the exposure length. Materials and Methods CT Scanner and Real Time Dosimetry System A Siemens Somatom Sensation 16 multislice helical CT scanner (Siemens AG, Forchheim, Germany) was used wi th a routine abdominal scan protocol. This protocol is a very diverse protocol, but was chosen to provide a direct comparison to overranging measurements by Tzedakis et al. and van der Molen et al. This particular protocol permits the selection of pitches between values of 0.50 and 1.50 in increments of 0.05; rotation speeds of 0.5, 0.75, 1 and 1.5 seconds per rotation; reconstruction slice widths of 0.75, 1, 1.5, 2, 3, 4, 5, 6, 8, and 10 mm per slice; and detector collimations of 12 and 24 mm. The abdomin al scan protocol was used with a tube voltage of 120 kVp and tube current of 140 mAs. The automatic exposure control was turned off for all scans.

PAGE 97

97 equivalent fiber optic coupled (FOC) real time dosimetry system was used for this stud y. While this dosimetry system accommodates up to five small dosimeter elements which can be read in real time with a temporal resolution of 10 ms, only three dosimeters were required for use in this study. The system performance has been previously descr ibed by Hyer et al. and is similar to the PMT PSD system described in Chapter 3 of this work 17 The dose linearity of the system has a correlation coefficient of 1.000 over exposures ranging from 0.16 to 57.29 mGy. Each dosimeter element is a small cylindrically shaped plastic scintillator 500 m in diameter and 2 mm in length which closely approxim ates a point detector. The scintillator does not suffer from angular dependencies with the dose measured free in air at isocenter varying less than 5% over an entire revolution. In order to reduce scatter contributions for the measurements performed in t his study, all dosimeters were suspended free in air. Dosimeter Positioning This methodology only involves a determining the distance betwee n scintillators. The maximum discrepancies between physical distances and console reading measurements must be no m ore than 1 2 m m as stipulated by quality control guidelines 98 However, this was explicitly tested by comparing distance measured with a calibrated linear measurement device between two lead bars against a measurement The remaining portion of the methodology is based upon the temporal spectrum obtained, where the limi ting resolution of this system is 10 ms which using a typical (pitch = 1) table speed of 30 mm/s yields an uncertainty of 0.28 mm 17 Assuming even the highest table velocities (pitch = 1.5) the uncertainty is slightly less than 0.50 mm. Thus assuming a

PAGE 98

98 high table velocity, the maximum uncertainty of the methodology is estimated as 0.7 mm which is obtained by the sum of the console measurement precision (0.5 mm) and the timing measurement precision (0.5 mm), added in quadrature. The unique ability of the point dosimetry system to provide real time information at v arious positions along the CT scan is exploited to analyze the temporal response of the CT equipment. Retrospective analysis of the event timeline provides all the information required in order to quantitatively evaluate the extent of overranging and the corresponding dose contributions. The response is initially characterized temporally, but is readily converted into longitudinal position with information about the table feed speed, which is also verified by the experimental measurements. In the experime ntal set up, the first of three dosimeters is positioned in the bore at a stationary longitudinal position while the second and third dosimeters are placed on the table, at isocenter height, and spaced 20 30 cm apart in the longitudinal direction. The bore dosimeter is suspended free in air the only restriction is that it stays within the fan angle for the entire helical scan. In other words, the stationary bore dosimeter will always be within the scan volume and its response is used to measure the exact scan duration. The table dosimeters move through the primary beam during the scan and are only within the scan volume for a portion of the total scan. After a scout scan, the two table dosimeters are located and used to define the boundaries of the VOI for the helical scan. The temporal response of the table dosimeters is used to measure the portion of the scan exclusively in the VOI. The table speed is measured and the total scan duration and VOI scan duration are converted into distances. Overranging leng th is determined by simply subtracting the VOI length from the total scan length. Figure 4 1

PAGE 99

99 shows a 3 D schematic of the setup, in particular the dosimeter placements within the bore and upon the table. Exposure Pattern As an absolute benchmark, the ove rranging length was determined by measurements using CR T he exact exposure pattern of the Siemens s canner was tested by scanning the CR plate at a low mA and developed at a low speed. Each of the physical measurement techniques provides different measurem ents. For example, the new methodology offers a measurement of beam on time and small region table velocity measurement. On the other hand, the console methodology used by Tzedakis et al. offers a beam on time and an overall velocity m easurement. Finally, the CR plate offers information on the length of the sca n, and the pattern of the scan. The path of the x ray tube will be determined by its exposure pattern from the CR plate. The total exposure length will be explicitly measured and compared with the mea sured lengths. Clinical Impact This methodology produced overranging lengths as measured in units of distance but provided no indication of clinical exposure. In order to measure dose, a number of metrics are available such as computed tomography dose inde x (CTDI), dose length product (DLP), and effective dose. CTDI and its main derivatives namely weighted CTDI (CTDI w ) and volume CTDI (CTDI vol ) are measured in units of milliGray (mGy); dose length product is measured in milliGray cm (mGy cm); and effective dose is measured in Sieverts (Sv). CTDI was first introduced and defined by the U.S. Food and Drug Agency as the average dose imparted by a single axial acquisition to a standa rd 100 mm pencil chamber dosimeter inside a CTDI phantom over 14 CT slices. 21 23, 45, 99, 100 However,

PAGE 100

100 CTDI is expected to vary with in the axial plane, so CTDI w was introduced and defined as a weighted average of the CTDI at peripheral and center locatio ns. 21, 45 The advent of helical CT magnified variations along the longitudinal axis, so CTDI vol was introduced and defined as CTDI w normalized to pitch. 21, 45, 100 CTDI vol can thus be used to express the average dose delivered to the scan volume 22 DLP was introduced to represent the total radiation exposure for a whole series of images and is defined as the product of CTDI vol and irradiated length. 21 23, 45, 99, 100 As explained by Dixon et al. CTDI values are simply not accurate at the edges of the scans. 21 23, 99 U sing DLP in particular would imply that the CTDI vol values would be the same for the entire length. However, for this application, the sharp dose gradients between the on and off values at the edges of the VOI allow DLP to be a useful first approximation to describe the overranging length. The DLP reflects the total energy absorbed and thus the potential biological effect attributable to the complete scan acquisition. Thus, an abdomen only CT exam might have the same CTDI vol as an abdomen/pelvis CT exam, but the latter exam would have a greater DLP, proportional to the greater z extent of the scan volume. 100 CTDI vol was not chosen to measure overranging because it only represents a single slice. 21, 22, 99, 100 On the other hand, DLP is proportional to scan length and will be directly dependent on overranging. In other words, overranging length will provide a larger DLP, with no effect upon CTDI vol Futhermore, while CTDI vol can accurately represent is the dose in the central slice within a long scan length it is not an accurate measure of dose on the edges of the scan. 21 23, 99 DLP is a useful metric specifically for overranging because it is dependent upon overall length. This study thus uses DLP in addition to r aw measurements of length to quantify clinical implications of overranging.

PAGE 101

101 Results Exposure P attern The CR image is used to determine the exact length of exposure. The ROI is marked with two metallic markers on the side closest to the bore and one metallic marker on the side further from the bore. These markers are measured center to center in o rder to provide a length to pixel calibration factor for the image of 0.04 mm/pixel. The average length over 7 images was 695 ( 2.3 pixels), which converted to 27.8 cm. If the exposed length is 27.8 cm, then the actual translation of the table would be 27.8 minus the beam collimation (half beam width at beginning and half beam width and end) or a 25.4 cm translation. In order to move 25.4 cm, with the steady state velocity already calculated, the time required is 5.25 seconds. This agrees with the bore d osimeter timing measurement as well as the console timing measurement. The overall CR measured length of 27.8 cm is greater than the console measurement of displacement length (26.7 cm) by 1.1 cm, or about one half of the 2.4 cm beam collimation. Thus, as expected, exposure is occurring outside of the overranging length displayed on the console. However, it is not the sum of table displacement and a full beam collimation as defined by Tzedakis et al. but only one half of a b eam collimation. Therefore, the overranging lengths obtained by this new methodology can be compared with results by Tzedakis et al. which are adjusted by one half of a beam collimation width. These Overranging Dependence upon Protocol Parameters Figure 4 2 shows the response of the dosimeters as a function of time for a 25.1 cm user selected scan length, for the lowest pitch of 0.5 and the highest pi tch of 1.5. In

PAGE 102

102 Figure 4 2 peak response represents the time required for the table to translate the two dosimeters into the primary beam. Table velocities are calculated by the quotient of the measured distance between table dosimeters and the time to translate between these dosimeters. The re is excellent linearity between table velocity and pitch, w ith a correlation coefficient squared (R 2 ) value of 0.999 and reproducibility for the three scans lengths performed of 19.7, 25.1, and 30 cm. This demonstrates the relationship expected between speed and pitch for a helical scan as well as providing valida tion of correct dosimeter placement and measurement. Figure 4 3 shows the overranging length as a function of pitch for three different scan lengths. The overranging length varies linearly with pitch, with R 2 values of 0.971, 0.942, and 0.979 for VOI scan lengths of 19.7, 25.1, and 30 cm, respectively. Overranging length ranges from 4.3 8 cm at a pitch of 0.5 and 6. 39 cm at a pitch of 1.5. Van der Molen et al. measured overranging lengths of 4. 26 cm at a pitch of 0.5 and 6. 43 cm at a pitch of 1.5. 96 Tzedakis et al. measured overranging lengths of 4.75 cm at a pitch of 0.5 and 7.45 cm at a pitch of 1.5. 94 The adjusted console readings were 3.55 cm at a pitch of 0.5 and 6 2 5 cm at a pitch of 1.5 Overranging lengths a s function of reconstruction slice thicknesses for two beam collimations of 12 and 24 mm are shown in Figure 4 4. Increasing the reconstruction slice width from 0.75 to 10 mm increased the overranging length from 11 .2 mm to 29.7 mm, respectively, using a b eam collimation width of 12 mm; increasing the reconstruction slice t hickness from 2 to 10 mm using a 24 mm beam collimation width

PAGE 103

103 increased the overranging lengt h increased from 25.4 mm to 52.5 mm, respectively. Note that the smallest reconstruction slice thicknesses for collimations for 12 and 24 mm are 0.75 and 2 mm, respectively. Two distinct regions are observed in Figure 4 4: a constant overranging length for smaller reconstruction slice thicknesses followed by a linearly increasing overranging length for larger reconstruction slice thicknesses. For example, the beam collimation of 12 mm has its constant region from 0.75 to 2 mm reconstruction slice thickness, where the overranging length remains 1.11 2.9% cm; overranging length is then linearly incr easing with an R 2 value of 0.99 between reconstruction slice thicknesses of 3 to 10 mm. Between the two linear regions, there is an abrupt 88% increase between reconstruction slice thicknesses of 2 and 3 mm. Similarly, for beam collimation of 24 mm, overranging length has its constant region from 2 to 4 mm reconstruction slice thickness, where the overranging length remains 2.55 3.3% cm; overranging length us then linearly increasing with and R 2 value of 0.87 between reconstruction slice thicknesses of 5 to 10 mm. Between the two linear regions, there is as an abrupt 77% increase between reconstruction slice thicknesses of 4 and 5 mm. Overranging length was independent of gantry rotation speed. For these measurements, Figure 4 5 shows the overranging length as a function of tube rotation time from 0.5 to 1.5 seconds per rotation, for three different pitches. Overranging length remained constant as a function of varying tube rotation times, even at different pitches, with measurements varying only 3.3 %, 2. 8 % and 1. 2 % for pitches of 1.0, 1.25, and 1.5, respectively.

PAGE 104

104 10% maximum at 1.85 and 3.85 seconds, respectively; these dosimeters were spaced 9.70 cm away. In other words, the table speed measured was 48.5 mm/s. The dosimeters were placed in the central region of the scan in order to avoid any acceleration and deceleration present at the beginning and end of the scans. This speed agrees well with the 19.7, 25.1, and 30 cm scans, which had s peeds of 48.4 mm/s. The bore dosimeter was above the 10% threshold from 0.55 to 5.80 seconds. The time of the scan is measured to be 5.25 seconds. The scanner selects a total scan length of 26.7 cm, with a predicted scan time of 5.24 seconds. This agrees well with the measured scan time. Analysis It has been long known that in helical scanning, volumes adjacent to the clinical VOI are also exposed to the primary beam. Selecting proper beam collimation, reconstruction slice width, rotation time and pitch c an help to minimize the detrimental effects of overranging lengths. Clinical benefit always outweighs the potential risk in clinical diagnosis scenarios. However, the appropriate VOI selection is dependent upon and there is certainly an amount of leeway especially near the VOI borders. The protocol must correctly select the VOI, while simultaneously ensuring correct alignment especially if the boundaries are close to radiosensitive organs. If for example, these clinical VOI boundaries are close to radiosensitive organs or tissues such as breasts, thyroid, eye lens, or gonads overranging may result in exposure to the primary beam from its inclusion in the overranging length, not to mention increased exposure to scatter. A s mentioned by Tzedakis et al. overranging

PAGE 105

105 lengths are frequently excluded altogether from deterministic simulations which only consider the clinical VOI, thus underestimating exposure. 93 95 Overranging lengths are particularly important considerations for deterministic Monte Carlo ca l cul ations of effective dose As noted by Tzedakis et al. typical CT simulations assume that the exposed length is equal to the VOI length. 93 95 In other words, while a simulation can accurately predict VOI dose, it will always underestimate DLP if it disregards the overranging lengths, which are always a part of a clinical scan. 45 The results obtained by a direct measurement methodology produce results which agree within 5% of measurements obtained from console readings and within 3% of an indirect measurement methodology. 93 96 Overall, a 5% difference in overranging length is not significant in the clinical context. For ex ample, the overranging length in this helical abdominal scan was found to be around 5 cm or a maximum difference (5%) of less than 3 mm. This could be attributed primarily to the finite 10 ms timing resolution and secondarily to measurement precision of VO I boundaries. The abdominal protocol at a reconstruction slice thickness of 5 mm, 120 kVp and 140 mAs has a CTDIvol measured by the ImPACT group of 13 mGy. 94 The 5% uncertainty is equivalent to a DLP of only 3.3 mGy cm. In terms of effective dose, a 3 mm margin would be difficult to incorporate primarily due to uncertainties from patient motion and positioning between the time of the scout scan where the VOI boundaries are defined and the act ual helical scan. T he accuracy associated with scout scan distance was tested following the QC procedure mentioned earlier with sub millimeter differences 98 It is important to consider the implications of an adjusted console reading. Specifically, if the console readings do indeed overpredict the overranging length, than

PAGE 106

106 the DLP is similarly too large. In terms of calculations, thi s inaccuracy must be corrected, but clinically, this can be viewed positively, in that volume exposed was smaller than originally calculated. When compared with the adjusted console measurements, the new methodology produces results which are always larger from 2% (at pitch of 1.5) to 20% (at pitch of 0.5). to the longitudinal axis in order to better approximate a point detector. Theoretically, irradiation anywhere along producing an uncertainty in the VOI measurement. In the worst case scenario of uniform response from the scintillator, unlikely due to the cylindrical geometry, this finite diameter imparts an uncertai nty of 0.5 mm, well below the limiting uncertainty imparted by the temporal resolution of 10 ms. Analysis of the temporal response was initially conducted based on the leading edges of the detected radiation peaks. Further analysis using both leading and trailing edges on the temporal response peaks resulted in similar results within 2.4% of the exclusive leading edge technique. Subsequently, the distance between the two table dosimeters was measured and divided by the average difference in leading and t railing edges of the table dosimeter temporal response in order to give the table speed. The table velocities were calculated from the quotient of the differences in the leading edge times obtained previously and the known distance between table dosimeter s, with R 2 values of 0.999 reproducible for the three VOI scans lengths performed (19.7, 25.1, and 30 cm). These values agree to within 1% to the table velocity selected on the console. The table dosimeters provide a measure of average

PAGE 107

107 velocity over a sm all region whereas the console measurement of table velocity is based on overall scan length and overall scan time. The table velocity measured over a small region should agree well with the table velocity measured over the entire scan by the console, whic h implies that the acceleration and decel eration regions are negligible. Figure 4 3 shows the linear relationship between overranging length and pitch for different scan sizes, with R 2 values of 0.971, 0.942, and 0.979 for VOI scan lengths of 19.7, 25.1, and 30 cm, respectively. Our methodology thus demonstrates the expected independence of overranging length upon different individual clinical VOI lengths; this will have obvious effects upon normalizing the (constant) overranging lengths to the (varying) clinical VOI lengths. The DLP measured for the three scan lengths of 19.7, 25.1, and 30 cm were 256, 326, 390 mGy cm. Figure 4 3 shows overranging lengt hs ranging from 4.38 to 6.39 cm, w ith an average around 5.5 cm; these correspond to DLP of 131 and 192 mGy cm, and an average of 165 mGy cm. The results are summarized in Table 1. One advantage of this new methodology is its complete independence from console measurements. Specifically, th e methodology prescribes a measurement of table velocity for each and every scan instead of relying upon the velocity and time calculated by the console. The Siemens scanner itself predicts the time required for the the scan is actually done. This is presumably done with a speed look up table which depends on a combination of tube rotation time, pitch, detector collimation, and reconstruction slice thickness. It is possible to refine the methodology to operate with o nly the bore dosimeter and eliminate the table dosimeters if independent speed measurements are not desired. If

PAGE 108

108 the system is operated in one dosimeter mode, the degree of precision and accuracy is exceptional and is limited only in the time resolution of one detector. With our current design using 10 ms resolution, the uncertainty is less than one half of a millimeter (for pitch = 1 the table velocity is ~ 48 mm/s). The console reading measurements and physically measured distances must differ by no more t han 1 2 mm as stipulated by QC guidelines, therefore the maximum error would be seen in a smaller scan, and is 1.1% for a 19.1 cm scan. In fact, this impressive number could be further improved by faster PMT binning. Using the one dosimeter mode would make it much simpler to use clinically and has exceptional uncertainty; but, as discussed, this will eliminate the complete independence of the dosimetry system. Figure 4 4 shows two distinct linear regions for smaller and larger sections this reconstruction slice width dependence has been measured by both van der Molen et al. and Tzedakis et al. 93 96 These two regions arise from different reconstruction schemes used by Siemens scann ers for smaller (cone beam corrected mode) and larger sections (z filtering). 96 beam thicknesses (0.75 2 mm reconstruction slice thickness for 12 mm beam collimation; 2 to 4 mm for 24 mm) while the faster S reconstruction slice thicknesses (3 10 mm for 12 mm; 5 10 mm for 24 mm). This methodology demonstrates the linear dependence of overranging length on pitch which has been observed in previous publications. 93 96 With a routine abdominal protocol, overranging length is clinically relevant and can be expected to contribute an average of 1 0% extra dose, or a DLP around 6 0 mGy cm (30 cm scan length with a CTDI vol of 20 mGy) Overall, the overranging lengths for varying pitch, reconstruction

PAGE 109

109 slice width and beam collimation measured using this methodology agree within 5% with measurements by van der Molen et al. 96 Further agreement in overranging length s was observed with console reading results obtained by Tzedakis et al. for 12 mm beam collimations; however, the overranging lengths obtained with 24 mm beam collimations were slightly lower than the console reading results obtained by Tzedakis et al. by around 5 10%. 93 95 Regardless, assuming a routine abdominal protocol, a 5% discrepancy in overranging length is equivalent to a DLP of only 3.3 mGy cm. The overestimate of effective dose by Monte Carlo methods is mentioned by van der Molen et al. and attributed to the simulated beam profile, which does not produce any noticeable differences until a larger beam collimation of 24 mm is used. 96 Thi s is discussed in more detail by van der Molen et al. 96 However, the CR exposure pattern test perform suggests that Tzedakis overestimated the overranging length simply because a whole beam collimation width was added rather than one half of a beam collimation width. Gan try rotation speed was not expected to contribute to overranging length because the actual scan length required is determined by the amount of reconstruction information required If the reconstruction scheme remains constant, this should remain fixed In other words, it should not change depending upon the speed of the gantry, unless the reconstruction scheme changed as a function of gantry rotation speed. This parameter was included in order to measure overranging for any operator controllable parameters This methodology only involves one physical measurement: distance betwee n scintillators. According to QC guidelines, any discrepancies between measured and

PAGE 110

110 physical distances should be no more than 1 2 m m. 98 However, this was explicitly tested by comparing distance measured with a calibrated linear measurement device between two lead bars against a measurement from the operator difference was less than 0.50 mm. The remaining portion of the methodology is based upon the temporal spectrum obtained, where the limiting resolution of this system is 10 ms which using a typical (pitch = 1) table speed of 30 mm/s yields an uncertainty of 0.28 mm 17 Assuming even the highest table v elocities (pitch = 1.5) the uncertainty is slightly less than 0.50 mm. Thus assuming a high table velocity, the maximum uncertainty of the methodology is estimated as 0.7 mm which is obtained by the sum of the console measurement precision (0.5 mm) and the timing measurement precision (0.5 mm), added in quadrature. There is good general agreement with other published techniques. However this et al. and also addresses a significant error m ade by Tzeda kis et al. Van der M olen et al. extrapolation method with a linear fit confidence interval of 0.95. The precision of the data used in order to perform this extrapolation was not discussed. Therefore, it is assume d that the error ba rs on those data were considered negligible by van der Molen et al. giving a precision on the order of 5% compared with maximum uncertainty of this new methodology of 1.1%. Discussion The methodology for direct measurement of overranging is an example of an application of the real time capabilities of this point dosimetry system. In this methodology, using two table dosimeters provided a measurement of average velocity over a small region whereas the console determines table velocity based on overall

PAGE 111

111 scan length and overall scan time. The table velocity measured over a small region agrees well with the overall measured table velocity, which is important because it demonstrates that the acceleration and deceleration regions are negligible. If implemented cli nically, this method could potentially be adapted to operate in conjunction with the console measurements of speed, thus eliminating the need for the two dosimeters. This would also reduce the number of measurements required, but introduce a strong code pen dence upon console accuracy. This method is limited in that it may not work for modern scanners which utilize dynamic collimators. However, neither the console readings by Tzedakis et al. nor the extrapolation method by van der Molen et al. would work in t his situation. A new method would need to be developed in order to account for the partially open or closed collimators. Overranging length can result in quite significant contributions to patient exposure. For a routine abdominal protocol, overranging len gth was determined to be clinically relevant with an average contribution of 10% extra dose, or a DLP around 165 mGy cm. In spite of this, DLP is traditionally measured only for the VOI, omitting overranging length. This technique has an advantage over c on sole reading simulations because it is scanner independent and does not require benchmarking or confidential proprietary information notably bowtie filter spectra ; furthermore, this method requires a single measurement while current physical methods require multiple measurements to establish an extrapolation baseline; lastly, this method avoids assumptions regarding immediate irradiation during table translation.

PAGE 112

112 This quick and direct methodology can be easily implemented in order to include overrang ing lengths for proper calculations of total DLP. Overranging length results have been measured for a Siemens Somatom Sensation 16, future research will include different vendors, scanner models, and post processing methods which have been shown to chang e overranging values by as much as 125%. 96

PAGE 113

113 Table 4 1. Measured overranging values [mm] for various techniques

PAGE 114

114 Figure 4 1 Schematic of PSD placement Figure 4 2 Sample responses for bore PSD and dosimeter PSDs

PAGE 115

115 Figure 4 3 Overranging length as a function of pitch Figure 4 4 Overranging length as a function of reconstruction slice width

PAGE 116

116 Figure 4 5 Overranging length as a function of x ray tube rotation time

PAGE 117

117 CH APTER 5 QUANTIFICATION OF ST ARTING ANGLE DOSE BI ASING IN HELICAL COM PUTED TOMOGRAPHY Background In helical CT, the x sinusoidal dose profile found in the helical dose profile. Newly innovated techniques known variously as starting angle methods or organ dose modulation, manipulate the pha se of the starting angle in order to pair the troughs of the periodic dose profile with a particular radiosensitive region. 4, 24, 25, 101, 102 This technique has achieved 60% organ dose reductions in the lens of the eye in computational studies by Zhang et al. 25 Organ dose modulation offers these dose savings come with almost no loss to image quality. Specifically a study by van Straten et al. concluded that for two given helical scans with two different starting angles, the images were deemed identical for all practical purposes by radiologists. 27 D ose reductions from organ dose modulation have not been realized because the x ray tube sta rting angle is both unpredictable and uncontrollable in any CT manufacturer, including popular vendors such as GE, Philip, or Siemens Recently, Toshiba has implemented one cardiac proto col in which the beam is gated to the cardiac cycle and is only on for a portion of the rotation Still no machines allow direct user specification of x ray tube starting angle This investigation focuses upon measurement uncertainty and its impact upon clinical dosimetry In contrast, Chapter 6 will derive the actual dose savings. Direct physical measurements of the helical dose was not possible given the random distribution of x ray tube starting angle. In the method presented, a mathematical dose expression on helical CT scan dose by Dixon et al. was further

PAGE 118

118 developed to calculate the helical dose for any given starting angle. 21, 99 The resultant expression decomposed the helical response into axial and longitudinal responses; after phas e shifting, these responses we re retroactively joined to represent a helical scan with any arbitrary starting angle. Starting angle dose bias (SADB) is a new metric which is introduced in order to express the observed difference in dose measurements as a d irect result of different x ray tube starting angles. The SADB calculation wa s validated with a cylindrical phantom and applied to an anthropomorphic phantom. Materials and Methods The first part of this study focuse d on dose measurements as a function of x ray tube starting angle by quantifying SADB as a function of pitch, detector length, detector depth and beam collimation. These were used to scale the longitudinal aspect of the helical dose expression. These dose profile responses were collected with a real time plastic water equivalent scintillating point detector system previously characterized by Hyer et al. 17 The second part of this study wa s validation of the dose expression, by measuring the axial responses in a cylindrical phantom homogenously composed of the soft tissue equivalent substitute (ST ES) material previously characterized by Winslow et al. 10 This STES material mimic ked the attenuation characteristics of soft tissue at diagnostic energies. The last part of this study wa s measuring dose in an anthropomo r phic phantom described in Chapter 2, which use d both bone equivalent and lung equivalent materials in addition to the STES. The organ doses wa s measured to all ICRP specified organs of interest for axial scans and derived for the helical s can. Helical organ doses we re calculated for each organ and all possible starting angles. The maximum, minimum and

PAGE 119

119 standard deviation in these measurements was due strictly to the starting angle. The total effective dose wa s calculated using the default be am parameters for five major scan protocols: head, neck, chest, abdomen, and pelvis. Patient positioning and region of interest (ROI) were determined using anatomical landmarks as approved by the American Society of Radiologic Technologists. 103 These are summarized in Table 5 1. CT Scanner and CTDI Body Phantom A Siemens S OMATOM Sensation 16 multislice helical CT scanner Siemens AG, Forchheim, Germany) was used in service mode Service mode additionally allowed the use of single axial scans, which wa s used to create the axial dose profiles. Prior to the helical scan, a scout image, a single planar image wa s typically obtained to delineate the region of interest. The scout image, which is an image taken with a stationary x ray tube and the tab le translating longitudinally, wa s used to create the longitudinal dose profiles. The helical scan protocol used for the cylinder was: 160 kV, 120 mA, pitch = 1, 0.5 second tube rotation time, 5 mm re construction slice width, and 24 mm (16 x 1.5) detector collimation. The protocols used for the anthropomorphic phantom are summarized in Table 5 2 Tube current modulation was not used in this study. The computed tomography dose index (CTDI) phantom has t raditionally served as the standard for measuring dose in helical CT. 45 The CTDI body phantom is a Lucite cylinder of 320 mm diameter and 150 mm height with five holes one in the center and four peripheral holes designed to accommodate a pencil ion chamber. 45 Measurement Phantom s A cylindrical phantom of similar dimensions was used for this study which was composed homogenously of the STES described in Chapter 2. The pha ntom was custom

PAGE 120

120 variety of different combined heights. On e of these cylindrical disks had a height of 30 mm while the remai ning four cylindrical disks had a h eight of 40 mm. Each pancake had a diameter of 320 mm with 6 mm slits along the radius in order to accommodate the dosimeters. In total, the assemble d phantom had the dimensions of a CTDI phantom: a diameter of 320 mm and a total height of 150 mm. These slits ha d 5 mm in crements marked for reproducible placement of the dosimeter. The isocenter of the phantom wa s aligned with the isocenter of the CT gantry. T wo views of t he STES cylindrical phantom are shown in Figure 5 1 The STES cylinder phantom was used instead of a CT DI Lucite phantom for four reasons: first and foremost, any dosimeter depth along the 320 mm diameter was possible. In contrast, a CTDI phantom is designed to accommodate only three depths in its transverse plane: one at isocenter and two peripheral depths Secondly, the STES material was a pliable medium which conformed around the dosimeter, as opposed to pencil ion chamber. Third, because it was constructed as a series of short cylinders, it had an adjustable height which allowed for easier access to central slices. Lastly, the STES material was designed to be soft tissue equivalent and will produce more clinically realistic dose measurements as opposed to the Lucite mater ial used in the CTDI phantom. Initial performance testing was done with a homogeneous STES cylindrical phantom. In order to more accurately represent the human body, the male 50 percentile reference anthropomorphic phantom described in Chapter 2 was used f or the next stage. As mentioned previously, the anthropomorphic phantom is composed of BTES,

PAGE 121

121 LTES in addition to STES. O rgan doses were measured to all ICRP specified organs of interest The number of mea surement points for each organ wa s specified previou sly in Table 2 1 Each desired measurement was taken for an axial scans and derived for the helical scan using the method below, which was simply an expansion of Dixon et al. formulation. 23 Original Helical Dose Profile Expression Dixon et al. produced the theoretical derivation for decomposing the helical dose profile into two independent constituent dose profiles, namely the axial and longitudinal dose profiles. 23 For a simple axial one slice scan, the dose distribution possesse d a sinusoidal shape whic h was periodic with the tube rotation time. In a more clinically applicable situation such as a helical scan, the tube continue d this periodic rotational motion in the axial plane while the table wa s simultaneously translating along what is conventionally known as the z axis or the longitudinal axis. In the methodology by Dixon et al. the total dose at a point located at longitudinal position, z, and depth in the axial plane, d wa s written as which wa s determined for a helical scan with a inst antaneous dose rate, As is the case in clinical cases, the table wa s translating along its longitudinal axis at constant velocity, for a scan time, Its single slice axial dose profile, wa s obtained with a x ray tube rotation time Introducing a simplifying change of variables, and including a longitudinal dose profile with an effective length of L written as Dixon et al. expressed the accumulated total dose along the axis of rotation as 23 (5 1)

PAGE 122

122 w here represents accumulated dose along the isocenter. Dixon et al. use d a step function defined with a width equal to the effective length L to represent the beam profile. 23 Extended Helical Dose Profile Expression In the extension of this me thod, a longitudinal beam profile, g L wa s used. The beam profile ha d a full width half maximum roughly equal to L and also include d scatter tails, which represents the overall effect of the beam upon the dose measured better than a step function. The beam profile wa s further rewritten as the convolution between the originally measured longitudinal dose profile, and a effective width step function with a length of L written as (5 2) This effective width step function, wa s used to scale the original measured longitudinal dose profile to represent different aspects of the beam profile such as detector length, beam collimation, or pitch. The effective width step function wa s useful for two reasons. First, it decreased the nu mber of measurements needed by convolving a single beam profile to represent varying different beam collimations or pitches. Secondly, it allowed larger detector lengths to be characterized by the response of the point detector through convolution. Incorpo rating the effective width step functions, the total dose along the isocenter wa s rewritten as (5 3)

PAGE 123

123 For the case of dose along the isocenter, a dependence on x ray tube angle, must be included in the singl e slice axial dose profile; furthermore, the beam profile must be rewritten for a specific depth in the axial plane, written as For example, i n at isocenter can be rewritten as Th us, a more general expression for total dose for an effective length at any ax is of rotation wa s rewritten as (5 4) In this general expression, the single slice axial dose profile now accounts for longitudinal position, angle, and axial depth of the detector. Using a symmetrical homogenous cylindrical phantom should eliminate this angular dependence, but angular dependence was quantified as part of the axial dose profile characterizat ion. Since all measurements were taken in the same longitudinal location in the center slice of the phantom, there wa s no longitudinal dependence. The dependence upon axial depth and effective length still remain, but E quation 5 4 wa s reduced to (5 5) In this case, the axial dose profile and longitudinal dose profiles were measured separately and combined in order to form a helical dose profile. By shifting the axial dose wa s obtained. Similarly, shifting the axial dose profile exactly 180 de grees out of phase with the peak in the beam dose profile, the minimum helical dose profile wa s

PAGE 124

124 obtained. Prior to convolution and integration, these dose profiles we re initially constructed as a function of time using the real time dosimetry system and la ter scaled to an arbitrary index. Helical dose wa s calculated by combining single slice axial responses and longitudinal r esponses as shown in Equation 5 5. Axial Dose Profile Characterization Temporal response was sampled at 10 msec intervals for both 16 x 1.5mm and 16 x 0.75 mm detector collimation. The time axis wa s converted to angular variation given the selected tube rotation speed. Over one rotation, the real time measurement thus yield ed the entire range of angles which are possible during a helical scan. A sample temporal axial response of the dosimeter in the cylindrical phantom is shown in Figure 5 2 Due to table attenuation, notably in its guide rails, dips were observed in the axial dosimeter response. If repeated, this axial measurement would be a repeating periodic function. The uncertainty which would be due to starting angle was quantified by calculating root mean square ( RMS ) of the counts as the analysis metric over one rotation. These counts were converted to dose by applying a dosimeter calibration factor. The RMS value gives a measure of the variation of the magnitude of a periodic function and for this application serve d as excellent metric for the uncertainty due to starting angle. The RMS value was calculated as a function of varying depth along both the AP and lateral axis for the axial cylinder measurements. In addition to the RMS value, which measured just the variation of the dose, the actual total dose was calculated. These RMS measurements used realistic clinical protocol values for the abdominal routine scan for applicability to common dosimetric methods: 160 mA, 120 kV, and tube rotation time of 0.5 seconds.

PAGE 125

125 T he RMS values were also be calculated for dosimeters at different angular orientations of the phantom in order to valida te the assumption of angular independence utilized in E quation 5 5 above. In other words, at any fixed distance from the isocenter in the plane of the bore, the response should be the same. This was tested by measuring response as function of constant dist ance from isocenter for two points along both anterior posterior and left right axes. Cylindrical Phantom Dose Axial dose responses were collected in the cylindrical phantom at depths ranging from 1 to 31 cm, in 1 cm increments. Using the real time dosim etry system and a simple scout scan, or topogram, the longitudinal dose beam profile was obtained by measured with a dosimeter at the given depth with 10 ms resolution. A sampl e response is shown in Figure 5 3 These profiles were collected at depths rangi ng from 1 (surface) to 16 cm (isocenter). This longitudinal beam profile is much more realistic than the step function used by Dixon et al. : containing a peak which is not uniformly flat, peak penumbras, and scatter tails. 23 A single axial scan, by definition, does not include scatter contributions from adjacent slices because it only contains one slice. These penumbras and scatter tails defined the scatter contributions from adjacent slices in the final helical dose response s. The time which a given detector is in the primary beam was determined by the width of the peak in the longitudinal beam profile. As mentioned earlier, Equation 5 5 has a component, the effective width step function with a length of L written as w hich was used to scale the original measured longitudinal dose profile to represent different aspects of the beam profile such as detector length, beam collimation, or pitch. Note that a measured beam profile with a high temporal resolution can be rescaled to

PAGE 126

126 different pitches by appropriate extent based upon the amount of distances covered by the beam. This distance at different depths was determined given the focus isocenter distance of 570 mm and the total effective length of detector array at isocenter (which In order to determine the maximum helical response, the axial profile peak was shifted in phase with the longitudinal profile peak and combined using Equation 5 5. Physically this re presented the x r ay tube position such that the beam wa s directly incident upon the detector as it passe d through the beam with the minimum attenuation between it and the x ray tube. For the cylinder, the minimum helical response wa s obtained when the axia l profile peak is shifted 180 degrees out of phase with the longitudinal profile in Equation E6. Physically, this represented the x ray tube in a position directly opposed to the beam with the maximum attenuation between it and the x ray tube. This phase s hifting is the premise of organ dose modulation. Need for Starting Angle Dose Biasing Metric The starting angle dose bias (SADB) is used in order to express the maximum observed difference in dose measurements from differing x ray tube starting angles. SAD B is greater than or equal to unity. It is defined as the quotient of the m ax imum dose possible divided by the m in imum dose possible. SADB was measured for effective width step functions ranging from 1 to 100 elements wide, to represent different beam coll imations, detector collimations, and pitch. SADB was determined by analysis of the helical dose response. For example, a 60% dose reduction was equivalent to a SADB of 1.67, determined by 1.0 divided by 0.4. As ei ther organ size increased or organ location became more isocentric, the SADB approached unity in other words the maximum dose and minimum dose were equivalent. 25 This

PAGE 127

127 represented a scenario where the starting angle has no effect For a larger organ, as the number of rotations increase, the uncertainty due to starting angle begins to be averaged out 25 On the other hand, a large dose savings can be obtained by manipulating pitch and detector collimation: higher pitch and larger detector collimation creates a longer low dose troug h with which could be pair ed with a larger radiosensitive region. Anthropomorphic Phantom Dose In the cylindrical phantom, the maximum instantaneous dose was delivered when the x ray tube is closest to the dosimeter; in the same way, the minimum instantaneous dose was delivered when the x ray tube is closest to the dosimeter For the cylindrical phantom, these two positions are a lways separated by 180 degrees. In the anthropomorphic study, this wa s not necessarily the case because the phantom was representative of a human and thus was heterogeneous in both composition and shape. The distance wa s better measured in terms of attenua tion pathlength. Contribut ions from LTES, BTES and air mad e the attenuation pathlength a unique function of angular and longitudinal position For example, the thymus a superficial point in the thick shoulder region has its maximum dose measured when t he x ray is directly anterior on the patient; however, its minimum dose c a me from a lateral position. Measurements of the anthropomorphic phantom provide d clinically relevant data which use d the default technical imaging settings parameters for five major CT scan protocols: head, neck, chest, abdomen, and pelvis. These exam parameters were previously summarized in Table s 5 1 and 5 2 However, the exam parameters listed by Romans et al. did not include table height. 103 Therefore, f or head scans, the phantom was aligned with the center of the brain at isocenter of the beam. For chest scans, the

PAGE 128

128 isocenter was placed in the center of the body at an axial plane near the center of the lungs. F or pelvis scans, the center of the prostate was placed at isocenter. To ensure reproducible results, the phantom was placed in a vacuum immobilization bag and any remaining gaps were eliminated using medical tape. As mentioned in Chapter 2, t he anthropomor phic phantom was set up and imaged in a close approximation to a real patient for the particular anatomical region being treated by reproducing the organ location from the original hybrid dataset. The exact measurement location for the scintillating elemen t was determined by dividing the organ by volume. L arger organs were subdivided into smaller segments with measurement locations for the centroid of each of these smaller segments. For the smaller organs, absorbed dose values were given by the measurement at the centroid. The number of measurement points was summarized previously in Table 2 1 For the larger organs, absorbed dose was given as the average of its constituent volumes. Measurements were taken as specified for calculation of effective dose by Publication 103 of the International Commission on Radiological Protection with the exception of bone surface, bone marrow, lymphatic nodes, skin, and muscle. 44 Each organ which was measured is explicitly listed in Table 5 3 along with the number of measurement points used Bone surface, bone marrow, lymphatic nodes, skin and muscle were excluded because it wa s physically difficult to obtain an average organ dose measurement and were not expected to have significant SADB values. Specifically, starting angle effects were already been shown to average out for organs which are either large longitudinally, or centrally located by Zhang et al. 24, 25, 83 Additionally, skin a nd bone surface have a weighting factor of 0 .01 which minimized

PAGE 129

129 potential impact from their exclusion. Muscle and lymphatic nodes are categorized as remainder tissues, along with 1 1 other tissues, thus minimizing any impact on effective dose. In the effect ive dose calculation, the average of the other 11 tissues was used as a suitable replacement. Ef fective dose was chosen as a metric because it is a simple calculation which is commonly used to compare between different procedures. 45 The axial scan wa s combined with the lon gitudinal scan using Equation 5 5 and all possible helical doses for differing start ing angles we re determined. As dictated by Equation 5 2 t he longitudinal beam profile was applied to incorporate beam penumbra and scatter E ach possible starting angle (limited by the temporal resolution, 10 ms) was simulated and the helical dose was determined. In order to account for geometric beam divergence, effective beam width functions such as beam collimation at isocenter, focal spot to isocenter distance, and distance from isocenter were simulated. Finally, the average and standard de viation of all the possible helical doses for different starting angles were found. The effective dose wa s calculated using the organ weightings specified by Publication 103 of the ICRP. 44 To study the effect of starting angle, the standard deviation of all organ dose measurements from varying starting angles is calculated using standard error propagation. Results SADB Measured in Cylindrical Phantom Dose as a function of time for a single axial scan was collected at depths ranging from 1 to 31 cm, in 1 cm increments. An axial dose profile was shown at a depth of 5 cm in Figure 5 2 The ordinate axis was con verted to angular position of the x ray tube given the tube rotation time in this case, 0.50 sec per revolution which produced a

PAGE 130

130 periodic function which depends on the angular position of the rotating x ray tube during a scan. The isocenter of the cyl indrical disks was aligned with the isocenter of the beam; in other words, a depth of 0 cm represented a surface location and a depth of 16 cm represented the both the central axis of the cylinder and the axis of rotation of the gantry. Both d=0 and d=32 c m were not calculated due to physical limitations of the dosimeter construction: at d=0 cm, the fiber was not able to be positioned on the surface of the cylinder similar to its position for the rest of the scans with the tip of the fiber perpendicular t result of the crimping required in the fabrication, required an additional 5 mm of clearance. The RMS values were calculated for depths in the cylinder ranging from 1 to 31 cm in 1 cm increm ents and are shown in units of dose in Figure 5 4 Total dose values ranged from 1.35 to 2.29 mGy and RMS dose values ranged from 0 .02 mGy at the isocenter to 0 .14 mGy at the surfaces. Validation of Dose Expression Equation 5 5 was tested by measuring the axial response as a function of distance from the isocenter in the plane of the bore for two points along both anterior posterior and left right axes. An intermediate radius of 6 to 11 cm was chosen and. This is shown in Figure 5 5 The data p oints are sum marized in Table 5 4 There was a maximum 0 .2 mGy variation in dose to STES. The root mean square dose is also shown in Table 5 5. The longitudinal response profile was obtained at depths ranging from 1 to 16 cm. A sample beam profile normalized to maximum response is shown at depths of 5 cm in Figure 5 3 The time axis was converted to a distance using the known topogram speed. The spacing between points is linearly interpolated to match the distance

PAGE 131

131 spacing of the axial profile. For each depth, the measured beam profile was then convolved with different effective width step functions from 1 to 100 elements wide. Figure 5 6 shows the SADB which wa s calculated for depths ranging from 1 to 16 cm and eff ective width step functions ranging from 1 to 20 elements wide. For each point, the helical response function wa s generated from a shifted axial response profile measured combined using Equation 5 5 with the result of the convolution of longitudinal respon se profile and effective width step function. SADB was near unity for nearly any detector located at isocenter. SADB reached a maximum of 5.70 for a 2 mm point detector located at a depth of 1 cm Increasingly isocentric positioning and longer detector le n gths produced SADB values which converge to 1.00, meaning the maximum and minimum doses delivered were the same regardless of x ray tube starting angle. SADB Measured in Anthropomorphic Phantom The axial scan is combined with the longitudinal scan using E quation 5 5 and all possible helical doses for differing starting angles are determined with resolution determined by 10 ms resolution and tube rotation time. The average and standard deviation of all the possible helical doses for different starting angle s are shown in Tables 5 6 and 5 7, respectively In addition, Table 5 8 shows the effective dose calculated using the organ weightings specified by Publication 103 of the ICRP 44 Note that Tables 5 6 through 5 8 do not include the lens of the eye because these are not included in ICRP 103. To study the effect of starting angle, the standard deviation of all organ dose measurements from varying starting angles wa s calculated using standard error propagation. If the measured dose was less than 1.0 mGy, it was not recorded. For example, while a measurement point was used for the organ dose to the brain for the thyroid

PAGE 132

132 exam was ignored in the t able. The average SADB value is also shown for each organ for each protocol. The largest standard deviation, in terms of percentage, is measured in the organ dose delivered to the salivary glands was 27.5 +/ 2.9 mGy, or 1 0.6% in a head protocol This was followed by the thyroid organ dose of 27.3 +/ 2.5 mGy, or 9.2%. Analysis Cylindrical Measurements The STES phantom was used instead of a CTDI Lucite phantom for four reasons: first and foremost, any dosimeter depth along the 320 mm diameter was possible. In contrast, a CTDI phantom is designed to accommodate only three depths in its transverse plane: one at isocenter and two peripheral depths. Secondly, the STES material was a pliable medium which conformed around the dos imeter, as opposed to pencil ion chamber. Third, because it was constructed as a series of short cylinders, it had an adjustable height which allowed for easier access to central slices. Lastly, the STES material was designed to be soft tissue equivalent and will produce more clinically realistic dose measurements as opposed to the Lucite material used in the CTDI phantom. Cylindrical P hantom Table 5 4 shows that total dose increases as distance from isocenter increases. Table 5 5 shows the same trend that RMS dose increases as distance from isocenter increases. Note that data for Table 5 5 were smaller than expected from typical CTDI measurements becau se measurements were made in a larger STES cylindrical phantom and were also done with an abdominal protocol with current lowered by the

PAGE 133

133 operator In other words, superficial locations have both higher dose and higher RMS dose in comparison with locations near isocenter This wa s expected from both higher geometrical attenuation as well as substantial attenuation by the medium for the centrally located points. Figure 5 2 shows the highly symmetrical shape of the dose distribution centered about the isocente r, which wa s located at a depth of 16 cm. The dependence of dose and RMS dose upon angular variation was important to examine because t he assumption of angular independence wa s required in Equation 5 4 a valid assumption given the homogenous composition a nd symmetry in the cylindrical phantom. Table 5 4 also shows the maximum difference between the doses measured from angular variation was 7.6%. Table 5 5 shows the maximum difference between RMS doses was 6.3%. This was determined to be due to precision in PSD placement. B oth Table 5 4 and Table 5 5 showed a n average 1 6 % change in total dose per 10 mm and 15% in RMS dose per 10 mm, respectively. The uncertainty in PSD distance from isocenter was estimated to be around +/ 3 mm due to uncertainty in isocenter localization as well as PSD placement within the phantom. This aspect of the investigation of cylindrical phantoms thus showed the independen ce of both total dose and RMS dose from angular variation These doses are dependent upon only absolute distance from isocenter in a cylinder. This thus validate d the assumption made in Equation 5 4 The width of the effective width step function used in Equation s 5 2 through 5 5 wa s determined by the time in the primary beam The time elapsed was a function of the beam collimation, detector collimation, pitch, and detector length. This was derived isocenter distance of 570 mm and the total

PAGE 134

134 effective length of 24 mm of detector array at isocenter which was located 16 cm deep in the cylindrical phantom B eam profiles were shown normalized because the purpose wa s to calculate SADB, which wa s itself a ratio. Also, the profile from 17 to 32 cm will be identical to that measured from 1 to 16 cm due to the angular independence confirmed earlier. Thus the response from 17 to 32 cm was not included in the figures. Next, the scout scan was incorporated into the effective wid th step function For example, assuming an isocenter of 16 cm, the effective width at 17 cm depth would still not be the same the scout scan at 15 cm depth While both are the same distance from the isocenter, due to geometric beam divergence as well as at tenuation the scout scan results made the results seem different This can be ignored because t he physical representation of the helical derivation explain ed that in a helical scan, the x ray tube w ould actually be rotating, so they would be equivalent and were represented by a phase shift. As seen in Figure 5 6 random x ray tube starting angle introduced a large (80%) range of possible helical dose for a cylindrical 2 mm detector at a depth of 1 cm. SADB converge d to unity as the position became more isocentric. In other words, starting angle had no effect upon measured precision. As depth approached isocenter there were almost no dose variations and the effect of varying x ray tube starting angle be ca me negligible. This reduc tion in the effect of starting angle was also seen by Zhang et al. 24, 25 Because SADB was calculated by shifting the axial response peak directly in phase wit h the longitudinal response peak, it wa s actually possible for SADB to be less than 1 for large effective w idth step functions. This occurred if the longitudinal response peak is wide enough to encompass the trough of the axial response peak and the two

PAGE 135

135 su rrounding peaks. This can occur most notably with a low pitch. This was physically interpreted as if the effective length becomes s o long that the detector crossed the beam twice. Anthropomorphic P hantom In contrast with the cylindrical phantom, the pathl ength through the anthropomorphic phantom was much more complicated than a homogenous STES cylindrical phantom. In addition to contouring which more accurately represented a human, c ontributions from LTES, B TES ma d e the attenuation pathlength a unique func tion of both angular and longitudinal position For example, the thymus an anterior point in the thick shoulder region would presumably have had its maximum dose measured when the x ray wa s directly anterior on the patient; however, its minimum dose wa s not 180 degrees away, but instead from a lateral position. Examination of Table 5 7 shows that t he largest one sigma standard deviation, as measured with the standard deviation in dose measurement, was 9% in the thyroid. The smallest SADB observed was 1.27 Zhang et al. has also specifically designated the thyroid as an excellent candidate for dose reduction using starting angle methods because it is both superficially located and small. 24, 25 Zhang simulated organ dose values at pitches of 0.75, 1 and 1.5. The measurements taken in this investigation were dard protocol values which were all below un ity. If there was no concern about image quality, the dose reductions could be increased by increasing pitch. Overranging Considerations As discussed in Chapter 4, o verranging lengths are the volumes adjacent to the clinical volume of interest which must b e collected in a helical CT for proper

PAGE 136

136 reconstruction In an effort to minimize the effect of starting angle, phase shifting could be specified using control of the x ray tube starting angle. This shift must account for the overranging present in the CT sc an. For example, if it was determined that a 30 degree x ray tube starting angle was optimal for SADB determination, the tube would need to be turned on around 20 degrees for a 30 cm scan, depending upon the table speed. This investigation of SADB was designed in order to cons ider every possible starting angle. It was not designed to find one optimal angle, rather it used brute force in order to calculate the average SADB for a given point measurement. Therefore, o verranging considerations are irrelevant for these physical measurements but need to be incorporated into clinical algorithms for accurate implementations. Discussion SADB represents a source of uncertainty which has been previously ignored. Specifically, the majority of computational models of helical sources take only one measurement of dose at a constant starting angle position. 25 X ray t ube starting angle introduced a large (80%) range of possible measured helical dose for a cylindrical 2 mm detector at a depth of 1 cm. A new metric of SADB was introduced as the quotient of the maximum dose possible divid ed by the minimum dose possible. At a certain depth, SADB converged to 1.00, which represented no dose variations due to different x ray tube starting angles. While dose reduction through organ dose modulation has been examined, the ac companying loss of precision due to the increase in SADB has not been accounted for. In other words, physical measurements will always have some baseline precision

PAGE 137

137 which cannot be overcome without hundreds of measurements in order to eliminate SADB through measurement of the entire range of possible doses. A n innovative and robust method was presented for physical measurement of the bias inherent in CT dose measurements due solely to differing x ray tube starting angles. Addressing this dose uncertainty is important for all precise measurements of organ dose, especially dosimeter research, clinical exposure logs, or Monte Carlo simulations.

PAGE 138

138 Table 5 1 Anatomy used for determining protocol scan ROI Table 5 2 Default beam parameters for different protocols

PAGE 139

139 Table 5 3 Number of measurement points used for each protocol

PAGE 140

140 Table 5 4 Total dose [mGy] to soft tissue as a function of phantom orientation Table 5 5 Root mean square dose [mGy] to soft tissue as a function of phantom orientation

PAGE 141

141 Table 5 6 Average measured organ dose [mGy] to soft tissue

PAGE 142

142 Table 5 7 Standard deviation in measured organ dose [mGy] to soft tissue

PAGE 143

143 Table 5 8 Effective doses for different pro tocols

PAGE 144

144 Figure 5 1 Two views of the STES cylindrical phantom Figure 5 2 Sample temporal axial response of PSD in cylindrical phantom

PAGE 145

145 Figure 5 3 Sample longitudinal response of PSD in cylindrical phantom Figure 5 4 RMS dose as a function of depth in phantom

PAGE 146

146 Figure 5 5 Single slice total dose variation between PSD orientations Figure 5 6 SADB as a function of different effective widths

PAGE 147

147 CH APTER 6 ORGAN DOSE AND INHER ENT UNCERTAINTY IN H ELICAL CT DOSIMETRY DUE TO QUASI PERIODIC DOSE DISTRI BUTIONS Background In helical computed tomography ( CT ), there is significant variation of organ and effective doses at surfaces. 21 23, 99 During helical scanning, there are overlaps and gaps in the primary exposure patt ern due to pitch, actual beam width, beam divergence, and the distance from isocenter. D oses will be larger or smaller within primary exposure pattern region overlap or gaps, respectively. Even a xially, in an anthropomorphic phantom, dose rate to a point i s not constant over a rotation of the x ray tube due to varying attenuation along a given chord from x ray tube to measurement point; this is compounded with varying beam diver gence. 45 The greatest dose rate occurs when the x ray tube is nearest as measured in attenuation length to the measurement point A t any moment in time during a helical scan, regions within a patient which are closer to the x ray tube position will have a higher dose rate than those regions farther away as measured using attenuation length It turns out that the cumulative dose distr ibutions at peripheral locations due to helical scanning are locally periodic in space with a fundamental period equal to the table translation per rotation which in turn, depend s up on pitch and detector collimation width. 24, 25 The magnitude of these quasi periodic distributions of dose are functions of : pitch, detector collimation width, beam divergence, and anatomical attenuation Similarly, the phase of these dose distributions is a function of x ray tube starting angle and starting edge of a scan. The phase is defined in terms of the location of the local dose maximum with respect to patient anatomy.

PAGE 148

148 Dixon et al. has derived mathematical expressions for accumulated dose distributions delivered in helical CT scans in cylindrical dosimetry phantoms, including one of the only expressions for the quasi periodic dose distributions on the peripheral phantom axes. 21 23, 99 These expressions are bas ed on a mathematically de rived single axial dose profile and the fundamental period of these dose distributions is equal to the table translation per gantry rotation. 21 23, 99 However, these expressions can be considered first order because they are valid for simple cylind ers. Furthermore, Dixon et al. gned to improve CTDI measurements and does not address p otential dose savings to radiosensitive tissues In other words, the use of an anatomically accurate anthropomorphic phantom such as the UF phantom series described by Winslow et al. can be used to me asure the theoretical dose savings by manipulating the phase of the quasi periodic dose distributions 1 0, 37 Zhang et al. has computationally measured dose delivered in surface and center positions in MDCT using Monte Carlo simulations of a 64 slice CT scanner, a body CTDI phantom model, and voxelized patient models 24, 25 This work was the first to quantify the magnitude of variability for the dose at the surface of the phantom models and discusses potenti al organ dose savings for a variety of pitch values beam widths, and tube starting angles. 24, 25 However, Zhang et al. used a CTDI cylinder which not only has a fixed diameter but limits the available posit ions for dosimeter placement. 24, 25 Additionally, Zhang et al. did not quantify the impact of beam divergence due to varying distance from isocenter 24, 25 S can settings can be manipulated in helical CT scans in order to minimize exposure to radiosensitive tissues which would lower the effective dose. 44, 90 This

PAGE 149

149 method also has the advantage of preserving image quality The extent and relative magnitudes of the dose modulations present in helical MDCT will be investigated in a realistic physical anthropomorphic phantom and compared with the results obtained in a computationa l cylinder by Zhang et al. 24, 25 R eal time dose measurements and a geometrically based model of peripheral dose will be combined in order to create a model which considers : nominal and actual beam widths, pitch, and beam divergence in helical CT scanning. The effects of patient positioning and scan geometry are also investigated Finally, t he potential dose savings to specific radiosensitive tissues is also measured and compared with the simulated estimates. M aterials and Methods CT Scanner and Measurement Devices A Siemens Somatom Sensation 16 multislice helical CT scanner (Siemens AG, equivalent fiber op tic coupled (FOC) real time dosimetry system as described in Chapter 3, was used for this study. This dosimetry system uses an FOC plastic scintillation dosimeter (PSD) with a 0.5 mm 3 scintillating volume and a time resolution of 10 ms. The system performance has been fully characterized by Hyer et al. 17 The UF male 50 percentile anthropomorphic phantom, as described in Chapter 2, was used for this study 10, 37 One of the advantages of this physical phantom is its computational twin which gives an accurate localization of anatomical features. Determination of Cumulative Dose Distribution The technique for determining dose distribution closely mi rrors the previously outlined method in Chapter 5. Essentially, the dose will first be measured for an axial scan and used to create a periodic dose rate profile. Specifically, the dose will first be

PAGE 150

150 measured at a peripheral location in a phantom at the ce nter of the beam collimation for an axial scan. Next, these measurements can be extrapolated to create a repeating periodic dose rate profile The total dose to a point in a helical scan acquisition can then be approximated by integrating over the region o f this curve defined by the amount of time the measurement point is in the primary beam. Th e integration time depends factors such as: pitch, actual and nominal beam widths, beam divergence, and distance from isocenter. The starting integration time can be shifted along the dose rate profile and is used to determine the minimum and maximum doses corresponding to a particular tissue. D ose modulations were expected to be most dramatic for peripheral tissues thus the radiosensitive thyroid and eye were chosen 90, 104, 105 Dose to the lens of the eye reflects scenarios where lens exposure cannot be avoided by tilting the gantry. The dose modulations observed by Tien et al. were largest for organs which were both peripheral and small therefore o ther radiosensitive candidates such as the breast, stomach, and testes were discarded 104, 105 It turns out that the thyroid and eye are the most likely candidates for dose reduction if start ing angle manipulation is made possible. Axial P oint D ose R ate Single axial scans were recorded with the FOC PSD dosimetr y system using the following technique: 120 kVp, 130 mAs, 12 x 1.5 mm beam collimation 9 mm reconstruction slice width, 0.75 second x ray tube rotation time kVp, tube current modulation off 1 scan, 2 images. The tip of the FOC dosimeter was placed in the a nthropomorphic phantom at physical positions which corresponded to the lens of the eye and the center of the thyroid. The measurements were repeated three times,

PAGE 151

151 averaged, and normalized for both tissues. Note that physical measurement of axial dose rate p rofile takes into account factors such as axial attenuation, x ray beam spectrum, and geometry which must be explicitly defined in computational simulations Helical P oint D ose R ate After collecting a xial point dose rates, the helical point dose rate is ob tained by integrating the profile for a given amount of time. The axial point dose rates were used as the basis for periodic functions used later in this analysis. Th e helical dose rate curve is used to represent the change in dose rate measured at a point due to the continual rotation of the x ray tube in helical scanning. At this point, the axial point dose rates were plotted as a function of an arbitrary index in order to accom mo date different gantry rotation times. Note that t able translation and beam w idth are accounted for in subsequent steps. Cumulative P oint D ose The time which a point would be in the primary beam was calculated using the table speed, distances from isocenter to the peripheral point and distance from isocenter to the x ray tube focal spot. Table speed was derived from pitch, detector collimation width, and gantry rotation time. 45 P it ch es of 1 and 1.5 were used The actual beam width value used was 28.3 mm as measured by Staton et al. for specific particular Siemens Somatom Sensation 16 CT scanner and a detector collimation width setting of 24 mm. 106, 107 The cumulative point dose was determined by integrating the helical dose rate curve over the time which a point would be in the primary beam. Underlying this method is the assumption that the peripheral dose could be calculated by the sum of a section of the helical dose rate curve which corresponds to the time which the point dosimeter

PAGE 152

152 would be within the primary beam of the x ray tube. The eff ect of scatter was determined by measuring the surface dose profile at isocenter for a CTDI head phantom in comparison with the dose profile measured free in air. Total T issue D ose The las t step to determine tissue dose also termed Total Tissue Dose i s to integrate the cumulative point dose curve over a region which represents the tissue. The length of integration corresponds to the length of each tissue. This is further refined by weighting each contribution to the sum using the distribution of mass wi thin each particular tissue. The metric of Total Tissue Dose allows quick comparisons between relative magnitudes of possible total tissue doses savings for the lens of the eye and thyroid by simply manipulating the phase of the dose distribution. The di stribution of mass within the lens of the eye along the z axis was approximated by using the chord lengths of a circle with a 10 mm diameter. 28, 108 The distributions of mass within the thyroid along the z axis were obtained more precisely from the segmented CT data set used i n creating the adult male physical phantom : the number of pixels for thyroid tissue located within each thyroid tissue containing CT slice was compared to the total number of pixels for the entire organ. In addition to the mass distributions, these data we re also used to determine the length of section of the cumulative point dose curve to integrate. There is a range of values in the Anterior Posterior (AP) axis for a reasonably defined isocenter for an anthropomorphic phantom. The nominal distance from is ocenter was 6 cm and 10 cm for the thyroid and lens of the eye, respectively. 28, 108 The range chosen for this study was 3 cm in either d irection along the AP axis. In other

PAGE 153

153 words, cumulative point doses and total tissue dose were also calculated for different distances to isocenter. R esults Axial Point Dose R ate R eal time axial measurements were recorded for points within the lens of the eye and the thyroid. The normalized plots of the axial dose rate to the eye and thyroid are shown in Figure 6 1. Since the gantry rotation time was 0.75 seconds and the axial scan is only one rotation, t he time between the peaks of this plot was 0.75 s econds. In t his plot the x ray tube position (nearest to the dosimeter for both organs posi tion before turning off. For both tissue locations, the troughs of the dose were less than 5% of the peak dose. Helical Point Dose Rate The axial scans were repeated for both tissue locations and used as the basis for periodic function s representing dos e rates at a point for a repeated axial scan or a hypothetical helical scan with a pitch of zero. The periodic functions for the lens of the eye and the thyroid are shown in Figure 6 2 plotted versus arbitrary units to facilitate different gantry rotation times ; this is in contrast to t he axial point dose rate curves in Figure 6 1, which within either organ while in the primary beam, during a helical MDCT scan for an arbitrary gantry rotation t ime. The phase of each curve is a function of the scan start position and the starting angle of the activated x ray tube. The helical point dose rate curve was derived including both primary and scatter radiation. Therefore, gafchromic film measurements we re performed in order to ensure

PAGE 154

154 that the scatter was negligible. It turns out that the dose measured at isocenter for a CTDI head phantom is less than 2% larger than the dose measured at isocenter free in air which suggests that the ratio of scattered to p rimary radiation is small enough at peripheral axes in a phantom to be negligible. Cumulative P oint D ose T he cumulative point dose curve represent s the dose at a point which has undergone a complete exposure by the primary beam Complete exposure is taken to mean during the course of the scan the point begins out of the primary field of exposure, enters the primary field of exposure, and exits the primary field of exposure. Cumulative point dose curves were calculated for two diff erent pitches as well as different phantom positions relative to isocenter. Specifically, the eye lens to isocenter distances selected were 7, 8, 10, and 13 cm; the thyroid to isocenter distances selected were 3, 6, 7.5, and 9 cm. Values were normalized ag ainst the nominal distance to isocenter distributions mentioned earlier. 28, 108 Figure 6 3 and Figure 6 4 show the possible cumulative point dose values for the lens of the eye and the thyroid for a pitch of 1 respectively. Figure 6 5 and Figure 6 6 show the possible cumulati ve point dose values for the lens of the eye and the thyroid for a pitch of 1.5. Again, within these Figures 6 3 to 6 6 tissue to isocenter distances were varied by changing the phantom positioning upon the table. The cumulative point doses were normali values if the primary beams were expected to overlap, while they were norm alized to values if the primary beams were expected to create gaps. For example, the nominal eye lens distance to isocenter of 10 cm did not result in p rimary beam exposure overlap, therefore the cumulative point doses were normalized to the maximum value On the other hand, the nominal thyroid distance to isocenter (6

PAGE 155

155 cm) result ed in primary beam exposure overlap, t herefore the cumulative point doses were normalized to the minimum value Geometric attenuation, which varies by the square of distance, was applied in order to account for different distances from tissue to the x ray tube. Beam divergence, which varies linearly with distance, was applied in order to account for different beam widths with different distances from tissue to the x ray tube. T he period of the local total dose distribution is equal to the product of detector col limation width which is simply table translation per full x ray tube rotation and pitch Therefore, t he phase as plotted for each cumulative point dose distribution is easily scalable. Each j th value of the cumulative point dose curve is the sum of a s ection of the helical point dose curve which begins at the j th value and continues another divergence corrected beam width The (j+1) th value is the sum of a section of the helical point dose curve which begins at the (j+1) th value and continues another d ivergence corrected beam width. In other words, the number of values which remains constant, it is simply the phase of the helical point dose curve which is shifted. T he range in magnitude of total point dose measurement values for all figures increased wi th pitch. An interesting case for the lens of the eye is the 8 cm distance from isocenter using pitch of 1, which showed no dose variability over the entire range. This is an example of a region where the effects of beam divergence and beam overlap seem to counteract each other : the cumulative point dose in this distribution decreased due to beam divergence while the increase in primary beam exposure overlap, and vice versa. This is seen in the thyroid at a distance of 7.5 cm distance from isocenter using

PAGE 156

156 p itch of 1. This demonstrates the variability based on initial positioning of the phantom positioning relative to isocenter. T he percent ripple, maximum, minimum, and average normalized values for cumulative point dose distributions for pitch es of 1 and 1. 5 are summarized in Table 6 1 and Table 6 2 for the lens of the eye and thyroid, respectively. Average normalized value refers to normalization to the maximum value for each pitch value. Averages are taken over a single period and percent maximum dose reduction refers to each distribution individually. Total T issue D ose Cumulative point dose represented the dose at a point which has undergone a complete exposure by the primary beam. The next step was to integrate the series of point measurements in order to represent total point dose. E ach point on the total tissue dose curve was designed to represent the total dose delivered to a tissue based up on position relative to starting phase The total tissue dose curves were calculated for different phantom positions and pitches of 1 and 1.5. Figure 6 7 and Figure 6 8 show the possible total tissue dose values for the lens of the eye and the thyroid for a pitch of 1, respectively. Fi gure 6 9 and Figure 6 10 show the possible total tissue dose values for the lens of the eye and the thyroid for a pitch of 1.5. These figures demonstrate the potential dose savings in tissue dose achievable with shifting only the phase. Note that t he se curves also contain information about dose savings in tissue dose achievable with differing phantom positioning T he range in magnitude of total tissue dose values increased with pitch as seen in Figures 6 7 to 6 10. Also, the distances closest to isocenter had the smallest range For the cases in which the phantom is positioned such that the distance from isocenter is

PAGE 157

157 greater than or equal to 3 cm from nominal distance, the reduction in average total tissue dose for lens of the eye was 16% and 14% for pitch es of 1 and 1.5. Similarly for distances which were greater than or equal to 3 cm from the nominal distance, the reduction in average total tissue dose for the thyroid was 13% and 12% for pitches of 1 and 1.5, respectively The percent ripple, maximum, minimum, and average normalized values for cumulative point dose distributions for pitches of 1 and 1.5 are summarized in Table 6 3 and Table 6 4 for the lens of the eye and thyroid, respectively. Similar to Tables 6 1 and 6 2, average normalized value refers to normalization to the maximum value for each pitch value. Averages are taken over a single period and percent maximum dose reduction refers to each distribution individually. Analysis This investigation was designed t o physical measure the potential tissue dose savings from manipulating the local dose distribution varying x ray tube starting position and phantom positioning relative to isocenter Dose savings from starting angle has already been measured in computa tion simulations by Zhang et al. 24, 25 Unfortunately, the physical measurements cannot serve as benchmarks because Zhang et al. used a different CT machine different beam collimations, and different phantoms. 24, 25 G eneral trends were observed between this study were seen and Zhang et al. for both organs exa mined For example, Zhang et al. reported percent ripple at the surface of the CTDI body phantom of 30% and 70% for pitch es of 1 and 1.5 respectively 24 The surface of the CTDI body phantom can be compared with the lens of the eye when

PAGE 158

158 positioned 13 cm fr om tissue to isocenter. The percent ripple found by this study is 25 % and 76% for pitch es of 1 and 1.5. Zhang et al. reported dose reduction s of 3, 7, and 11% for the lens of the eye at a pitch of 1 for three different phantoms used; 37, 49, and 52% for t he lens of the eye at a pitch of 1.5 for three different phantoms. 24 All these values compare favorably with the ranges found in t h i s study which found values which rang ed from 0 % to 20% for a pitch of 1 and from 59 % to 71% for a pitch of 1.5. Also, Zhang et al. reported dose reductions of 3, 4, and 3% for the thyroid at a pitch of 1 for three different phantoms used; 20, 20, and 4% for the thyroid at a pitch of 1.5 for three different phantoms. 24 All these values compare comparably with the ran ges found in this study, which found values which ranged from 0% to 4 % for a pitch of 1 and from 14 % to 19% for a pitch of 1.5. These values are especially impressive considering the difference in experimental setups mentioned previously: different CT scan ners, different beam collimation and different phantoms. Zhang et al. used CTDI phantoms for studies which facilitated precise isocenter placement. 24 By using physical phantoms, a range of table heights was visually chosen by a technologist as possible vertical heights for proper patient positioning Different tissue to isocenter depths were considered when calculating each of the total tissue dose savings. Furthermore, the horizontal plane could reasonably vary by more than 3 cm, which is large enough t o create large differences in tissue dose as evidenced by Tables 6 1 through 6 4. In order to compensate for beam penumbra, beam collimations are larger than their nominal width. 45 Therefore, p rimary beam exposure overlaps occur for scans even

PAGE 159

159 with pitch 1 near isocenter. As the distance from the tissue to isocenter increases, the size of the overlapp ing region decreases and actually becomes a gap between primary beam exposures. Moreover, t he distance from tissue to isocenter at which the overlap becomes a gap, the smallest range of total tissue dose savings is observed. This further enhances the effec t of phantom positioning As mentioned in C hapter 5 point measurements are much more subject to the periodic nature of d ose distributions in helical CT larger volume measurements simply average out any periodic variation. Specifically, caution must be t aken with small d osimeters such as TLDs, OSLs, semi conductors, and MOSFETs. While t his problem can be partially mitigated by using an average of multiple dosimetry points this presumes a good distribution in other words, small contribution of minimum o r maximum points. Accurate measurement of a region using the average of multiple point measurements is dependent upon measurement location as well as distribution and number of points. Another way to minimize the variability in point dose measurements is t o run multiple scans. The Siemens Somatom Sensation 16 was found to have an unpredictable but possibly biased activated x ray tube starting angle during helical scans which would lead to dose distributions which are not averaged out by using multiple scans Both averaging and starting angle dose biasing are discussed in C hapter 5 It was demonstrated that p eripheral points within a phantom during helical MDCT have periodic dose distributions based upon phantom posit ioning and x ray starting angle. Due to this periodic distribution it is feasible that aliasing could occur For example, the RANDO and ATOM phantoms measurement points could conceivably

PAGE 160

160 have the period of the local dose distribution with a phase of 2.5 cm, which would alias with the p hantom slice spacing of 2.5 cm. 30, 31 D ose reductions due to manipulation of local dose distribution ph ase would come at no cost to image quality. For the lens of the eye, up to 20% and 71% reduction in dose was calculated for scans using pitch es of 1 and 1.5, respectively. Similarly, f or the thyroid, up to 19% dose reduction was calculated for a pitch of 1 .5. These two organs are especially radiosensitive and have a high weight in effective dose calculation. T his method can thus efficiently lower the effective dose of the entire scan. Note that regardless of the phase of the local dose distribution, the dose to both the lens of the eye and thyroid can be reduced by shifting these tissues closer to isocen ter when setting up the patient This introduces the idea of adjustment of patient setup in relation to isocenter in order to protect asymmetrically d istributed radiosensitive tissue such as the lens of the eye and thyroid, which are positioned on the anterior side Discussion This study described a method ology in order to physically measure the dose savings to tissue located a peripheral locations usin g real time dosimetry. The dose savings was achieved through both phase shifting and patient positioning. Controlling the starting tube angle is one method of phase shifting. However, it is currently a source of uncertainty when using point dosimeters as d iscussed by Dixon et al. Tien et al. and also addressed in C hapter 5 21 23, 99, 104, 105 Dose to the lens of the eye and thyroid can be minimized by positioning patients so these tissues are closer to isocenter. Subs equent studies, including Monte Carlo studies similar to those of Zhang et al. ., should be designed to further evaluate the

PAGE 161

161 potential dose savings or image quality costs resulting from positioning patients off isocenter. Potential total tissue dose reducti on was quantified Phase shifting has an advantage over other dose reduction techniques, notably shielding and tube current modulation, because there is reduction in dose reduction with no accompanying loss of image quality. Tissue dose savings were observ ed for both the lens of the eye and thyroid two especially radiosensitive organs studied with this methodology. 44, 90 As mentioned earlier, phase shifting is not currently viable because the ability to dictate x ray tube starting is not clinically available. In order to incorporate this method into clinical practice, manufacturers would need to devise a method for sensing and controlling the x ray tube starting angle If this was accomplished, dose distribution could be determined using sc an paramet ers and a scout image. Presumably, the lowest points of the tissue dose would then be overlaid over the radiosensitive organs. This work was supported by the U.S. Department of Energy under project award number DE GF07 05ID14700

PAGE 162

162 Table 6 1 Lens of the eye, cumulative point dose distribution: percent ripple, maximum, minimum, and average normalized values current 2009), Table 5 1, p.121) Table 6 2 Thyroid, cumulative point dose distributions: percent ripple, maximum, minimum, and average normalized values and prediction in tube current 1, p.121)

PAGE 163

163 Table 6 3 Total tissue dose, lens of the eye : percent ripple, maximum, minimum, and average normalized values assessment and prediction in tube current 2, p.122) Table 6 4 Total tissue dose, t hyroid : percent ripple, maximum, minimum, and average normalized values (Data reproduced with permission current 2, p.122)

PAGE 164

164 Figure 6 1 Normalized axial dose rate ( Figure courtesy of J Winslow C onstruction and application of anthro pomorphic phantoms for use in CT dose studies PhD Dissertation (20 09), Figure 5 2 p. 114) Figure 6 2 Normalized helical dose rate ( Figure courtesy of J Winslow onstruction and application of anthro pomorphic phantoms for use in CT dose studies P hD Dissertation (20 09), Figure 5 3 p. 115)

PAGE 165

165 Figure 6 3 Lens of the eye, pitch of 1, cumulative point dose for different positions ( Adapted with permission from J Winslow onstruction and application of anthro pomorphic phantoms for use in CT dose studies PhD Dissertation (20 09), Figure 5 4 p. 115)

PAGE 166

166 Figure 6 4 Thyroid, pitch of 1, cumulative point dose for different positions (Adapted with permission from J Winslow onstruction and application of anthro pomorphic phantoms for use in CT dose studies PhD Dissertation (20 09), Figure 5 7 p. 117)

PAGE 167

167 Figure 6 5 Pitch of 1.5, l ens of the eye, cumulative point dose for different positions ( Figure courtesy of J Winslow CT dosimetry due to qua si periodic dose distributions Med Phys 3 8 3177 3185 (2011), Figure 5, p. 3181)

PAGE 168

168 Figure 6 6 Pitch of 1.5, t hyroid, cumulative point dose for different positions ( Figure courtesy of J Winslow dosimetry due to quasi periodic dose distributions Med Phys 3 8 3177 3185 (2011), Figure 6 p. 3181 )

PAGE 169

169 Figure 6 7 Normalized total tissue dose, l ens of the eye, pitch of 1, for different positions ( Figure courtesy of J Winslow ent uncertainty in helical CT dosimetry due to quasi periodic dose distributions Med Phys 38 3177 3185 (20 11 ), Figure 7 p. 3182 )

PAGE 170

170 Figure 6 8 N ormalized total tissue dose, t hyroid, pitch of 1, for different positions ( Figure courtesy of J Winslow CT dosimetry due to quasi periodic dose distributions Med Phys 38 3177 3185 (20 11 ), Figure 8, p. 3183)

PAGE 171

171 Figure 6 9 Pitch of 1.5, Lens of the eye, normalized total tissue dose for different positions ( Figure courtesy of J Winslow Organ dose and inherent uncertainty in helical CT dosimetry due to quasi periodic dose distributions Med Phys 38 3177 3185 (20 11 ), Figure 9, p. 3183)

PAGE 172

172 Figure 6 10 Pitch of 1.5, Thyroid, normalized total tissue dose for different positions ( Figure courtesy of J Winslow CT dosimetry due to quasi periodic dose distributions Med Phys 38 3177 3185 (20 11 ), Figure 10, p. 3183)

PAGE 173

173 CH APTER 7 A PRELIMINARY STUDY OF TUBE CURRENT MODULATION USING PHYSICAL MEASUREMENTS AND DIC OM HEADER EXTRACTION Background The sixth project of this work was to determine the dose savings from tube current modulation techniques (TCM). The purpose of this investigation was to generate a libra ry of data which will be used to develop a predictive computational algorithm for TCM. This will eventually be used to predict the dose delivered during any CT procedure. T ube current modulation (TCM) adjust s the x ray tube current during a scan in order to adapt to varying attenuation Previously, tube current and voltage were fixed for the entire scan. These parameters were chosen based upon patient size patient weight and image quality requirements. Without TCM, varying body circumference and tissue a ttenuation at different longitudinal locations lead to variable attenuation though the scan which result ed in different numbers of signal carriers reaching the detectors at different table positions. Ultimately, this lead to unacceptably low image quality in h igh attenuation regions and excessively high image quality in low attenuation regions. Images with too much noise obscure low contrast lesions or tumors that would normally be visible in less noisy images In turn, this could lead to misdiagnoses or t he need to rescan the patient, exposing them to unnecessary radiation. More often, f ixed techniques suffer from underexposure which results from photon starvation and resultant artifacts which are manifested during reconstruction. 45, 109 Previously, in order to avoid photon starvation artifacts, the tube current for the entire scan area was increased as opposed to simply increasing tube current in a specific region as done with

PAGE 174

174 TCM. With the advent of computing power, it is now possible to use TCM in order to tailor the x CT dose has been traditionally associated with computed tomography dose index (CTDI) o r one of its derivatives which were developed to maintain relevance as CT progressed from axial to helical scans and from single detector to multi detector scans. 45, 98, 99, 102, 110, 111 Volume CTDI ( CTDI vol ) is the most recent iteration, which is measured using a 10 cm pencil chamber in a CTDI phantom and is intended to represent the average absorbed dose along the z axis from a series of contiguous iterations. 6, 7, 9, 102 CTDI vol is simply the quotient of CTDI w and pitch. 6, 7, 9, 102 In turn, CTDI w is intended to approximate average dose in the x y plane and is calculated using a weighted average of measurements in different CTDI locations. 6, 7, 9, 45 Overall, CTDI numbers are commonly used in the clinic a s a single number which can be used for quality assurance and provides simple comparisons in tube output across scan parameters. 6, 7, 9, 45, 98, 99, 102, 110, 111 CTDI vol only esti mates the average radiation dose within an irradiated volume with similar size, shape, and attenuation as a CTDI phantom. 7 In many cases, CTDI numbers ha ve been criticized as being too simple for fully quantifying dose : a single number does not easily or accurately convert into a clinically measurement relevant metric as well as effective dose or organ dose. 6, 7, 9, 45, 98, 99, 102, 110, 111 This complaint has only increased as TCM has been incorporated. 22, 23, 99 With TCM, tube current is neither constant across the scan nor is a patient very similar to a homogenous Lucite cylinder. In other words, CTDI vol is not an adequate parameter for absorbed dose along the z axis in a ny scan which employ s TCM of an anthropomorphic patient. 102

PAGE 175

175 The physical dosimetry system described previously in Chapter 3 can be used to accurately measure a point dose delivered during a CT scan. Alternatively, other approaches rely upon the computational power available through computational Monte Carlo methods. However, to date, these have been limited to fixed tube current methods. 46, 109, 110, 112 T he ImpaCT group of the UK has commercialized a CT dose calculator for clinical CT procedures based upon Monte Carlo calculations. This software uses inputs of: CTDI values, imaging pr otocol techniques, beam characteristics, and geometric measurements characteristics The output result is organ doses resulting from CT imaging procedures. However, these Monte Carlo data are computed from axial, fixed tube current scans performed on styli zed geometric phantoms compensate for this, an average or maximum tube current used in a scan is used. Much of the ImpaCT software is based on conversion factors based upon CTDI numbers, wh ich has been proven to underestimate actual delivered doses 21 23, 46, 99 It is absolutely feasible and accurate to obtain o rgan and effective doses for CT procedures using Monte Carlo computational simulations. However, i n corporating TCM into the source modeling of Monte Carlo simulations has been challenging due to the lack of availability of data matching tube current to scan position This portion of the investigation aims to measure the tube current throughout different protocols using a modified cylindrical phantom. These data will provide the data required to produce a predictive algorithm of tube current throughout the scan. There are currently two major TCM strategies employed by manufacturers : angular and z axis modulation The Siemens SOMATOM

PAGE 176

176 Sensation 16 which was studied uses CARE Dose 4D. CARE Dose 4D actually utilizes modulation is based up on at tenuation value measured from the previous 180 degree rotation. This development was made possible due to advances in electronics namely processing speed and smaller size Z axis modulation is based on each slice as determined from a scout radiograph. Be fore the helical scan is done, a scout scan also referred to as a topogram is performed and estimated attenuation through each Z axis modulation is designed to keep image quality uniform in each reconstructed slice. This processing is not done in real time like the within a single rotation Tube current data was recorded for both z axis and angular TCM schemes. Materials and Methods A custom MATLAB program was created in order to extract tube current in each slice for reconstructed images produced by the scanner. This was done by parsing the Digital Imaging and Communications in Medicine (DICOM) header produced in the reconstructed i mage for the tube current tag used by DICOM for distributing and viewing any kind of medical image 45 While the first stages of this investigation utilize PSDs, many of the data are extracted from parsed on board electronic current measurement reports. Therefore, the first parts of this project describe the utilization of PSDs while the latter parts descri be the modeling work which will not require PSDs. The rationale for using DICOM headers was that PSDs do not measure tube current, but instead measure counts. As described in Chapter 3, these counts are subsequently converted to dose through a series of c alibration factors. Therefore, while

PAGE 177

177 PSDs offer excellent time resolution (10 ms), they do not give information on the actual tube current. PSDs can also be used to collect data about the tube current output. CTDI Phantom The first step was to calibrate measurements to a CTDI phantom with a PSD. CTDI phantoms have four peripheral holes and one central hole in order to accommodate a pencil ion chamber. When not in use, a Lucite plug is inserted into the hole and simply taped into place. The plug eliminates air gaps. Therefore, in order to eliminate any air gaps which would occur from PSD measurements, a Lucite plug was grooved using a dremel in order to accommodate the FOC architecture. An axial slice of the CTDI phantom is shown with the PSD and Lucite plu g in Figure 7 1 For these scans, the CTDI phantom was placed in the head holder of the CT scanner. The head holder is made of relatively radiolucent material, in this case carbon oes not position the middle of the head near isocenter. For the CTDI phantom, the dose delivered during both helical and axial scans was determined at a given lo ngitudinal z position. For the helical scans, measurements were taken with TCM activated using a reference mAs of 150. Additionally, data were taken with the TCM deactivated. TCM is not used for axial scans. The standard deviation of the dose was calculate d for 8 10 repeated measurements. CTDI Phantom with Elliptical Add Ons Elliptical add ons for the CTDI phantom were constructed of the soft tissue equivalent substitute (STES) described in Chapter 2 of this work. The attenuation properties of this material are similar to Lucite. 11 Two add ons were added to enclose an

PAGE 178

178 existing CTDI head phantom. Thus, the resultant oval shape had the same height of 15 cm as the CTDI head phantom along with a minor axi s of 16 cm. 45 Three different major axes were used in the add on piece, thus creating a variable stair step phantom. Due to the elliptical attenuation, the phantom was expected to challenge the TCM in angular modulation. The elliptical phantom concept, design and fabrication was described in detail by Fisher. 113 Fisher crea ted a family of elliptical add ons with increasing major axis size which maintain the same minor axis in order to be used in conjunction with a CTDI head phantom. Id The major axis lengths used in each of the five phantoms were 26 cm, 28.5 cm, 31.25 cm, 32. 6 cm, and 37.25 cm (+/ .25 cm variation from top to bottom); these phantoms are shown in Figure 7 2 113 In the creation of these elliptical add ons, Fisher et al. i nitially used the methodology of stacking foam blocks des cribed in Chapter 2 However, the foam cutouts were proven to be too flimsy to withstand the weight of the STES. Specifically, the Fisher et al. discovered that the wax paper designed to provide a seal within the foam bulged and, in some cases, ripped. 113 Therefore, plywood was used instead to create a more stable mold which would withstand the weight of the STES. 113 A similar process was followed, using the engraver to cut elliptical templat es 113 Instead of wax paper, however, the interior was lined with a rubber sheet 113 Using wooden blocks as spacers between the three main plywood supports, the entire apparatus was const ructed to have a height of 15 cm, which corresponds to the height of a CTDI head phantom. This construction process is shown in Figure 7 3.

PAGE 179

179 The CTDI head phantom was placed within the center of the mold and filled with STES. This was too bulky in order to fit into the head holder, and therefore was elevated to isocenter using radiolucent foam blocks. In an improvement to the single sized elliptical add on, an STES step phantom was constructed. This step phantom is a series of elliptical add ons with varyin g major axis lengths. Essentially, this is representation of varying patient diameter. A picture of the CTDI head phantom inside the step phantom is shown in Figure 7 4 Figure 7 5 shows a topogram. STES Phantom In addition to the Lucite CTDI head phanto m, a cylindrical phantom with identical dimensions was constructed made of the STES material described in Chapter 2. This was done in order to create a completely uniform STES ellipse if necessary. This cylinder was constructed in a similar manner to the e lliptical add ons with stacked supports. The topogram of the final STES cylinder with in the elliptical add on is shown in Figure 7 6 The STES cylinder phantom also has four peripheral holes and a center hole for pencil chamber placement. STES plugs were a lso constructed in order to eliminate air gaps when the holes were not being used for pencil chamber measurement. A thin radial hole was drilled at mid height, 7.5 cm, parallel with the circular surface in order to aid in localization. Also, four thin hole s were drilled normal to the circular surface. The four holes can be seen in the axial view of the STES cylinder phantom. Note the U shaped form in this figure which surrounds the cylinder is the carbon fiber head holder.

PAGE 180

180 DICOM Information Extraction Mana gement of DICOM f iles Each slice of the helical reconstruction was output into an image ( *.ima ) file. Each *.ima file represents a reconstructed slice with height equal to reconstruction slice width. Thus for each typical abdominal scan, with slice thickne ss resolution of 5 mm, about 50 files were generated. After even a few scans, it became difficult to manage all of these files. Therefore, a MATLAB program was written (file_sorter.m), which was dedicated to sorting the considerable number of output data f iles for each scan. The source code for this file can be found in Appendix B. Siemens uses a specific file naming convention for naming its *.ima files. The *.ima files followed by the sc an protocol type followed by the scan series number followed by the number within the scan series followed by the date followed by the time followed by two large numbers which were related to the position of the slice within the volume. This was followed b y a period and a sequence of numbers which related to the archival information. performed on November 1, 2010 at 2:01 pm, and it was the 3 rd scan performed, the 7 th image in the sequen ce would begin with There were a few important exceptions. For example, files with the series number 501 were dedicated to patient prot ocol information while the series number 601 was dedicated to information about raw data storage method. It was discovered that the

PAGE 181

181 topogram performed at the beginning of each study was always designated scan number 1. Also, if a fast reconstruction algori thm was not available for the particular scan, then these scans would begin numbering again with a 100 prefix. For example, the first non The slow and fast designations are summarized in Table 7 1 Folders were cr eated to store the data for future use in format, X IMA Y, where X represents the series number and Y represents the number of reconstructed slices which exists in this series. Tag identifier a ssociations The second MATLAB program file in this code was dedicated to reading the DICOM encoded file and pulling information out by the tag identifier of each of the *.ima files and, for this particular file (readoutv1_7.m), outputs the slice location, x ray tube c urrent, and exposure for each scan. The source code for this file can be found in Appendix C. Unfortunately, the DICOM headers for the *.ima files did not include some information such as reference mAs and effective mAs for a given scan. The effective mAs was found on the console and recorded in the patient protocol (prefix 501) rather than the *.ima file. Other important parameters omitted in the DICOM information were estimated scan time and CTDIvol. Therefore, these parameters were separately recorded a nd manually added for completeness. The Siemens SOMATOM series of CT scanners chose one out of two reconstruction algorithm s based upon the reconstruction width and beam collimation selected during each scan. 114 reconstruction modes. id For correct data read out, it was crucial for both of these codes

PAGE 182

182 that the appropriate reconstruction mode is selected. The mode of reconstruction is determined primarily by beam collimation thickness and reconstruction slice height. The reconstruction modes as a function of beam collimation and reconstruction slice heig ht for an abdominal scan are summ arized in Table 7 1 Z axis TCM Method The Siemens SOMATOM Sen sation 16 which was studied used CARE Dose 4D. CARE Dose 4D actually utilize d axis modulation. However, there are two protocols which use d only one of these techniques. As it turns out, the Extremities protocol relied exclusively upon modulatio n while the Head protocol relied exclusively upon only z axis modulation. 114 These TCM methods were perform ed independently of each other. Therefore, using the Head and Extremities protocols, different algorithms could conceivably be tabulated and modele d separately in different algorithms After validation, they could be combined into one complete algorithm. A s based up on attenuation value measured from the previous 180 degree rotation. This recent development was made p ossible due to the advances made in electronics. Z axis modulation wa s based on each slice as determined from a scout radiograph. Before the helical scan wa s done, a scout scan also referred to as a topogram wa s performed ; estimated attenuation through each wa s determined. Z axis modulation wa s designed to keep image quality uniform in each reconstructed slice thus the tube current was observed to remain constant within a single rotation 114 In contrast, angular TCM was changed within each rotation as information from the previous 180 degrees was processed on line.

PAGE 183

183 Th is stage of the investigation fo cuse d on obtaining data to develop the first stage of modeling TCM, namely z axis modulation. In other words, the undulations seen using did not need to be discriminated. The first stage approach was to isolate just the z axis modulation function and thus required only the head protocol This investigation anticipated the next stages and data was obtained for other common protocols such as neck, abdomen, chest/thorax, and pelvic scans. The only infor mation required for the algorithm development was tube current information averaged over a slice. This further justified the choice of using DICOM instead of using the PSD system because the immediate output wa tube current information For t he first stage of algorithm development, using the DICOM headers provided a superior methodology for quantifying tube current using real time PSD measurements for three reasons. First, in this way, even if angular TCM wa s being used, only the average curre nt over the slice will be displayed. Secondly, all information was found without using any extra dosimetry systems Lastly, the tube current wa s displayed as electrical current rather than normalized dose. Note that for further algorithm development, the P SD system would provide an independent source of much higher sampling rate of tube current per slice Experimental Physical Measurements Protocol s election longitudinal z axis, and ignore the angular attenuation profile 114 Conversely, the applications us with no z axis TCM 114 Other than

PAGE 184

184 using one of those protocols, a nother method of isolating only angular modula tion is to perform a scout scan with an empty gantry. Next, the desired object to be scanned would be positioned in the gantry and scanned. Because the software has no axis composition from its blank scout scan the sy stem will not incorporate z characterize the TCM applied. The Image Quality Reference mAs corresponds to the effective tube current value that the technologist would apply for a reference patient without TCM. While initial dose savings may not be dramatic, the tube current will be her image quality. Siemens defines a reference patient as a 70 to 80 kg for adults and 20 kg for a child. 114 The change in the tube output due to z axi s TCM is much larger in comparison with the angular TCM. In a sense the z axis TCM establishes a baseline while the angular TCM oscillates around this value as the x ray tube rotates around the gantry. This investigation was designed to provide data which would be used to create a predictive computational algorithm for TCM. Therefore, in order to accommodate this need, this investigation focuses upon the first stage of this modeling: the z axis modulation. As the algorithm is refined and developed, the angu lar modulation will be incorporated along with more physical measurements. It is also important to correctly model the z axis modulation first because the angular modulation is heavily dependent upon highly variable parameters, notably pitch

PAGE 185

185 and beam colli mation. On the other hand, the z axis modulation depends simply upon the attenuation measured in the scout topogram. This investigation was designed with this in mind because the output current is av eraged over one complete reconstructed slice. Symmetrica l scan v alidation An easy way to test the robustness of the computational algorithm would be to scan the cylindrical phantom in both cranial caudal and caudal cranial orientations In this way, for the stair step phantom, the attenuation is either increasi ng or decreasing depending upon the orientation sele c ted. This was easy to implement and challenge the algorithm to perform with both increasing and decreasing attenuation values. Reconstruction m eth od e ffects s based primarily upon beam collimation and reconstruction slice height. These paramet ers are summarized in Table 7 1 for an abdominal protocol. The immediate effect of using a thinner beam collimation wa s more overall x ray tube rotations in order to cove r the same distance. With smaller reconstruction slice thicknesses, more slices were available for a given scan volume. This offered more sampling points of current which in turn provided more data to study the behavior, with better statistics. More data were added to the reco nstruction by using smaller beam collimation. Results CTDI Phantom The cylindrical CTDI head phantom was scanned to establish a n order of magnitude measurement estimate for electrical current. Even with the TCM turned on, this phantom did not challenge t he z axis TCM algorithm because its attenuation did not

PAGE 186

186 vary along the z axis. Within the CTDI phantom, the coefficient of variation was 1.07 and 4.39 with the Lucite PSD plug and with an air cavity, respectively. The se images were previously shown in Figu re 7 1 Furthermore, the symmetry of the simple cylinder did not challenge the angular TCM algorithm. The results using different reference mAs values is shown in Table 7 2 The nomenclature of R/E represents the values of the reference mAs and effective mA s, respectively. This will be used for the rest of the t ables. As can be seen in Table 7 2, the current did not change along the z axis. CTDI and STES Phantom s with Elliptical Add ons Using the PSD system, the dose delivered was measured using the fibers embedded within the Lucite PSD plug in each of the CTDI positions (4 peripheral and 1 central locations). This was performed in both helical and axial scans. The dose to Lucite was measured at l ongitudinal z positions of 4, 7, and 13 cm which correspond ed to increasing major axes in the x y plane of the phantom For the helical scans, measurements were taken with TCM activated using a reference mAs of 150. Additionally, data were taken with the TCM deactivated. TCM is not used for axial scans. The standard deviation of the dose to Lucite was calculated for 8 10 measurements. Axial, helical without TCM, and helical with TCM measurements are shown in Table s 7 3 7 4 and 7 5 respectively Again, t he nomenclature of R/E represents the values of the reference mAs and effective mAs, respectively. This procedure was repeated using the STES phantom with elliptical add ons. Reconstruction Methods Siemens has two main methods of reconstruction for its im ages which are determined by the beam collimation and reconstruction slice thickness. The estimated

PAGE 187

187 CTDI vol is shown as a function of reconstruction mode, beam collimation, and reconstruction slice thickness in Table 7 6 Using a more tightly collimated be am in the fast reconstruction scheme provide s the best statistics in terms of number of samples per study. Using a 3 mm reconstruction slice thickness with a 12 mm beam collimation would result in 11% more dose delivered. Therefore, thin slices should be u sed only for initial development purposes because it is unrealistic for clinical cases. Z Ax is Modulation Characterization U sing Routine Head Protocol The Siemens routine head protocol actually consisted of two scans performed in succession: first a cerebr um followed by a base scan. This protocol provided the simplest basis for construction of the current TCM for any scan in the head category is performed with only z axis modulation. For certain exams, such as the Adult Hea d Protocol, Siemens uses limited TCM where the tube angular attenuation profile. 114 In the production of the predictive algorithm, exclusively z axis TCM ( no angular TCM) would be the easiest aspect to model as all information for the TCM is based upon the scout scan. The head routine actually wa s divided into two scans: base and cerebrum. The default for scout scan wa s lateral. This still challenge d the TCM because the regions possess ed different HUs. The scout scan wa s performed in the typical manner with a cranial translation. However, during t he actual scan, the table translates in a caudal direction. The results for a thin (3 mm) head base protocol is shown in Table 7 7 The results for a routine cerebrum and base protocol s are shown in Table 7 8 and 7 9 respectively. T h e three different atte nuation regions produced by the elliptical add ons are distinct. The default scan protocol parameters are shown in Table 7 10. The

PAGE 188

188 default effective mAs values when using no TCM and corresponding tube curr ent are summarized in Table 7 1 1 Angular TCM Cha racterization Remember that this method of DICOM extraction outputs the average current for a given reconstructed slice. Therefore, it was expected that with about 2 full rotations of the x ray tube (typical beam collimation of 24 mm with a typical reconst ruction slice width of 5 cm), the variations between slices would not be evident in the results. This was true : each measurement of current was the same as the previous slice as the standard deviation in many of these measurements was 1.5% or smaller. Re sults Using Both Z Axis and Angular TCM Overall, s cans were performed for the five major protocols: head, neck, chest, a b domen, and pelvic. The head protocol was previously described and data were provided for both thin (12 mm beam collimation, 3 mm recon struction slice thickness) and default (24 mm beam collimation, 5 mm reconstruction slice thickness). However, as mentioned previously, the thin protocol wa s not clinically realistic and was meant to provide better sampling for the initial development of t he algorithm. N eck, ch est, abdomen, and pelvic scans use both z axis and angular TCM. The data for the other four major protocols neck, abdomen, chest/thorax, pelvic are summarized in Tables 7 12 through 7 15 respectively The default effective mAs values with no TCM and corresponding tube current were previously summarized in Table 7 11 Reverse Abdominal A routine abdominal s c an was performed caudal cranial rather than the default cranial caudal The data are summarized in Table 7 16 It was mo re instructive to plot

PAGE 189

189 t he average slice currents as a function of position for both thick side first and thin side first protocol which typically relies upon both z axis and ang there was nothing scanned by the topogram. In this manner, no information about z axis attenuation can be used to create z axis TCM and any TCM w hich was performed would be due solely to real The results of only angular TCM are shown in Figures 7 7 7 8 and 7 9 for reference mAs of 160, 200, and 260 respectively, for the abdominal scan with two different orientation s. Figures 7 1 0 and 7 1 1 show the results for different table translation direction s with both angular and z axis TCM In each F igures 7 7 to 7 11 there is an over response at the end, which was most dramatic when examining the different translation direc tions c ases with a maximum overshoot of 82 mAs when scanning thick side first (cranial caudal in clinical setting ) and a maximum overshoot of only 34 mAs when scanning thin side first (caudal cranial in clinical setting) Analysis Using only the Lucite cylinder or STES cylinder phantom does not challenge either of the TCM methods used by the system z axis and angular. Therefore, in order to accomplish this, elliptical add ons were added to these simple cylinders, creating three di stinct regions of attenuation. Additionally, this modified the shape of the object from a simple cylinder to a more complex ellipsoid. In addition to being convenient to implement, both the z axis and angular TCM methods were challenged by this design. Th e different attenuation regions of the cylindrical phantom with elliptical inserts can be observed in Tables 7 7 through 7 9 and Tables 7 12 through 7 15 The plan in

PAGE 190

190 order to develop an algorithm involves the first generation of the algorithm model ing onl y the z axis TCM. This is used by Siemens for head protocols. These data are shown in Tables 7 7 and 7 9 which showed the base and cerebrum protocols. Next, only the angular TCM would be modeled. This can be accomplished by using an extremities protocol be cause Siemens uses only angular TCM for its modulation And finally, the z axis and angular TCM codes will be incorporated into one. Th is is used by Siemens for neck, chest/thorax, abdomen, and pelvic scans. Th ese data are shown in Tables 7 12 through 7 15 In Table 7 7 a smaller slice than default was used in order to increase the sampling of the tube current data However, using a 3 mm reconstruction slice thickness with a 12 mm beam collimation would result in 11% more dose delivered. Therefore, thin slices should be used only for initial development purposes to provide higher resolution in longitudinal sampling but is unrealistic for clinical cases. The typical and inverted abdomen d ata are shown plotted in Figure s 7 10 and 7 11 I n Figures 7 10 and 7 11, the same phantom was used along with the same effective mAs The only difference was the type of protocol, which suggested that the difference was due to the angular TCM present in the abdominal case but absent in the head case. These figures show tha t it is better to insert a patient in a certain direction if inverted direction (thin side first). In a clinical setting, this would be equivalent to a cranial caudal scan rather tha n a caudal cranial scan. Angular TCM Method was studied by a novel method. This method involved using a protocol which typically

PAGE 191

191 relies upon both z axis and angular TCM. The difference was taking a blank scout scan. In other words, there was nothing scanned by the topogram. In this manner, no information about z axis attenuation can be used to create z axis TCM and any TCM which was performed would be due solely to on line angular processing. This experiment was designed to eliminate the z axis TCM in order to isolate the was done for completeness. It was anticipated th at Z axis TCM w ill produce much larger modulation results than angular modulation, especially given the relatively simple geometry of this p hantom, notably in its purposeful lack of high or low attenuation regions. It was anticipated that the angular TCM will be the most challenging aspect of this predictive algorithm as a ray trace program must be developed in order to determine the attenuation information being processed by the scanner during online TCM. After incorporation of angular TCM, there was a consistent over r esponse near the the end of the scan, which was most dramatic when examining the thin side first cases. This was an example of the unexpected results from angular TCM. The thick side first cases still have a jump, but the overall magnitude was not as big. Overall, the integrated area under the reversed cases is smaller than the routine cases. The angular TCM does this because it operates on feedback of the previous 180 degrees. At the end of the scan, for an attenuating slice, the angular TCM is compensatin g and putting out a current. When the scan reaches the end of the patient, all attenuation is removed but the higher current is still activated.

PAGE 192

192 The more dramatic results are seen when ending in the thick slices because the overall current must be higher. In other words, the percentage overshoot is almost equivalent, around 30%, but in terms of overall magnitude, ending with thick slices has a much higher overshoot. This could be investigated as another future source of dose reduction and could be implemen ted immediately in a clinical setting. Machine R eproducibility It was noticed while incorporating angular TCM that the reproducibility degraded from 4% to 1.5% observed with the z axis TCM This was due to variable starting angle as mentioned in Chapter 5 of this work. The extent of deviation was visualized by taking a series of 10 consecutive scans and is shown in Figure 7 10 This was repeated for the other table translation direction, with a series of 10 consecutive scans and is shown in Fig ure 7 11 TCM will simply add in quadrature with the uncertainties of the isolated z axis TCM. Discussion This portion o f the investigation recorded the tube current throughout different protocols. These data will provide the basis of a predictive algorithm of tube current throughout the scan. It is absolutely feasible and accurate to obtain organ and effective doses for CT procedures using Monte Carlo computational simulations. However, incorporating TCM into the source modeling of Monte Carlo simulations has been challenging due to the lack of availability of data matching tube current to scan position. The only informatio n required for the algorithm development was tube current information averaged over a slice. In many Monte Carlo simulations, the source is simply defined as normalized value. Therefore, mAs can be defined as a certain number

PAGE 193

193 of source particles. The PSD d ose measurements were replaced with a custom MATLAB program which extract ed tube current from DICOM headers. This should prove easier for algorithm development as the information involve s number of source particles whereas PSD do se measurements would requi re complete simulations to determine detector interactions. For this first stage of algorithm development, using the DICOM headers provided a superior methodology for quantifying tube current using real time PSD measurements for three reasons. First, in th is way, even if angular TCM was being used, only the average current over the slice will be displayed. This eliminates many of the high axis TCM. Secondly, all in formation was found without using any extra dosimetry systems. Lastly, the actual tube current was displayed as current, not normalized dose, which offers the advantage in easier source modeling as mentioned above. Note that for further algorit hm developme nt, the PSD system c ould provide an independent source of much higher sampling of tube current per slice. Specifically, with 100 Hz sampling and a table speed of 48.5 mm/s as measured in Chapter 4, counts could theoretically be measured with sub millimeter bins instead of 5 mm bins This would be enough to isolate the high measurements from the PSD system have been shown to resolve the gantry rotation, as seen in Chapter 4. Elliptica l add ons were used to transform cylindrical phantom into phantoms which challenged both z axis and angular TCM. The phantom was made with relatively simple geometry, notably in its purposeful lack of high or low attenuation regions. Data were

PAGE 194

194 gathered for head, neck, chest/thorax, abdomen, and pelvic scans. Thin slices were clinically unrealistic because dose was 11% higher with 3 mm slices rather than 5 mm slices. However, a smaller slice than default was used for some scans in order to increase the sampling of the data on the current Lastly, t here was a significant overshoot observed in the abdominal scans which was attributed the angular TCM present. This overshoot was based upon percentage rather than absolute magnitude, t herefore it is recommended that the lowest attenuating region be scanned at the end of a scan. This can be implemented in the clinic immediately by diligent patient positioning. This experiment was the first stage of data collection for an algorithm which predicts TCM behavior. It i s anticipated that the angular TCM will be the most challenging aspect of this predictive algorithm as a ray trace program must be developed in order to determine the attenuation information being processed by the scanner during online TCM.

PAGE 195

195 Table 7 1 Reconstruction mode as a function of beam collimation and reconstruction slice height for an abdominal scan Table 7 2 C ylinder with no elliptical add ons, in routine abdominal scan in notation

PAGE 196

196 Table 7 3 Point dose to Lucite using CTDI head phantom with step phantom as a function of CTDI hole position and longitudinal position for 16x1.5 cm axial scan Table 7 4 Point dose to Lucite using CTDI head phantom with step phantom and no TCM as a function of CTDI hole and longitudinal position for 15 cm helical scan

PAGE 197

197 Table 7 5 Point dose to Lucite using CTDI head phantom with step phantom, with TCM, using 150 mAs reference, as a function of CTDI hole and longitudinal position for 1 5 cm helical scan Table 7 6 Console estimated helical scan dose to Lucite as a function of reconstruction mode, beam collimation, and reconstruction slice thickness for 15 cm scan

PAGE 198

198 Table 7 7 H ead base protocol with thin slices using cylinder with elliptical add ons for different TCM mAs R to mAs E (R/E) values

PAGE 199

199 Table 7 7 C ontinued

PAGE 200

200 Table 7 8 Routine cerebrum scan, cylinder with elliptical add ons, tube current as a function for different TCM mAs R to mAs E (R/E) values

PAGE 201

201 Table 7 9 Routine head base scan, cylinder with elliptical add ons, tube current as a function for different TCM mAs Rto mAs E (R/E) values

PAGE 202

202 Table 7 9 C ontinued

PAGE 203

203 Table 7 10 Default beam parameters Table 7 11 Default effective mAs values when using no TCM

PAGE 204

204 Table 7 12 Routine neck scan, cylinder with elliptical add ons, tube current for different TCM mAs reference to mAs effective (R/E) values

PAGE 205

205 Table 7 13 Routine abdominal scan, c ylinder with elliptical add ons, tube current as a function for different TCM mAs reference to mAs effective (R/E) values

PAGE 206

206 Table 7 13 C ontinued

PAGE 207

207 Table 7 14 Routine chest thorax scan, cylinder with elliptical add ons, tube current for different TCM mAs reference to mAs effective (R/E) values

PAGE 208

208 Table 7 15 Routine pelvic scan, cylinder with elliptical add ons, tube current for different TCM mAs reference to mAs effective (R/E) values

PAGE 209

209 Table 7 15 C ontinued

PAGE 210

210 Table 7 16 Backward routine abdominal scan, cylinder with elliptical add ons, tube current for different TCM mAs reference to mAs effective (R/E) values

PAGE 211

211 Figure 7 1 CTDI phantom with and without Lucite PSD adapter inserted Figure 7 2 Cylindrical elliptical add on inserts family ( Photo courtesy of R Fisher Tissue equivalent phantoms for evaluating in plane tube current modulated CT dose MS thesis University of Florida (20 06 ), Figure 3 5 p. 56)

PAGE 212

212 Figure 7 3 Construction using plywood mold ( P hoto courtesy of R Fisher Tissue equivalent phantoms for evaluating in plane tube current modulated CT dose MS thesis, University of Florida (20 06 ), Figure 3 3 p. 55) Figure 7 4 CTDI phantom with in step phantom in CT scanner

PAGE 213

213 Figure 7 5 Topogram of CTDI phantom with in step phantom Figure 7 6. STES CTDI head phantom

PAGE 214

214 Figure 7 7 At 160 mAs R, average slice current as a function of position for two orientations Figure 7 8 At 200 mAs R, average slice current as a function of position for two orientations

PAGE 215

215 Figure 7 9 At 260 mAs R, average slice current as a function of position for two orientations Figure 7 10 Typical phantom orientation, 10 measurements with abdominal protocol

PAGE 216

216 Figure 7 11 Inverted phantom orientation, 10 measurements with abdominal protoco l

PAGE 217

217 CH APTER 8 FEASIBILITY AND CONS TRUCTION OF PROTOTYP E CCD BASED DOSIMETRY SYSTEM FOR USE AT DI AGNOSTIC ENERGIES Background Plastic scintillation detectors (PSDs) have become popular in dosimetry applicati ons because of their unique combination of low cost, temperature independence, water equivalence, small volume, dose rate and energy independent response, and real time capabilities. 12, 17, 65 72, 115 Successful designs have coupled a PSD to an optical fiber which is subsequently coupled to either a photomultiplier tube (PMT) or, more recently introduced, charge coupled devices (CCD s ) and photo diodes. 17, 65 67 CCDs offer two main advantages over the PMT design described in Chapter 3 First, the method for eliminating background noise and stem effect with a PMT design is reference with no scintillator from that of a 66, 70 A CCD could use this subtraction method or switch to another method which takes advantage of the spectral information. 66 Secondly, massively parallel read out is econo mically and physically possible, comparable to Cirio et al. system, which reads 1024 detectors simultaneously. 116 With the average cost of a PMT around $4000, even an array of only 1% of this size (10 PMTs) would cost around $40,000. Each PMT is 3.5 cm wide x 6.0 cm high x 5 cm depth in size and requires its own read out channel. Thus, in addition to a large physical volume, 10 PMTs would require a rather exotic read out hub with an enormous data buffering capab ilities. Furthermore, two PMTs for each measurement point. Using a high performance camera, Lacroix et al. has demonstrated a PSD system could theoretically handle over 15,000 detectors with a precision of 1%

PAGE 218

218 on a 30.7 x 30.7 mm 2 CCD imaging area. 117, 118 Typically, PSD systems have precision around 2 3% using around 10 30 detectors. 92, 118 T he PMT system described in Chapter 3 of this work had a real time ability which was demonstrated in its utility and application in the ophthalmic projec t. Specifically, the real time monitoring was used to ensure patient sa fety during delivery of the x ray treatment. On the other hand, using a CCD does not easily lend itself to real time monitoring. 119 Furthermore, the PMT had extremely fine sampling with up to 100 Hz available T ypically, 150 ms bins were used, so during a typical CT scan which takes about 8 seconds, there will be about 54 measurements. In contrast, acquisitions using a CCD are typically one long single exposure. 119 PSDs have enjoyed success in radiotherapy applications, notably by Beddar et al. 12, 13, 67, 68, 70, 72, 120 However, problems with low signal at diagnostic energies are foreseen due to the lower exposure rate, lower exposure, and lower treatment times. I n radiotherapy the exposures are co mmonly at least one order of magnitude higher and treatment times are measured in minutes rather than milliseconds. In other words, the acceptable signal levels measured by Beddar et al. in radiotherapy applications may not be seen at diagnostic level ener gies F irst the theoretical SNR will be derived and further demonstrate the feasibility of the pro t o type CCD based dosimetry system for standard diagnostic energies and dose rates. Beddar et al. have completed a analogous study where the performance of PM T based PSDs was characterized using SNR. 67 Lacroix et al. later expanded upon this method in order to also characterize the performance of CCD based PSDs. 67, 117, 118

PAGE 219

219 This investigation will then discuss the development of a prototype CCD based dosimetry system and preliminary results obtained. Materials and Methods CCD Theoret ical Signal Level Overall, the goal of optimizing the performance of the system (as measured by the SNR) is accomplished by maximizing the number of photons which are incident upon the CCD. For Poisson statistics, which are an accurate description of radiation detection measurements SNR of a single pixel is defined as ( 8 1 ) he signal can be defined as the product of quantum efficiency, integration time, and photon fluence per unit time per single pixel. 121 The variance in the noi se can be defined as the quadrature sum of the main sources of CCD noise: photon shot noise, dark noise, readout noise, and electronic noise. 121 ( 8 2 ) Applying Poisson statistics to photon shot noise; multiplying dark noise rate, D by integration time; incorporating readout and electronic noise into one term, RE gives ( 8 3 ) Photon fluence per unit time per single pixel can be further broken down a product of: total number of PSD photons produced ; coupling efficiency including geometric considerations for isotropic scintillation as well as optical transmission

PAGE 220

220 between PSD and waveguide, ; waveguide transmission efficiency, ; coupling efficiency between waveguide and CCD lens, ; and pixel area, A A collection of of typical efficiencies between PSD and waveguide for common optical materials and polishing techniques is given by Ayotte et al. 86 Yu et al. also has shown that a correction is required for apodization. 122 More discussion concerning optical coupling will be addressed later in this derivation. An a podization function g is used to model the non uniform illumination or transmission profile that approaches zero at the edg es. 122 ( 8 4 ) The apodization term can be written as the inverse of the total number of pixels on the CCD, n tot and pixel surface area, A 122 ( 8 5 ) This cancels out the pixel area dependence, substituting this into the overall SNR relationship yields ( 8 6 ) Many of the factors in this SNR equation can be estimated using manufacturer acceptance values or by other work using the same materials. These factors include: total number of photons pr oduced by the scintillators (adjusted for gain setting), coupling efficiency between the scintillators and waveguide, CCD quantum efficiency, total number of CCD pixels, pixel area, integration time, dark current rate. Similarly,

PAGE 221

221 noise terms are commonly c haracterized CCD performance metrics: dark noise, readout and electronic noise. Therefore, the SNR for each pixel can be calculated. For this application, photons will be incident upon groups of pixels rather than just one pixel. Thus, if n individual pixe ls (which are assumed to have the same SNR) are used to gather the information, then the SNR 0 improves by a factor of Klein et al. 115 Furthermore, this CCD also allows f multiple frames to be combined to create a single image, which further improves the SNR 0 by a factor of Explicitly written, the SNR relations hip becomes ( 8 7 ) Yu et al. have studied the lens coupling efficiency for a variety of digital radiography schemes. 122 An appropriate first order approximation from Yu et al. is ( 8 8 ) In this scenario the lens coupling efficiency is derived with the assumption of the fiber as a point source and a uniform distribution of light within the cone defined by numerical aperture. In Equation 8 8, m is the magnification, which is defined as the quotient of image and object sizes; F is the f number. In other words, the lens coupling efficiency is roughly inversely proportional to the square of the f number. F number will be discussed in the next secti on. At this point, an electron to signal conversion coefficient, w is introduced and defined as the number of electrons required to register a single digital grayscale level. In

PAGE 222

222 order to maximize performance, w is chosen by the manufacturer such that the full well depth is proportional to the full bit depth. This is analogous to a window level setting such that a smaller w can give much higher sensitivity with a tradeoff in total range. This coefficient reduces the SNR ratio by a factor of Substitutin g Equation 8 8 into Equation 8 7 gives an explicit expression of scintillation photon fluence per unit time per single pixel. The final expression for SNR can be written as ( 8 9 ) Equation 8 9 will be used for experimental validation. In each expression for SNR, for the limit of high photon fluence, photon shot noise becomes the dominant noise term. In simpler terms, this expression is given as the root of the photons produced, coupling efficiencies, and collection time ( 8 10 ) Using the appropriate values, Equation 8 10 shows that the SNR is about 5% of the SNR derived in Chapter 3 for the successful PMT based dosimeter. This SNR is still acceptable and Equation 8 10 shows that a CCD design is at least feasible. Furthermore, man y of the factors in this Equation 8 10 were estimated using the lower range of manufacturer acceptance values or other work using the same materials. Therefore this SNR calculated for the CCD should be used only as an order of magnitude measurement. Note that the equations governing objective lenses were derived using the thins lens approximation which holds in the limit of the distances between the source, lens,

PAGE 223

223 and observer being much larger than the size of the lens Practically, most of assumptions in the thin lens approximation do not change the results. 123 However, an aspect which could potentially alter the results is vignetting, which is a reduction of an image's brightness or saturation at the periphe ry compared to the image center. 123, 124 This approximation is mentioned by Ray et al. Smith et al. and Lacroix et al. but regarded as a negligible effect when utilizing a high performance lens. 118, 123, 124 Therefore, the theoretical SNR derivation shown in Equation 8 10 should be used only as a first order approximation to the optimization of CCD design. CCD Physical System Design PSD and w aveguide The pro toty pe CCD based PSD system involved the same design as the fiber used previously for the PMT based system described in Chapter 3. This design involved fiber optically coupling the PSD to a waveguide. The PSD was the same water equivalent scintillating fib er (BCF 12, Saint Gobain Crystals, Nemours, France). The PSD is a cylinder with 500 micron diameter and 7 mm height. Note that i n anticipation of lower signal, a longer PSD was used than the PMT based system which employed a cylinder with 500 micron diamet er and only 2 mm height. BCF 12 scintillating fibers emits in the blue region with a peak at 438 nm with a polystyrene core, a polymethyl methacrylate (PMMA) cladding, and a refractive index of 1.59; ESKA waveguides are plastic optical fibers with a PMMA c ore, a refractive index of 1.49, with loss of 375 dB/km at 450 nm. 77, 78 Note the water equivalence (1.05 g/ cm 3 ) of both the fiber and the waveguide. Absence of surface polishing reduced the light collection by approximately 40% in a study by Ayotte et al. therefore t he ends of both the scintillator and waveguide we re

PAGE 224

224 polished for 20 to 30 minutes per surface using sequentially finer grit lapping paper (12, 3, and 1 micron) in order to facilitate light transmission. 79, 86 The other end of the scintillator was also polished, furthermore it was coated with reflective paint (EJ 510, Eljen Technology, Sweetwater, TX) in order to capture more scintillation photons. 79 Schematics and a picture of a completed fiber w ere shown p reviously in Figure s 3 9 and 3 10 respectively In addition to the longer scintillator length the fibers used for the CCD system differ from th os e described in Chapter 3 because they were directly incident upon the optics which are used to couple the op tical waveguide to the face of the CCD instead of the earlier design which was coupled using SMA connectors. W aveguide s were held securely in place using a post mountable v clamp (VC1, Thorlabs, Newton, NJ). The waveguide was centered in the field of view and placed 8 mm away from the optical set up. The lens used is described in more detail below. CCD c amera An electron multiplying (EM) CCD (Luca, Andor Technology, Belfast, Ireland) was used in order to capture the images. An important feature of this CC D was its EM ability In an EM system, a gain register is placed within each stage of the CCD capture which multiplies the number of electrons, thus making collection of signal from small numbers of electrons possible. The gain setting used was 225 (out of a maximum 255) for all image captures. This number is unitless and varied even among cameras of the same model. The brightness of the image as a function of gain was tested explicitly and is shown below in Figure 8 1 One drawback of EMCCD cameras is the require ment of a cooling system to maintain an extremely low temperature. T h is EMCCD required a cooling system which operated the system at 20 degrees Celsius.

PAGE 225

225 CCD h ousing An electronic enclosure box ( 1415 Hammond Manufacturing Cheektowaga, NY ) was used to shield the CCD from ambient or background light. In addition to protection, the housing also provided a convenient carrying module. Necessary holes for the CCD system included: filtered power entry mount (1 large, 2 small), USB port ( 1 large, 2 small), fiber hole (3 small), L backet array support (3 medium), and breadboard either M5 or 1/16 inch drill bit are 1/4 custom openings with edges on the order of 1 inch. In addition, new holes were drilled into the top lid of the enclosure for incorporation of convenient 6 32 (size) thumb screws. Holes were drilled and then a tap screw was used to create the threads for the thumb screws. tin lead solder with no clean flux core (N CCW2, Amerway Inc, Altoona, PA) and shrink tubing. The housing box is composed of three main parts: a body and two lids. Thus seams and corners on the body were first sealed with two coats of black rubber cement and then taped. All three pieces were painted with pri mer which was then followed by two coat s of flat black paint in order to min i mize reflection. The bottom lid was permanently attached to the body of the housing using black rubber cement and the seams were painted again. Thin rubber strips which serve as a gasket were placed along the top of the body in order to prevent light leakage. Additionally, rubber feet were attached to the bottom of the box and Velcro strips w ere used for organizing cables. A small hole was drilled in the front of the box in order to insert the optical fiber. The fiber entry hole is shown in Figure 8 2 with a coin to provide a frame of reference

PAGE 226

226 Four 1/4 20 x 2 inch screws were used to attac h the breadboard. The v clamp was attached to the breadboard using one 1/4 20 x 1 inch screw. 1/4 20 x 1 screws were used for attaching the L bracket array support as well. The completed CCD housing with power entry module, USB port, and L bracket support is shown in Figure 8 3 The was tested. Optical c oupling As mentioned earlier, the end of the fiber was directly incident upon the detection system. A singlet plano convex was used to focus the output of the fiber onto the CCD accurately. There are traditionally six types of singlet lenses: plano convex, bi convex, plano concave, bi concave, positive meniscus, and negative menisc us. These are shown in Figure 8 4 Of these, the plano convex, bi convex, and positive meniscus lens are known as positive lens, meaning they are used for focusing. 123 On the other hand, bi concave, plano concave, and nega tive meniscus lens are known as negative lens, meaning they are used to spread light. To improve the signal, the ends of the waveguides were place d within the focal region of a traditional plano convex spherical lens spherical lens (LA1116 A, Thorlabs Inc Newton, NJ). These lens were constructed of a polymer and anti reflective coating designed for transmission at 350 700 nm. The results from this simple plano convex spherical lens were promising, but were immediately improved by a switch to an aspheric lens which provided images with less aberration There are four types of aberrations which can be improved using aspheric lenses : spherical, marginal, distortion, and chromatic. Spherical Aberration causes blur because light rays passing through the edges focus sooner than other rays which pass passing through those rays passing through the center of the lens. 123 This is shown in

PAGE 227

227 Figure 8 5 below. Marginal aberrations are caused when parallel beams are not incident complete ly perpendicular with the lens. 123 This creates two focus points and is also known as astigmatism. Distortion is caused because rays passing through the edges rays at the edge increase magnification. 45 This is what causes the pincushion effect commonly seen in fluoroscopy. Chromatic aberrations are found when there is a li ght spectrum rather than a monochromatic source. In this aberration, c olors with shorter wavelengths are bent quicker and sharper than the longer wavelengths. 122 The aspheric lens (Geltech C140TME A, LightPath Technologies, Orlando, FL) was mounting within a lens tube (SM1L10, Thor labs Inc, Newton, NJ) which used SM1 threading. This was connected to the CCD which used an c mount threading using a simple adapter (SM1A9, Thorlabs Inc, Newton, NJ) Th e lens has an anti reflective coating and transmits 350 700 nm. The lens was first mou nted in an optics adapter (S1TM12, Thorlabs Inc, Newton, NJ) disk which was subsequently threaded into the lens tube mentioned previously and held in place using a retaining ring. Figur e 8 6 shows the lens inside the optics adapter which is threaded inside the lens tube. The position of the lens disk within th e tube was manipulated using a custom spanner wrench (SPW909, Thorlabs Inc, Newton, NJ). For this investigation, using one lens made this design feasible. Using the aspheric lens further decreased the FWHM by around 15% in comparison with the plano convex lens. Signal improvement t echniques The signal was improved using two categories of techniques: improving the scintillation efficiency and improving the detection efficiency. The transmission effi ciency was increased with a substitute waveguide wh ile implant ation of a novel lens system was undertaken in order to improve detection efficiency by calibrating and focusing the

PAGE 228

228 light through manipulation of a lens system. At the time of publication, t h is lens system has not been fully characterized, therefore it will not be discussed, but will be di scussed extensively in Chapter 10 in the future work section. Waveguide r eplacement In order to increase the signal, a different s ilica waveguide optical fiber (400 UV, Ocean Optics Inc, Dunedin, FL) was used. This was the original waveguide used by Hyer et al. while the current waveguide as described in Chapter 3 is plastic 17 Compared with the plastic waveguide previously used, this waveguide is more expensive and does not optimize the geometrical coupling efficiency be cause the s ilica waveguide is slightly smaller in diameter (400 microns) than the PSD diameter (500 microns). However, the improved transmission at the blue 400 nm wavelength was expected to outweigh this. 125 Th e brown s ilic a wav eguide is shown in Figure 8 7 on the left while the clea r plastic waveguide is shown in Figure 8 7 on the right The transmission spectrum of the silica w aveguide is shown in Figure 8 8 Calibration In order to position the waveguide accurately within the field of view (FOV) of the CCD, a waveguide was illuminated using a strong light source. This was done for alignment in the x y plane (parallel with CCD face) as well as for proper focusing within the z plane (normal to CCD face). Figure 8 9 shows a s eries of ima ges of fiber to lens distances of 0.5, 1.5, 2.5, and 3 mm (left to right) from the lens. Note that the light source wa s actua lly so intense that it saturated the illuminated area as the waveguide comes into focus.

PAGE 229

229 Baseline Performance Testing U sing Portable X ray Unit The portable SR 115 x ray unit described previously in Chapter 3 was used in order to provide diagnostic energy photons. 81 A PSD was placed in the field, at 100 kVp, 1.50 mAs. Since the exposure time was purposely set to 1.0 seconds, remote triggering was not necessary. The fully assembled CCD unit is shown with the po rtable x ray unit in Figure 8 1 0 A background image was taken for five minutes prior to each set of exposures. The first set of exposures consisted of response for different gains. The background was then subtracted from the response in order to form a signal curve. Results As mentioned in the introduction, PSD systems based upon CCDs have enjoyed extensive success in radiotherapy applications by many groups, notably with its introduction by Beddar et al. 12, 68 However these systems have two advantages in producing signal with respect to diagnostic systems: higher exposures and longer exposure times. Typically, therapy ap plications use current on the order of amps, with treatment times on the order of minutes. 45 On the ot her hand, diagnostic scanners use current on the order of milliamps, with treatment times on the order of 500 milliseconds. The most important aspect of this investigation was attaining an acceptable SNR. The SNR was derived theoretically in Equations 8 1 through 8 10 to be high enough to be detectable. While this CCD methodology has been used successfully in radiotherapy applications, acceptable images have not been attained by any other group operating in the diagnostic energy regime T h gain which raise the signal by two orders of magnitude and the custom housing which lower the ambient signal by an order of magnitude are expected to bring the SNR high enough for appropriate discrimination

PAGE 230

230 Waveguide R eplacement The original plastic fiber had a transmission loss of 0.22 dB per meter at 400 nm while the new silica fiber had an attenuation of 0.09 dB per meter at 400 nm. 78, 125 Over the 2 m fiber, this resulted in an attenuation of 0.44 dB in the plastic fiber while there was an attenuation of 0.18 dB in the silica fiber. Because of the logarithmic scale on which decibels are defined, this was significant. More importantly, the silica waveg uide was much easier than the plastic waveguide to work with because of the higher melting point. The silica and plastic waveguides had operating temperatures the temperatures with no degradation in optical properties of 300 and 70 degrees Celsius. 78, 125 Note that the operating temperature is slightly lower than the temperature at which physical melting occurs, at which poi nt the optical properties are almost completely destroyed. 78, 125 As detailed in Appendix A, the fabrication process is heavi ly reliant upon using heat to conform the outside tubing in order to create the primary coupling between the waveguide and PSD. Additionally, heat shrink tubing was used to couple the fiber to the SMA connectors and for the bifurcation. And, lastly, heat s hrink tubing was used to create a smaller, more point like detector which did not displace the measurement medium. Housing Performance A 250 x 250 pixel ROI was obtained using a blank capture using the housing with the lights off and yielded a mean of 502 +/ 8.9 (1.8%) counts. This was repeated with the lights on and yielded an identical mean of 502 +/ 8.9 (1.8%) counts. This independence to ambient lighting shows excellent shielding by the box. Using no housing, with the lights off, yielded a 808+/ 40 (5.0%) counts. If the lights were turned

PAGE 231

231 on, the CCD would be saturated. This further shows the shielding by the box in still lowering the counts to the CCD. This is summarized in Table 8 1 Gain Performance The linearity of maximum pixel value was measur ed as a function of the EM gain. This wa s previously shown in Figure 8 1 Before saturation, t he CCD showed a multiplication of a factor of 97.3 with a linear correlation coefficient of 0 .976. During saturation, the maximum pixel value was 16383. T he PSD produced a signal level such that a gain of 225 out of 255 was the highest EM gain which remained in the linear region and did not saturate the detector Lens Effect Figure 8 11 shows the output of the CCD both with and without the lens The maximum numbe r of counts measured was 74 66 with the lens and 1 51 1 without the lens. Centering a 250 x 250 pixel region of interest (ROI) around the center cylinder yields 3.43 x 10 8 and 7.35 x 10 7 counts for lens and no lens, respectively It was difficult to align the fiber within the x y plane and this fine tuning will be discussed in future work of Chapter 10 The mean value is 5489 +/ 868 (15.8%) and 1176 +/ 141 (12.0%) counts for lens and no lens. The image with the lens had 494% more max counts than the image wi th no focusing lens with only minimal loss (3.8%) of precision. The SNR is improved by a factor of 3.5. Initial Testing I mage s were obtained from the portable x ray unit at 100 kVp and 15 mAs. Furthermore the SNR is not very good. The mean value of the s ignal fiber wa s 1434 +/ 792 counts while the surrounding background value wa s 504 +/ 855 counts Note that ROIs are 50 x 50 pixels. In an attempt to obtain acceptable SNR special techniques

PAGE 232

232 such as pixel binning and longer shutter times were attempted. Using electronic binning into 2x2 pixels rather than 1 pixel, t he mean value of the signal fiber wa s 1935 +/ 1069 counts while the surrounding background value wa s 685 +/ 830 counts U sing longer CCD shutter timing t he mean value of the signal fiber is 1026 +/ 884 counts while the surrounding background value is 639 +/ 1092 counts These results are summarized in Table 8 2 along with the measured exposure s which ranged from 1.8 to 3.1 R. This range of exposures was not sufficient to produce an adequate SNR to visually distinguish the fiber CCD Radiation Shielding Initial images possessed a streaking artifact across the image, which was noticeable only after applying a chromatic filter across the entire image. A chromatic viewing filter (CVF) uses the entire color spectrum rather than a grayscale and makes it easier to visually distinguish patterns. These streaks were completely parallel and horizontal across the image. An example is shown in Figure 8 1 2 They were non uniform in intra streak spacing. Exposures taken under the same conditions possess streaking in different locations and different spacings. This was attributed to a register error or electronics issue. However, this was disproven when blank background images with no irradiation did not p ossess this feature. It was then theorized that this artifact was due to possible saturation. However, this artifact was not seen in images taken with fibers using the calibration fibers which were much brighter saturating the CCD without any gain. After adding steel blocks between the CCD housing and the x ray tube, these artifacts were no longer present in the image Therefore, this artifact was attributed to stray radiation.

PAGE 233

233 Groundbreaking Resultant Images T he CCD and fibers were placed within the light protection box and the large PSD fiber was used. Measure ments were taken with steel blocks placed between the x ray tube and the CCD to prevent the horizontal streaking artifact seen earlier The gain was empi rically determined to be optimally placed at 225 by measuring the signal and maximizing the gain without saturation and while still within a reasonably linear region. The resultant image is shown in Figure 8 1 3 and represents the first successful implement ation of diagnostic application of a CCD PSD system. The CCD performance w as next tested by measuring a 25x25 pixel 2 area within the center of the image. The average value was referred to as the pixel intensity This was first measured as a function of cu rrent which was varied from 3.0 to 6.75 mAs. The resul ts are extremely linear, with a correlation coefficient ( R 2 ) of 0.9980. This is shown in Figure 8 1 4 and the results are shown in Table 8 3. Next, the stability of pixel intensity was tested at a curren t of 4.5 mAs and 100 kVp. The standard deviation between measurements was 0.5% and the r esults are plotted in Figure 8 1 5 The pixel intensity was next measured as function of mAs for 80, 90, and 100 kVp. The results are shown in Figure 8 16 The results h ad R 2 linearity of 0.9996, 0.9995, and 0.9997 for 80, 90, and 100 kVp, respectively. The calibration factor between mAs and pixel intensity is further plotted as a function of current between 3.0 and 7.5 mAs in Figure 8 1 7 MATLAB Automated ROI Analysis An important aspect of this investigation was to demonstrate proper sampling within the ROI. Essentially, for each ROI, it was desirable to maximize the SNR, typically by maximizing signal This was accomplished by proper selection of the pixel

PAGE 234

234 sampling area A versatile definition of this area was demonstrated as a value dependent upon only the measured FWHM. The source code for this program is included in this work as Appendix D. The first part of this MATLAB program developed an initial estimate of the ROI average value using a user defined center located at coordinates (c1,c2) within the image. The program applied a smoothing filter and line fit using local regression based upon weighted linear least squares and a 1st degree polynomial model. This aspect of the linear least squares function assigns a zero weight to any data which are greater than 6 sigma away in the moving average. After this, the function goes through and measures the FWHM using linear interpolation. This process is done for the 10 rows a bove and 10 rows below the user defined center. If the variation between these FWHMs is too large, then the program recognizes this, exits, and suggests a different definition for the user defined center (c1,c2). F igure 8 18 shows the results with the mean pixel intensity as a function of ROI sampling area. The values in this particular figure vary from 5 to 150% of the constraint. This figure shows an expected maximum value as the ROI expands beyond the illuminated area. Furthermore, this figure shows that the value converges to the background value which is around 700 counts Again, the constraint was defined as the average FWHM value minus a buffer of 10 pixels and then divided by the square root of 2 This correction factor is used because the radius of the circle is larger than the ROI rectangle by a factor of square root of 2 This constraint value is 712 or 5041 pixels sampled, where the mean pixel intensity value is 1140 counts The maximum pixel intensity value is seen with 900 1000

PAGE 235

235 pixels, where the mean pixel intensity is 1740 counts Therefore, a 30x30 array would be an excellent choice for this particular setting. The robustness of this program was tested by changing the focal length of the CCD by varying the lens to detector distance. The impo rtance of appropriate ROI selection has been demonstrated using only the mean pixel intensity. The standard deviation can also be used as a metric, and has been included in the coding as another method of data analysis. In the future, this entire ROI tool must be one of the first tests run on an acquired image. This investigation demonstrates the ability of this code to provide a mathematically sound way to define a proper ROI for each image or portion of an image. Analysis After adding shielding, changing the PSD measurement fiber, and creating a custom housing, this prototype CCD has produced excellent results. Testing has already been undertaken on its next generation which utilizes a lens system and promises even better results. This prototype came afte r much develop ment: t he very first physical results were not encouraging, with large uncertainties seen in the measurements. Continued work was justified because this prototype CCD PSD system used the same Andor CCD as other successful groups such as Bedda r et al. and his group (Archambualt et al. and Lacroix et al. ). 12, 13, 65 68, 70, 86, 92, 117 120, 126 While this application used a much lower current and much smaller exposure time, it was shown to be theoretically possible. It was difficult to discriminate the signal from the background in many cases and a CVF was used for easier visual distinction of pa tterns. After applying these CVFs there was extremely noticeable horizontal streaking across the entire face of the images which

PAGE 236

236 was eliminated using steel blocks in order to shield the CCD acquisition system from stray radiation. The average CCD pixel intensity was found extremely linear as a function of current, with an R2 of 0.9980 from 3.0 to 6.75 mAs. Next, the variation between measurements of pixel intensity was found to be excellent, at 0.5%. The pixel intensity as a functio n of mAs was extremely linear for a variety of energies with an R 2 value of 0.9996, 0.9995, and 0.9997 for 80, 90, and 100 kVp, respectively. system which was used at the diagno stic energy level. Using this PSD at diagnostic energies instead of radiotherapy applications does provide some advantages. Notably, there is no stem effect for Cerenkov because energies are below the 178 keV threshold for PMMA. 13 background subtraction. Background subtraction for removing stem effect is not as precise as chromatic filtering. 92 Howeve r, chromatic filtering was not used because requires an extremely high SNR. 92 Note that the chromatic filtering technique described by Archambault et al. is not the same as a chromatic viewing filter (CVF). 92 Chromatic filtering is used to remo ve Cerenkov radiation and other wavelength dependent artifacts from images while CVFs are used to simply display images in color rather than in grayscale. The limiting resolution factor comes in either longest dimension or in this case normal to lengt h of fiber due to requiring measurement signal from both fibers. The use of a blank fiber does lower spatial resolution. However, this is as not as important as in therapy applications for two reasons: (1) orientation is not as hard to change in diagnosti

PAGE 237

237 acquisitions and (2) penumbras are nowhere near as sharp percentage wise as in brachytherapy applications which demand high spatial resolution. Prototype Development Compu tational o ptimization of SNR The next stage of development should also include improvements in SNR This would be accomplished by proper sampling within the ROI. A versatile definition of ROI was demonstrated as a value heavily dependent upon FWHM. A MATLA B code has been developed to automate this process. There is an important geometric correction factor based upon using a rectangular ROI instead of a circular ROI. Using a rectangle instead of a circle has two advantages: first, computationally it is much easier and faster to incorporate. Secondly, any statistical anomalies across the circular area will tend to be averaged out For example, if an artifact were present only at the 70% isodose line this would not be detected by a circular ROI. But a rectangul ar ROI with the same area 2 area would mean a circle with radius of 10 pixels or a rectangle with length of 17.7 pixels. CCD lens s ystem An ideal fiber for focusing would have no divergence. Ho wever, because this fiber is not an idealized case, there is some degree of divergence to it. However, this can easily be accounted for by using negative lens. Negative lenses were mentioned previously, but were described as those which spread out a beam o f light. If focused on an already diverging source, a negative lens can be used for collimation. This creates a beam with a focal spot of infinity and makes the light directly incident perpendicularly upon the next object. 123

PAGE 238

238 The concept of collimation is easy to visualize as a backwards version of the typical positive lens focusing. In the positive scenario, a beam of light is incident upon the curved portion of the lens and focused in to a point. Conversely, in a negative scenario, a point source is incident upon the planar portion of the lens and spread into a beam. The effectiveness of this diverging lens was tested by moving the lens closer and further away from the CCD while maintaining the fiber at a focal length away If the collimation is occurring correctly, the image width should not change regardless of distance of the lens from the CCD. This was initially conducted by flipping the aspheric lens such that the light from the fiber was incident upon the planar side rather than the convex side. From general optics, the numerical aperture for a single mode fiber is 123 (8 11) It is given that refractive index, n is equal to 1 in air. 123 Furthermore, the NA of the fiber is specified as 0.22. 125 Therefore the half angle of the emission (or acceptance) cone, is 12.7 degrees. At a distance of one focal length away from the collimating condenser lens, F C then the collimated beam then possesses a width of (8 12) Note that the factor of two comes in because the emission cone is specified as a half angle. Therefore, with the specified focal length of 8 mm for the collimating conden ser lens, the collimated beam width is 3.6 mm. 127 Finally, with the 400 nm

PAGE 239

239 wavelength of light being transmitted and the 20 mm focal length for the focusing condenser lens, F F the spot size, d can be determined by 123 (8 1 3 ) This leads to a spot size of approximately 2.8 x 10 6 m. In other words, the theoretical spot size is on the order of single digit microns. The pixels themselves are 10 microns. After this quick proof of concept, a condenser lens was ordered. The set up is shown in Figure 8 19 Aspheric condenser lenses have a non spherical curved face but, by definition, have a short focal length which is ideal for both positive and negative focu sing, such as beam collimation. 123 For optimal coupling, the condenser collimation lens should have a higher or equal NA in comparison with the focusing lens. 123 The aspheric condenser len s chosen (ACL 108, Thorlabs Inc, Newton, NJ) had an NA of .547 while the focusing aspheric lens described previously had an NA of 0.500. 127 This specific series (ACL, Thorlabs Inc, Newton, NJ) of lenses was designed for high efficiency collimation. This particular lens was also coated with an anti reflective coating which allowed transmission of light in the 350 to 700 nm range. 127 Novel target adjust ment d evice Figure 8 2 0 shows a schematic of an innovative proposed target adjustment device (TAD). This device would be used to quickly and precisely tune the position of the output of the lens system. It is envisioned that this device would be attached to the end of the collim ated light housing. There would be four holes in each of the corners of

PAGE 240

240 the TAD. The exact amount of adjustment would be dependent upon the threading of the screws used. It is recommended that a large thread pitch (with respect to Unified standards) is ch osen. This corresponds to a small thread pitch with respect to metric threads. In either case, this corresponds to a large number of threads per inch. Additionally, set screws are also recommended because they are fully threaded, thus providing maximum ut ility in what will most likely be a very short screw. Figure 8 2 0 shows a socket screw simply because it was easier to illustrate. The space between the end of the collimated light housing and the TAD should be light tight and it is suggested that some sor t of o ring be used in order to maintain sp acing. This is shown in Figure 8 2 0 Either a cloth or rubber o ring should be sufficient. Discussion this is the first system to acco mplish these measurements at diagnostic level energies. The SNR levels of the PSD CCD was shown to theoretically to satisfy the discrimination of an image Next, the system was made suitable for measurements by constructing a custom housing, a higher signa l PSD fiber, and scatter shielding. This prototype is the first generation in an expected series of systems. The average CCD pixel intensity was extremely precise, with less than 1% change between measurements. Furthermore, the pixel intensity was found t o be linear as a function of exposure for a variety of energies This investigation has shown the feasibility of a CCD based PSD system. Future work on this system will include the incorporation of a lens system for improved optical

PAGE 241

2 41 coupling betw een the fi ber and the CCD. The incorporation of a condenser lens has been started and is considered to be the second generation This investigation has also demonstrated the ability of an ROI analysis code to provide an optimized selection of a proper ROI for each i mage or portion of an image. A n improved ROI algorithm is under development to process the images in a more automated and faster method. The next generation will also include the capability for multi ple parallel fiber measurements most likely by utilizing the novel target adjustment system. This research was funded by Oraya Therapeutics.

PAGE 242

242 Table 8 1 Counts as a function of housing and lighting Table 8 2 Effects of special techniques upon counts Table 8 3. Linearity of pixel intensity as a function of current

PAGE 243

243 Figure 8 1. Gain linearity Figure 8 2 Size of fiber entry hole coin used for reference

PAGE 244

244 Figure 8 3 Completed CCD system housing ( top ) and a schematic ( bottom) showing 1) power supply, 2) optical post, 3) focusing lens, 4) CCD camera, 5) power entry mount, and 6) USB panel mount

PAGE 245

245 Figure 8 4 Different types of lenses (from left): simple plano convex, biconvex, plano concave, biconcave, meniscus, and plano conv ex Figure 8 5 E ffect on edges by aspheric lens ( Photo courtesy of Figure 8 6 Lens mounted in optical adapter threaded into lens tube

PAGE 246

246 Figure 8 7 Brown silica waveguide (left) and clear plastic waveguide (right) mounted on v clamp inside focal region of lens Figure 8 8 Silica waveguide transmission vs wavelength ( Photo courtesy of Polymicro,

PAGE 247

247 Figure 8 9 Resul tant images with fiber to lens distances of 0.5, 1.5, 2.5, and 3 mm (left to right) Figure 8 10 Completed CCD housing ready for x ray source benchmarking

PAGE 248

248 Figure 8 1 1 Lens effect showing focused image (left) and unfocused image (right) Figure 8 12 Streaking artifact

PAGE 249

249 Figure 8 1 3 Final image produced with proper housing, fiber selection, and shielding shown with out (left) and with (right) CVF Figure 8 1 4 Pixel intensity linearity as a function of mAs

PAGE 250

250 Figure 8 1 5 Stability o f pixel intensity from multiple trials Figure 8 1 6 Pixel intensity as a function of mAs for different peak voltages

PAGE 251

251 Figure 8 17 Calibration factor between mAs and pixel intensity for different peak voltages Figure 8 18 Mean pixel intensity vs ROI sampling area

PAGE 252

252 Figure 8 19. CCD double lens set up ( top ) and a schematic ( bottom) showing 1) optical post, 2) condenser lens, 3) focusing lens within lens tube, 4) CCD camera

PAGE 253

253 Figure 8 20. Novel target adjustment device

PAGE 254

254 CH APTER 9 MULTIPIXEL PHOTON CO UNTERS Background Chapter 8 investigated the next generation of prototype PSD detectors using a CCD based design. The next investigation is devoted to evaluation of another different prototype system using a multipixel photon counters (MPPCs) based design. MPPCs were recently innovated and were not available in wide spread commercial mo dules until very recently. P reliminary results are examined and performance is compared with the curr ent PMT based design. The idea for using MPPCs w as borrowed from current detectors which are installed in the prominent Large Hadron Collider experiment. MPPCs are based upon avalanche photodiodes (APDs). These detectors were invented in Russia in the 1980 They have an internal gain of 10 5 to 10 6 are extremely robust, and have detection efficiencies as high as 60%. 128 131 On the other hand, they do have a high crosstalk and high dark count rate. 129 131 Their size can be characterized as either an advantageous (compact, high resolution) or a disadvantage (small cross section). In the case of PSDs, the small size is not an issue because the waveguide is still smaller in diameter (500 microns) than the active area (1000x1000 microns 2 ) MPPCs operate as an array of avalanche photodiodes (APDs). APDs operate with their voltage bias above breakdown voltage, in Geiger mode. 131 Therefore, the APD is a binary device either a photon creates a charge or it does not. In turn, APDs are not dependent upon the number of photons which impinge upo n the sensitive area; they are therefore count rate limited. On the other hand, common detectors are amplifiers and

PAGE 255

255 apply a gain to photons, such that the amount of charge created is proportional to the number of photons which enter. 45 APDs have extremely high sensitivity for single photons and some modules even have an adjustable trigger level such as 0.5 photoelectrons (pe), 1.5 pe, and 2.5 pe. 128 The increments begin at half of particle in order to provide a proper threshold for each event with a reasonable uncertainty range. For example, if a 0.5 pe threshold should be sufficient to discriminate against noise while allowing some leeway in the charge created by one photoelectron. Note that the output of each APD pixel is considered. If the number of photons is low and they arr ive at a time interval shorter than the recovery time of the pixel, the APD can output pulses that are equal to a single photoelectron. However, if the photon flux is high or the photons arrive with a frequency which is higher than the recovery time, the p ixel outputs will begin to pile up. This is when the threshold of 1.5 pe, 2.5 pe, or 3.5 pe is used. The recovery time has been shown to be less than 100 ns for most APDs arrays. 129 131 APDs have extremely fast timing properties with resolution available down to 100 ns range. Most commercial modules offer resolution down to 1 ms, which is an order better than th e 10 ms maximum resolution available with commercial PMTs. Remember that APDs are operated in Geiger mode which is about 10 20% above breakdown voltage, which makes essentially immune to small bias voltage drifts. 131 In general, the low voltage requirements for APDs is also an advantage and most modules ns. 128 Thus they require neither an extra power supplies nor other cables except the necessary readout cable.

PAGE 256

256 While there is quite complicated circuitry, overheating has not b een a concern in the module and can be operated for long periods without overheating concerns. However, while APDs do not overheat, their temperature can fluctuate. O ne potential sensitivity to temperature changes. Using Geiger mo de does mitigate this effect, but temperature stabilization or compensation should still be considered. One method of reducing dark current in any detector is to lower the dark current noise and a common way of doing this is to use cooling. 45 APDs possess a large advantage by using Geiger mode which offers high intrinsic gain. This makes another PMTs (or a ny other type of high performance amplifiers) unnecessary. On the other hand, the principle disadvantage of APDs is that Geiger mode is a binary operation. All that can be determined from an avalanche is that one electron/hole pair initiated the breakdown, but not the number of electron/hole pairs. The concept behind MPPCs is to use 500 1000 independent APDs within a small (1 to 9 mm 2 ) area. 128 131 Each of the APDs are connected in parallel and, to first order, the sum signal of all the cells added is, to first order, directly proportional to the number of photons impingin g upon the whole sensor surface. 128 131 However, this is only true if the number of photons is small compared to the number of pixels. 128 131 Preliminary Results The detection unit (C10507 11 025U, Hamamatsu, Bridgewater NJ) chosen was a commercially available photon counting module. This module comes preinstalled with the hardware required to control the MPPC (S10362 11 025U, Hamamatsu, Bridgewater, NJ). This MPPC is a square 1 mm 2 with 1600 pixels which are 25 microns x 25 microns each, with a nominal gain of 2.75 x 10 5 128 Because the pixels are so small, the fill factor is relatively low (0.38). 128 However, remember that the linear output

PAGE 257

257 relies upon the number of pixels being much larger than the number of photons. The time resolution, which is defined as the minimum time difference that can be detected the APD is 200 ps. 128 This is essentially the reset time of the APD. This is not the same as the electrical readout binning resolution, which is 1 ms. The peak sensitivity is at 440 n m. 128 Within 20 to 30 degrees Celsius, there have been no discernable changes in stable analog output. 128 This module has an adjustable trigger level between 0.5 pe, 1.5 pe, and 2.5 pe, or it can be disabled. The module is shown with an SMA adapter installed in Figure 9 1. The typical MPPC response to a computed radiography machine is shown in Figure 9 2. The source was the same portable x ray tube used previously (filtration of 2.7 mm Al). The background noise in the MPPC was substantially higher than that of the PMT system. Typically, the PSD fibers had around 5 10 background counts which was enoug h to trigger an event in the MPPC due to its low threshold as described previously As shown in Figure 9 2, background measurements remained very consistent. Over 16 different mAs values, each with three trials, the background was 2601 0.4% counts. Howe ver, these background counts were relatively high in comparison with the signal. It was observed that the number of counts changed with the amount of ambient light. Specifically, the number of counts was higher in the presence of ambient light. Therefore, it is assumed that there is light leakage somewhere in the fiber. However, this was mitigated somewhat by taking all measurements with overhead lights turned off. This effect can be further suppressed by constructing an enclosure similar to the ones detail ed in Chapter 3 for the PMT and Chapter 8 for the CCD. Another test which

PAGE 258

258 was tried was by simply putting a sealed SMA connector onto the MPPC. The resultant counts were still very high, suggesting some light leakage in the threading or connection between the MPPC and the SMA. Next, the portable x ray generator was used to generate a signal in the PSD to be recorded by the MPPC. The mAs was varied from 2 to 9.75 mAs. The number of signal counts as a function of mAs was determined to be roughly linear, with an correlation coefficient (R 2 ) of 0.9557. This is shown in Figure 9 3. Analysis and Discussion The MPPC must have better performance than the PMT system in order to justify a switch. Requirements to be fulfilled include: robust and stable, easy calibrati on, blue sensitive, low cost (optimally including initial investments), compact, low power consumption, no extra cooling necessary, linear photon detection efficiency. Other useful properties include insensitivity to magnetic fields and self powered. Whil e the PMT offers much better SNR by a factor of almost 100, each channel costs $4000. MPPCs offer an intriguing option due to its price. The MPPC array which seems to fit the criteria bets is a 16 channel (4x4) package (S11827 3344MG, Hamamatsu, Bridgewate r, NJ). 132 Each channel in this module has a large 3x3 mm 2 sensitive area comprised of 3600 pixels with 50 micron pitch. 132 In addition this is chip is fabricated in a monolithic fashion which means that the gaps are small in fact the fill factor (.615) is 150% higher than the single channel module used in this proof of concept (.38). 128, 132 Each channel has two output pins which can be fitted into a module and this array costs $1300. 132 In order to control the MPPC, it is recommended that the module version (C11206 0404FB[X], Hamamatsu, Bridgewater, NJ) is purchased which

PAGE 259

259 raises the cost to $4000. The module versio n contains the control electronics, preamp, and DAC array. The MPPC response is shown with the PMT response overlaid in order to make a direct comparison between their performance in Figure 9 4. As mentioned previously, the MPPC suffers from much more sig nificant background noise than the PMT. However, it offers still offers a linear response to current and has other advantages, most notably cost and robustness. It is another viable option which should be considered especially if the background noise can b e reduced in a manner similar to the CCD based design or by using cooling.

PAGE 260

260 Figure 9 1. MPPC with fiber connected through SMA connector (top, black with tape) and firewire (bottom, blue)

PAGE 261

261 Figure 9 2 Typical MPPC response Figure 9 3 MPPC cou nts vs mAs

PAGE 262

262 Figure 9 4 PMT and MPPC response normalized to max counts

PAGE 263

263 CH APTER 10 CONCLUSIONS Results of T his Work The overall goal of this work was to develop the tools and methods necessary to measure radiation delivered during helical computed tomography (CT) exams using p lastic scintillation dosimeters (PSDs) These PSDs were used to characterize the performance and behavior of n ovel methods of dose reduction, specifically organ dose modulation (ODM) and tube current modulation (TCM) i n helical CT. This research was broken down into development of a measurement medium in Chapter 2 and the fabrication of the dosimetry system in Chapter 3 Next, utility of the PSD system was shown in applications of x ray radiosurgery, CT overranging, CT starting angle studies, and CT organ dose modulation in Chapters 3 through 6, respectively. Specifically, Chapter 3 of this work showed an example of application of PSDs to age related macular degeneration, which is currently the leading cause of severe v ision loss and blindness in those over age 65. 60 This unique application uses the PSD system to measure the delivered x ray radiosurgery dose to ensure precision and safety. A novel, independent method for meas uring overranging was described in Chapter 4. The importance of overranging in accurately predicting CT dose has recently been acknowledged, most notably in the recently released reports of American Association of Physicists in Medicine (AAPM) Task Group ( TG) 111 and AAPM TG 23. 6, 9 Additionally, AAPM has advocated a move away from CT dose index (CTDI), towards a new metric known as equilibrium dose. 6 The methodology for determining this metric is similar to the cylindrical derivation which was introduced in the starting angle dose biasing

PAGE 264

264 (SADB) methodology described in Chapter 5. The SADB needs to be incorporated into reports of CT dose in order to account for uncertainties in x ray tube starting angle. Chapter 6 describes a different application of the random x ray tube starting angle in which dose savings are shown to be poss ible and, in some cases, quite significant. These studies should be presented to vendors in order to show the dose savings possible with specification of x ray tube starting angle T he data necessary to develop a predictive tube current modulation (TCM) a lgorithm w ere collected for different scan protocols and recorded in Chapter 7 These measurements will be used for a predictive computational TCM algorithm, a novel idea which has not been successfully modeled by any other research group. Lastly, a worki ng prototype PSD system based upon CCDs was shown to be both theoretically and physically possible in Cha pter 8. PSDs have traditionally been used almost exclusively only in radiotherapy applications. At these higher megavoltage energies, higher dose deliv ery rates, and longer exposure times make a CCD based system much easier to incorporate than with diagnostic applications. Using a combination of SNR improvement techniques, t his work has provided the first successful CCD imagi ng for PSD applications at di agnostic level energies. In addition to these contributions to the current state of knowledge, t here are many promising opportunities for research featuring P SDs including system development different applications, and new methods which will be briefly discussed below. Opportunities for Future Work and Development Anthropomorphic Phantom Developments The dimensions for the male anthropomorphic phantom represented a 50th percentile adult male as defined by ICRP publication 89. 28 Chapter 2 details the

PAGE 265

265 fabrication process which can be used to produce a physical phantom from any set o f segmented CT data. These data would be drawn from the 50 adult computational phantoms which are part of the UF Advanced Laboratory for Radiation Dosimetry Studies (ALRADS) library. 41 In addition to creating an entire family of physical phantoms which represent patients of all sizes and shapes, this introduces many other exciting possibilities. For example, a project which is currently underway is the development of a series of physical add on s to the 50th percentile female anthropomorphic physical phantom which would represent a pregnant female at different stages of gestational development. These phantoms could be u sed to accurately measure fetal dose The physical phantom currently uses three different tissue equivalent substitutes. It has been made more realistic using a new material which simulates the composition of the fat layers, adipose tissue equivalent subs titute (ATES). 46 Another investigation has been recently completed which involves proper representation of breast tissue using a brea st tissue equivalent substitute (BrTES). Currently, the new B rTES and the fetal substitute are under development and are expected to be in the next generation of phantoms produced. Commercial PSD Usage for Radiation Monitoring The PSD system described in Chapter 3 was implemented in the IRay TM designed by Oraya Ther apeutics Inc. It would be interesting to check for yellowing in the PSD which is characteristic of radiation damage whenever a machine is brought in for its yearly calibrations An other proposed application of the PSD system would be as a voltage monitor. This would involve a simple setup: one measurement in air (PSD air ) while the other measurement would be after collimation material (PSD coll ) It is expected that the ratio between PSD air to PSD coll should remain constant because the only

PAGE 266

266 differen ce between the two measurement points is the filtration and the beam quality at both of these points should remain constant if the voltage is not changed. Therefore, if the ratio changed, this could be used to detect a voltage ripple and be connected to a gating event which would provide yet another safety feature. There have been plans to incorporate these measurement points by simply creating channels in the current collimator used for the x ray tube. Continuing in the idea of a voltage monitor, the ratio between the two measurements can be used as an indicator of machine degradation. Specifically, if the ratio between PSD air to PSD coll response became higher, this would suggest that the spectrum became softer One possible reason would be the tungsten mat erial of the anode melting. On the other hand, if the ratio between the PSD air to PSD coll became lower, it would suggest that the spectrum became harder, a possible reason would be the attachment of the electron cloud or plasma to the window filtering the beam. Regardless, a change in the ratio indicates a change in filtration or beam quality and should be investigated immediately. The SNR was excellent in measurements (1000 4000). I t may be more practical to utilize two separate channels using only two PM Ts. Typically two PMTs are required per channel one for signal and one for background subtraction. If this is done, the voltage monitor would not require any additional modifications or material acquisition other than another signal fiber. The voltage mo nitor was not tested during the last visit due to th e lack of an extra signal fiber. Overranging Real time, point measurements of PSDs were used to quantify the overranging in a Siemens Somatom Sensation 16. In th e methodology described in Chapter 4 two

PAGE 267

267 table dosimeters provided a measurement of average velocity over a small region whereas the console determines table velocity based on overall scan length and overall scan time. If implemented clinically, this method could potentially be adapted to operate in conjunction with the console measurements of speed, thus eliminating the need for the two dosimeters. This would also reduce the number of measurements required, but introduce a strong code pendence upon console accuracy. The purpose of o ver ranging is t o collect sufficient projection data on each side of ROI introduced in which x ray beam can be changed over the course of the scan, thus producing a trapezoidal beam profile at the beginning and end of the scan. 133 It c ould be feasible to use a PSD methodology to measure or verify the dose reduction, as well as quantify overranging. A PSD based approach would need to account for the partially open or closed collimators in real time. Also, while o verranging length results have been measured for a Siemens Somatom Sensation 16, future research should include different vendors, scanner models, and post processing methods which have already been shown to change overranging values by as much as 125%. 96 Starting Angle and Organ Dose Modulation S tarting angle dose bias (SADB) is a new metric which was introduced in Chapter 5 to represent a source of uncertainty in x ray tube starting angle which ha s been ignored. Current measurements of dose do not account for the accompanying loss of precision due to SADB. In other words, physical measurements will always have some baseline precision which cannot be overcome without hundreds of measurements in orde r to eliminate SADB through measurement of the entire range of possible doses. It would be interesting to repeat this investigation for different vendors and scanner

PAGE 268

268 models in the same manner described in Chapter 5 (head, neck, chest, abdomen, and pelvic p rotocols for all organs used in effective dose calculation). Furthermore, this would be extremely interesting to study across a wide library of phantoms. The concept of ODM was described in Chapter 6. In this study, d ose savings was calculated for two orga ns: lens of the eye and thyroid These organs represent ideal candidates d ue to their peripheral location, small size and radiobiological sensitivity Subsequent studies should be designed to evaluate ODM in more organs, and if possible, for different ven dors and scanner protocols. Also, another area of investigation would involve the impact of off isocenter positioning because these calculations assume isocentric positioning. This would also be extremely interesting to study across a wide library of phant oms. Predictive Algorithm for Tube Current Modulation Chapter 7 measured the tube current as a function of longitudinal positions for common scan protocols, as defined in Chapter 5 (head, neck, chest, abdomen, pelvis). These in scan data will provide the basis of a predictive algorithm of tube current. It is absolutely feasible and accurate to obtain organ and effective doses for CT procedures using Monte Carlo computational simulations. However, incorporating TCM into the source modeling of Monte Carlo s imulations has been challenging due to the lack of availability of data matching tube current to scan position. This experiment was designed to gather data for an algorithm which characterized longitudinal TCM. In reality, there are actually two different types of TCM in Siemens CT systems However, i t is anticipated that the other type, angular TCM will be the mo re challenging aspect to model. Therefore, the first generation of this code would involve only long itudinal TCM. Angular TCM is based upon the detector fluence in the past 180

PAGE 269

269 degrees of rotation. One method to determine this would be creating an x ray source and using a ray trace to determine the transmission for each projection After incorporating bo th types of TCM for Siemens machines, other work should be done with other vendors such as Toshiba. In fact, Toshiba uses a completely different method for TCM which is based upon data from two orthogonal projections (AP and lateral) of the patient. Ellipt ical add ons were used to transform a cylindrical phantom into phantoms which challenged both z axis and angular TCM. The phantom was made with relatively simple geometry, notably in its purposeful lack of high or low attenuation regions. As the algorithm becomes increasingly accurate for simple shapes, more and more complicated add ons can be created, culminating in a realistic anthropomorphic phantom, such as the ones described in Chapter 2. PSD CCD Prototype Dosimetry System Chapter 8 was an investigatio n into the feasibility of a CCD based PSD system. This was shown to be both theoretically and physically possible. This chapter also described the rationale and fabrication process for a prototype CCD PSD system to supplement the PMT PSD system. Essentiall y this would provide an economical massively parallel read out. The second generation prototype is currently under development will include the incorporation of a lens system using a condenser lens for improved optical coupling between the fiber and the CC D. This investigation has also demonstrated the ability of an ROI analysis code to provide an optimized selection of a proper ROI for each image or portion of an image. An improved ROI algorithm is under development to process the images in a more automate d and faster method. The next

PAGE 270

270 generation will also include the capability for multiple parallel fiber measurements, most likely by utilizing the novel target adjustment system. Multi Pixel Photon Counters MPPCs operate as an array of avalanche photodiodes (APDs) which have extremely high sensitivity for single photons. 128 The recovery time has been shown to be less than 100 ns for most APDs arrays. 129 131 In turn, most modules offer resolution down to 1 ms which is an order better than the 10 ms maximum resolution available with commercial PMTs. The background noise in the MPPC was substantially higher than that of the PMT system. The SNR of the MPPC based system was only around 2.5, while the SNR of the PMT was around 500. T he PSD fibers give around 5 10 background counts even in the PMT based system. However, 5 10 background counts was enough to trigger an event in the MPPC due to its low threshold. The MPPC must have better performance than the PMT sys tem in order to justify a switch. As mentioned previously, the MPPC suffers from much more significant background noise than the PMT. However, it offers still offers a linear response and has other advantages such as cost, low power consumption, and robust ness It is another viable option which should be considered especially if the background noise can be reduced using cooling and a housing similar to that constructed for the CCD based system. In comparison with the PMT based system, the largest advantage of MPPCs is their low price. While t he PMT offers much better SNR by a factor of more than 100 each PMT channel costs $4000. On the other hand, a $4000 MPPC module (C11206 0404FB[X], Hamamatsu, Bridgewater, NJ) has 16 channels. 132 Each channel in this 4x4

PAGE 271

271 array has a 3x3 mm 2 sensitive area comprised of 3600 pixels 132 This module controls the MPPC ( and S11827 3344MG Hamamatsu, Bridgewater, NJ) and contains the control electronics, preamp, and DAC array. Final Thoughts The research presented in this dissertation highlights the development of the tools and methods in order to measure radiation delivered during CT exams using novel designs based upon PSDs. T he average dose per scan has fallen in each of the past decades since the inception of CT technology. In order to remain viable, the benefits of an appropriately ordered CT exam should outweigh those risks. Quantification of radiation exposure from CT scans should be accurate in order to accurately assess the collective risk and to provide patient specific organ doses for use in eit her retrospective epidemiologic or prospective risk estimation studies. It is hoped that the tools and methods described in this study can be used to innovate other new methods and continue to contribute to the field of medical physics.

PAGE 272

272 APPENDIX A FIBER CONSTRUCTION METHODOLOGY This section of the dissertation details the fabrication process for fiber optic coupled (FOC) plastic scintillator dosimeters (PSD) to plastic waveguide. This waveguide is next a ttached to an SMA connector. If a different connector is used, simply stop before the step which involve s SMA epoxy. This methodology is currently the most efficient coupling available, however it is anticipated that it will be further refined. The surface preparation and coupling techniques are similar to those described by Ayotte et al. 86 Note that another version of this methodology is available with pictures as an internal laboratory document. Gather cons truction materials. In terms of fiber materials need ed mail order: : approximately 430 cm of plastic fiber transmission material (Fiber Optics Center, Eska CK 20), 3 cm of plastic scintill ator (Saint Gobain Crystals BCF heat shrink tubing m), reflective coating (Eljen Technology, EJ 510), 5 mL of SMA epoxy (FiberFin, 1656 resin, #80 hardener), an Amphenol hand puck (Ocean Optics), a glass lapping surface (Ocean Optics) and 3 grits of lapping paper (Angstrom Lap, 12, 3, and 1 m). In terms o f fiber materials available at a local store: transparent crazy glue, scissors, masking tape, electrical tape, marker, and ruler. Also required is a microscope (Carl Zeiss, Standard 20) with at least 3,2/0,07 magnification and a heat gun (Wagner). Cut 2 f ibers of slightly more than 210 cm of plastic fiber. Use the floor tiles as a rough reference, each of which is 30 cm. Cut a 3 cm of scintillator Note: when making an initial cut of the fiber, use the sharp side (not the beveled side) of the knife and mov e in a smooth, downward motion. Take one end of each fiber and prepare for optical coupling with the lapping paper. Insert the end through a 500 m SMA connector and screw it into the hand puck. Begin the lapping process with the highest grit and ending w ith lowest grit: 12, 3, 1 m. The lapping should be done by holding the fiber firmly outside the SMA connector and moving the hand puck in a figure 8 movement. This will symmetrically smooth out any jagged edges in the end of the fiber. Repeat this step wi th one side of the scintillator. This polishing aspect is one of the most

PAGE 273

273 important parts of the fabrication process. An example of the progress made is shown in Figure A 1. The surface should be at least to this degree of smoothness. Next, cut off about 3 shrink tubing which you will use as a coupler. Tape this down on the table and begin to heat in order to create a tighter coupling. To make the next few steps easier (notably the meeting of the fiber and scintillator), cut the coupler at a 45 0 angle. See Figures A 5 and A 6 Test to make sure that the 500 m fiber can fully make it through the coupler. Tape the coupler down and mark the center of coupler with a marker Place the edges of the scintillator and the fiber (th onto the black mark, and mark on the fibers where the edges of the coupler are. Super glue the scintillator and the fiber up to the edge marks and place them into the coupler. Note that the glue will set within 2 seconds, so make sure to get the components located at the center of the coupler quickly. Measure 2 mm of scintillator as measured from the black mark and make a cut. Polish the end of the scintillator with the lapping paper, starting from 12 m and ending with 5 m. Aga in, move in figure 8 motions and check the smoothness with the microscope. Dip the end of the scintillator into the reflective coating. Only a small amount is necessary, enough to cover the end of the scintillator. After waiting about 10 minutes, scrape th e edges of the coating off so only the top of the scintillator is covered. fiber. Insert both the scintillating and reference fibers into the 200 cm medium sh rink tubing until both fibers pass through the other end of the tubing. Push both fibers through the heat shrink tubing until the scintillating side is at the edge of the tubing. Both fibers should have passed through the other side of the 200 cm tubing. Continue carefully pushing the scintillator through until it is flush with the end of the tubing Tape the end of the fiber which overlaps the edge of the fiber by 6 8 mm. Roll the electrical tape and crimp the end with your finger. T he end can be sealed w ith crazy glue more stability is desired Now moving to the other side of the fiber, insert each fiber (both scintillating and reference) into the small shrink tubing.

PAGE 274

274 Tape the fibers together at the divider of the PVC heat shrink tubing. shrink tubing and place it onto each of the fibers. Dispense equal amounts of the SMA epoxy (1:1 of resin and hardener) and mix thoroughly. Put SMA epoxy into the SMA connect or and push the fibers through. Secure the fibers and allow the epoxy to set for at least one night 24 hours is best. Cut the protruding fibers to about 2 mm. Remove the previous SMA connector on the hand puck and use the new SMA connectors which have just been glued to the reference and scintill ating fibers to secure the edge of the fibers to the hand puck and polish the ends of the fiber. shrink tubing onto the ends of the SMA connectors. Begin to heat the heat shrink tubing in order to complete the light tight co nnection. In order to protect the ends of the fibers, I usually cover the ends with the rubber cap originally included with most SMA conne ctors. The safest way to store the fibers is to wrap them into a circle.

PAGE 275

275 Figure A 1 Stages of e liminatio n of jagged edges of fiber

PAGE 276

276 APPENDIX B TUBE CURRENT MODULAT ION DICOM SORTER CODE ( SORTERV1 2 .M) clear all target='Aug 17 abdcylinder HH'; % k='slow'; k='fast'; numseries=10; % ============================================================ fnames=dir(strcat(target,' \ *.ima')); [sizeL sizeW]= size(fnames); % scanparameters=zeros(numseries,10); disp(' -----------------------------------------------------------------------'); disp(' File sorter program v 1.1 '); disp(' '); disp('Creating folders..'); disp(' '); if k=='slow' for i=101:101+numseries 1 mkdir(strcat(target,' \ atcm',num2str(i))) disp(i); end disp('Folders created successfully.' ); disp(' '); elseif k=='fast' for i=1:numseries mkdir(strcat(target,' \ atcm',num2str(i))) disp(i); end disp('Folders created successfully.'); disp(' '); else disp('Error: specify reconstruction mode (slow or fast)'); break; end disp('Moving files..'); disp(' '); for i=1:sizeL a=strcat(target,' \ ',fnames(i,1).name); temp=dicominfo(a); seriesnum=temp.SeriesNumber; movefile(a, strcat(target,' \ atcm',num2str(seriesnum))); end disp('Files sorted succe ssfully'); disp(' ');

PAGE 277

277 APPENDIX C TUBE CURRENT MODULAT ION DICOM READER CODE (READOUTV1_7.M) clear all target='Aug 17 abdcylinder HH'; numseries=10; reconstructionmode='fast'; disp(' -----------------------------------------------------------------------'); disp(' DICOM readout program v 1.7 '); disp(' '); disp('Creating xls output..'); disp(' '); % MSGID=MATLAB:xlswrite:AddSheet; % warning('off', MSGID); warning off all; file_name = char(['ATCM output_',target,'.xls']); heading = cellstr(char('Series #','kVp', 'slice t', 'coll','tscan', 'Pitch', 'Ref mAs', 'Eff mAs', 'CTDIvol')); xlswrite(file_name,reshape(heading,1,9),'beam parameters',['a','1']); for i=1:numseries heading3 = cellstr(char('Series #','kVp','slice t')); xls write(file_name,reshape(heading3,3,1),strcat('Sheet',num2str(i)),['a','1']); heading2 = cellstr(char('Z pos','mA','Exposure')); xlswrite(file_name,reshape(heading2,1,3),strcat('Sheet',num2str(i)),['a','9']); end disp('Output xls file created s uccessfully'); disp(' '); beamparameters=zeros(numseries,6); disp('Processing series number..'); disp(' '); for i=1:numseries disp(i); if reconstructionmode== 'slow' a=strcat(target,' \ atcm',num2str(i+100),' \ *.ima'); elseif reconstru ctionmode== 'fast' a=strcat(target,' \ atcm',num2str(i),' \ *.ima'); else disp('Error: please specify reconstruction mode (slow or fast)') break end fnames=dir(a); [sizeL sizeW]= size(fnames);

PAGE 278

278 if sizeL==0 disp('Target directory not correctly specified'); break end info=zeros(sizeL,3); for j=1:sizeL if reconstructionmode== 'slow' b=strcat(target,' \ atcm',num2str(i+100),' \ ',fnames(j,1).name); el seif reconstructionmode== 'fast' b=strcat(target,' \ atcm',num2str(i),' \ ',fnames(j,1).name); else disp('Error: please specify reconstruction mode (slow or fast)') break end temp=dicominfo(b ); info(j,1)=temp.SliceLocation; info(j,2)=temp.XrayTubeCurrent; info(j,3)=temp.Exposure; end % heading3 = cellstr(char('Series #','kVp','slice t')); xlswrite(file_name,info(:,1),strcat('Sheet',num2str(i)),['a','10']); xlswrite(file_name,info(:,2),strcat('Sheet',num2str(i)),['b','10']); xlswrite(file_name,info(:,3),strcat('Sheet',num2str(i)),['c','10']); beamparameters(i,1)=i; beamparameters(i,2)=temp.KVP; beamparameters(i,3)=temp.SliceThickness ; end disp('Scan series written '); disp(' '); disp('Writing series parameters to file..'); disp(' '); % heading = cellstr(char('Series #','kVp','slice t', 'Pitch', 'Ref % mAs', 'Eff mAs', 'CTDIvol', 'coll')); for i=1:numseries disp (i); xlswrite(file_name,beamparameters(i,1),strcat('Sheet',num2str(i)),['b','1']); xlswrite(file_name,beamparameters(i,2),strcat('Sheet',num2str(i)),['b','2']); xlswrite(file_name,beamparameters(i,3),strcat('Sheet',num2str(i)),['b','3']); % xlswrite(file_name,beamparameters(i,4),strcat('Sheet',num2str(i)),['b','4']); % xlswrite(file_name,beamparameters(i,5),strcat('Sheet',num2str(i)),['b','5']);

PAGE 279

279 % xlswrite(file_name,beamparameters(i,6),strcat('Sheet',num2str(i)),['b','6']); % xls write(file_name,beamparameters(i,7),strcat('Sheet',num2str(i)),['b','7']); end disp('Series parameters written'); disp(' '); disp('Writing beam parameters to file..'); disp(' '); xlswrite(file_name,beamparameters(:,1),'beam parameters',['a','2']); xlswrite(file_name,beamparameters(:,2),'beam parameters',['b','2']); xlswrite(file_name,beamparameters(:,3),'beam parameters',['c','2']); % xlswrite(file_name,beamparameters(:,4),'beam parameters',['d','2']); % xlswrite(file_name,beamparameters(:,5),'beam parameters',['e','2']); % xlswrite(file_name,beamparameters(:,6),'beam parameters',['f','2']); % xlswrite(file_name,beamparameters(:,7),'beam parameters',['g','2']); disp('Beam parameters written'); disp(' '); disp('Output xls file completed successfully '); disp(' ');

PAGE 280

280 APPENDIX D AUTOMATED ROI ANALYSIS FOR CCD ACQUISITIONS CODE (ROI_V6.M) clear all fid=fopen('datfiles20110328 \ image25.dat'); B=transpose(fread(fid,[658,496],'int16')); % imagesc(B); axis image off; % **** INITIAL PARAMETERS ************* ****************************** c1=242; c2=351; i1=4; i2=4; sigmaconstraint=0.33; fwhmBOUNDS=200; fwhmROWS=10; % ******************************************************************* sigma=16383; avgsample=16383; % *************INITIAL AVERAGE VALUE DETERMINATION ****************** while sigma > avgsample*sigmaconstraint avgsample=mean2(B(c1 i1:c1+i1,c2 i2:c2+i2)); sigma=std2(B(c1 i1:c1+i1,c2 i2:c2+i2)); i1=i1+1; i2=i2+1; if i1>10 || i2> 10 disp(' WARNING: initia l peak sampling is becoming large'); disp('please consider recentering sampling location or relaxing sigma constraints'); elseif i1>20 || i2>20 disp(' ERROR: initial peak sampling is too large. '); break end end % % visualization of FWHM 45/55 percentile outline % FWHM=zeros(496,658); % for rows=1:496 % for cols=1:658 % if B(rows,cols) > avgsample*0.45 && B(rows,cols) < avgsample*0.55 % FWHM(rows,cols)=1; % end % end % end fwhmlines=zeros(fwhmROWS*2+1,fwhmBOUNDS*2+1); fwhmlines(:,:)=B(c1 10:c1+10,c2 fwhmBOUNDS:c2+fwhmBOUNDS);

PAGE 281

281 for i=1:fwhmBOUNDS*2+1 fwhmindex(i)=i; end % **************** FWHM ANALYSIS ************************************ % smoothing filter and line fit using powerful local regression using % weighted linear least squares and a 1st degree polynomial model, % additionally it assigns a zero weight to any data which are greater than % 6 sigma away. After this, the function goes through and measures the FWHM % using linear interpolation. disp(' Analyzing 21 rows around c1 defined center to determine FWHM..'); for rows=1:fwhmROWS*2+1 yy2=smooth(fwhmlines(rows,:),0.1,'rloess'); y=yy2/max(yy2); N=length(y); if y(1)<0.5 % find if below pulse center [temp,centerindex]=max(y); Pol=+1; else [temp,centerindex]=min(y); Pol= 1; disp('WARNING: Pulse Polarity = Negative') end i=2; while sign(y(i) 0.5) == sign(y(i 1) 0.5) i=i+1; end %first crossing is between v(i 1) & v(i) interp=(0.5 y(i 1))/(y(i) y(i 1)); tlead=fwhmindex(i 1) + interp*(fwhmindex(i) fwhmindex(i 1)); i=centerindex+1; %start search for next crossing at center while ((sign(y(i) 0.5) == sign(y(i 1) 0.5)) && (i <= N 1)) i=i+1; if i==N 2 disp('WARNING: No end detected'); break break end end if i~=N Ptype=1; interp=(0.5 y(i 1))/(y(i) y(i 1)); ttrail=fwhmindex(i 1)+interp*(fwhmindex(i) fwhmindex(i 1)); fwhm(rows)=ttrail tlead; else

PAGE 282

282 Ptype=2; disp('WARNING: Pulse only has one edge, FWHM IS undefined') ttrail=0 ; width=0; end disp(num2str(rows)); end disp('average FWHM is '); averagefwhm=mean(fwhm); disp(averagefwhm); if std(fwhm)/averagefwhm > .05 disp('WARNING: Coefficient of variation of FWHMs measured is greater than 5%'); elseif std(fwhm)/averagefwhm > .1 disp('ERROR: Coefficient of variation is greater than 10%. Select a different (c1,c2). '); break end % *********** ROI analysis ****************************************** % the mean and standard deviation are cal culated for varying sizes, all % centered along the c1,c2 reference point. A third dimension for the ROI % array (:,:,1 4) can be incorporated for different centers. % ROI=zeros(10,2,4); ROI=zeros(10,2); L=(averagefwhm 10)/sqrt(2); % temp2=linspace(0.5,0 .95,20); % for i=1:20 % index=floor(temp2(i)*L); % ROI(i,1)=index*index; % ROI(i,2)=mean2(B(c1 index:c1+index,c2 index:c2+index)); % ROI(i,3)=std2(B(c1 index:c1+index,c2 index:c2+index)); % end temp2=linspace(0.05,1.45,60); for i=1:60 index=floor(temp2(i)*L); ROI(i,1)=index*index; ROI(i,2)=mean2(B(c1 index:c1+index,c2 index:c2+index)); ROI(i,3)=std2(B(c1 index:c1+index,c2 index:c2+index)); end

PAGE 283

283 plot(ROI(:,1),ROI(:,2),'.'); title('Mean pixel intensity vs sample area'); xla bel('pixels sampled [pixel^2]'); ylabel('pixel intensity');

PAGE 284

284 LIST OF REFERENCES 1. NCRP, "Ionizing radiation exposure of the population of the United States, NCRP Report 160," edited by NCRP (2009). 2. A. de Gonzalez, K. Kim and J. Samet, "Radiation induced cancer risk from annual computed tomography for patients with cystic fibrosis," Am J Respir Crit Care Med 176, 970 973 (2007). 3. D. J. Brenner and E. J. Hall, "Computed tomography an increasing source of radiation exposure," The New England Journal of Medicine 357, 2277 2284 (2007). 4. C. H. McCollough, M. R. Bruesewitz and J. J. M. Kofler, "CT dose reduction and dose management tools: overview of available options," Radiographics 26, 503 512 (2006). 5. M. Kalra, M. Maher, T. Toth, L. Hamberg, M. Blake, J. Shepard and S. Saini, "Strategies for CT radiation dose optimization," Radiology 230, 619 628 (2004). 6. AAPM Task Group 111, "AAPM Report No. 111: Comprehensive methodology for the evaluation of radiation dose in x ray comp tued tomography," (American Association of P hysicists in Medicine, College Park, MD, 2010). 7. AAPM Task Group 204, "AAPM Report No. 204: Size specific dose estimates (SSDE) in pediatric and adult body CT examinations," (American Association of Physicist s in Medicine, College Park, MD, 2011). 8. International Electrotechnical Commission, "IEC 60601 2 44," (2003). 9. AAPM Diagnostic Imaging Council CT Committee Task Group #23, "AAPM Report No. 96: The measurement, reporting, and management of radiation do se in CT," (American Association of Physicists in Medicine, College Park, MD, 2008). 10. J. F. Winslow, D. E. Hyer, R. F. Fisher, C. J. Tien and D. E. Hintenlang, "Construction of anthropomorphic phantoms for use in dosimetry studies," J App Clin Med Phys 10 195 204 (2009). 11. F. H. Attix, Introduction to Radiological Physics and Radiation Dosimetry (Wiley VCH, Hoboken, NJ, 1986). 12. A. S. Beddar, T. R. Mackie and F. H. Attix, "Water equivalent plastic scintillation detectors for high energy beam dosimetry: II. Properties and measurements," Phys Med Biol 37 1901 1913 (1992).

PAGE 285

285 13. A. S. Beddar, T. R. Mackie and F. H. Attix, "Cerenkov li ght generated in optical fibers and other light pipes irradiate by electron beams," Phys Med Biol 37 925 935 (1992). 14. J. M. Edmund and C. E. Anderson, "Temperature dependence of the Al2O3:C response in medical luminescence dosimetry," Radiat Meas 42 1 77 189 (2007). 15. L. Struelens, F. Vanhavare and K. Smans, "Experimental validation of Monte Carlo calculations with a voxelized Rando Alderson phantom: A study on influence parameters," Phys Med Biol 53 5831 5844 (2008). 16. L. A. Benevides and D. E. Hi ntenlang, "Characterization of metal oxide semiconductor field effect transistor dosimeters for application in clinical mammography," Med Phys 33 514 520 (2006). 17. D. E. Hyer, R. F. Fisher and D. E. Hintenlang, "Characterization of a water equivalent fi ber optic coupled dosimeter for use in diagnostic radiology," Med Phys (2009). 18. M. A. Clift, P. N. Johnston and D. V. Webb, "A temporal method of avoiding the Cerenkov radiation generated in organic scintillator dosimeters by pulsed mega voltage electro n and photon beams," Phys Med Biol 39 253 263 (2002). 19. R. L. Dixon and A. C. Ballard, "Experimental validation of a versatile system of CT dosimetry using a conventional ion chamber: beyond CTDI100," Med Phys 34 3399 3413 (2007). 20. D. Zhang, A. A. S avandi, J. J. Demarco, C. H. Cagnon, E. Angel, A. C. Turner, D. D. Cody, D. M. Stevens, A. N. Primak, C. H. McCollough and M. F. McNitt Gray, "Variability of surface and center position radiation dose in MDCT: Monte Carlo simulations using CTDI and anthrop omorphic phantoms," Med Phys 36 1025 1038 (2009). 21. R. L. Dixon, "A new look at CT dose measurement: Beyond CTDI," Med Phys 30 1272 1280 (2003). 22. R. L. Dixon, "Restructuring CT dosimetry A realistic strategy for the future Requiem for the pencil c hamber," Med Phys 33 2006 (2006). 23. R. L. Dixon, M. T. Munley and E. Bayram, "An improved analytical model for CT dose simulation with a new look at the theory of CT dose," Med Phys 32 3712 3728 (2005).

PAGE 286

286 24. D. Zhang, A. S. Savandi, J. J. Demarco, C. H. Cagnon, E. Angel, A. C. Turner, D. D. Cody, D. M. Stevens, A. N. Primak, C. H. McCollough and M. F. McNitt Gray, "Variability of surface and center position radiation dose in MDCT: Monte Carlo simulations using CTDI and anthropomorphic phantoms," Med Phys 36 1025 1038 (2009). 25. D. Zhang, M. Zankl, J. J. DeMarco, C. H. Cagnon, E. Angel, A. C. Turner and M. F. McNitt Gray, "Reducing radiation dose to selected organs by selecting the tube start angle in MDCT helical CT scans: A Monte Carlo based study," Me d Phys 36 5654 5664 (2009). 26. C. J. Tien, J. F. Winslow and D. E. Hintenlang, "A methodology for direct quantification of overranging in helical computed tomography," Med Phys (2009). 27. M. van Straten, H. W. Venema, J. Hartman, G. J. denHeeten and C. A. Grimbergen, "Reproducibility of multi slice spiral computed tomography scans: an experimental study," Med Phys 31 2785 2786 (2004). 28. ICRP, "Basic anatomical and physiological data for use in radiological protection Reference values, Publication 89 ," Vol. 37, (The International Commission on Radiological Protection, New York, 2002). 29. L. M. Hurwitz, T. T. Yoshizumi, P. C. Goodman, D. P. Frush, G. Nguyen, G. Toncheva and C. Lowry, "Effective dose determination using an anthropomorphic phantom and metal oxide semiconductor field effect transistor technology for clinical adult body multidetector array computed tomography protocols," J Computed Assit Tomogr. 31 544 549 (2007). 30. Computerized Imaging Reference System, "ATOM Dosimetry Phantoms: A Fam ily of Dosimetry Phantoms," (Computerized Imaging Reference System, Norfolk, VA, 2008). 31. Phantom Lab, "The RANDO Phantom, RAN100, and RAN110," (Phantom Lab, Salem, NY, 2008). 32. A. K. Jones, D. E. Hintenlang and W. E. Bolch, "Tissue equivalent materi als for construction of tomographic dosimetry phantoms in pediatric radiology," Med Phys 30 2072 2081 (2003). 33. A. K. Jones, T. A. Simon, W. E. Bolch, M. M. Holman and D. E. Hintenlang, "Tomographic physical phantom of the newborn child with real time d osimetry I. Methods and techniques for construction," Med Phys 33 3274 3282 (2006). 34. D. R. White, "The formulation of tissue substitute materials using basic interaction data," Phys Med Biol 22 889 899 (1977).

PAGE 287

287 35. D. R. White, R. J. Martin and R. Darl ison, "Epoxy resin based tissue substitutes," Br J Radiol 50 814 821 (1977). 36. C. Lee, D. Lodwick, J. Hurtado, D. Pafundi and W. E. Bolch, "Development of a series of hybrid computational phantoms and their applications to assessment of photon and elect ron specific absorbed fractions," (2008 Annual Meeting of the Europena Association of Nuclear Medicine, Munich, Germany, 2008). 37. J. F. Winslow, Construction and application of anthropomorphic phantoms for use in CT dose studies ." (Doctoral dissertatio n). University of Florida, 2009. 38. C. Lee, D. Lodwick, J. L. Williams and W. E. Bolch, "Hybrid computational phantoms of the male and female newborn patient: NURBS based whole body models," Phys Med Biol 52 3309 3333 (2007). 39. C. Lee, D. Lodwick, J. L Williams and W. E. Bolch, "Hybrid computational phantoms of the 15 year male and female adolescent: applications to CT organ dosimetry for patients of variable morphometry," Med Phys 35 2366 2382 (2008). 40. C. Lee, C. Lee, S. H. Park and J. K. Lee, "Development of the two Korean adult tomographic computational phantoms for organ dosimetry," Med Phys 33 380 390 (2006). 41. C. Lee, K. P. Kim, D. Long, R. F. Fisher, C. J. Tien, S. L. Simon, A. Bouville and W. E Bolch, "Organ doses for reference adult male and female undergoing computed tomography estimated by Monte Carlo simulations," Med Phys 38 1196 1206 (2011). 42. M. Cristy and K. F. Eckerman, "Specific absorbed fractions of energy at various ages from int ernal photon sources," (1987). 43. International Commission on Radiation Units and Measurements, "Tissue substitutes in radiation dosimetry and measurement," (International Commission on Radiation Units and Measurements, Bethesda, MD, 1989). 44. ICRP, "R ecommendations of the International Commission on Radiological Protection, Publication 103," Vol. 37, (The International Commission on Radiological Protection, New York, 2007). 45. J. T. Bushberg, J. A. Seibert, E. M. L. Jr and J. M. Boone, The Essentials Physics of Medical Imaging (Lippincott Williams and Wilkins, Philadelphia, 2002). 46. R. F. Fisher, Dose assessment and prediction in tube current modulated computed tomography ." (Doctoral dissertation). University of Florida, 2010.

PAGE 288

288 47. H. Churei, K. Oh kubo, M. Makajo, H. Hokotate, Y. Baba, J. Ideue, K. Miyagawa, H. Nakayama, Y. Hiraki, T. Kitasato and N. Yabe, "External beam radiation therapy for age related macular degeneration: two years' follow up results at a total dose of 20 Gy in 10 fractions," Ra diat Med 22 398 404 (2004). 48. R. P. Murphy, "Age related macular degeneration," Ophthamology 93 969 971 (1986). 49. A. Capone and Macugen AMD Study Group, "Intravitreous pegaptanib sodium (Macugen) in patients with age related macular degeneration (AMD ): Safety and pharamacokinetics," Invest Ophthalmol Visual Sci Annual Meeting Abstracts 46 2362 (2005). 50. Macular Photocoagulation Study Group, "Visual outcome after laser photocoagulation for subfoveal choroidal neovascularization secondary to age rela ted macular degeneration. The influence of initial lesion size and initial visual activity," Arch Ophthamol 112 480 488 (1994). 51. P. J. Rosenfeld, D. M. Brown, J. S. Heier, D. S. Boyer, P. K. Kaiser, C. Y. Chung and R. Y. Kim, "Ranibizumab for neovascul ar age related macular degeneration," N Engl J Med 355 (2006). 52. VPT Study Group, "Photodynamic therapy of subfoveal choroidal neovascularization in pathologic myopia with vertepofin. 1 year results of a randomized clinical trial VIP report no. 1," Oph thamology 108 841 852 (2001). 53. D. M. Moshfeghi, P. K. Kaiser and M. Gertner, "Stereotactic low voltage x ray irradiation for age related macular degeneration," Br J Ophthalmol 95 185 188 (2010). 54. NCRP, "Biological effects and exposure limits for ho t particles," NCRP Report No. 130 (1999). 55. M. P. Avila, M. E. Farah, A. Santos, Z. Kapran, J. P. Duprat, B. W. Woodward and J. Nau, "Twelve month short term safety and visual acuity results from a multicentre pospective study of epiretinal strontium 90 brachytherapy with bevacizumab for the treatment of subfoveal choroidal neovascularisation secondary to age related macular degeneration," Br J Ophthamol 93 305 309 (2009). 56. RAD Study Group, "A prospective, randomized, double masked trial on radiation therapy for neovascular age related macular degeneration (RAD study). Radiation therapy for age related macular degeneration," Ophthamology 106 2239 2247 (1999).

PAGE 289

289 57. H. J. Zambarakji, A. M. Lane, E. Ezra, D. Gautheir, M. Goitein, J. A. Adams, J. E. Munzen rider, J. W. Miller and E. S. Gragoudas, "Proton beam irradiation for neovascular age related macular degeneration," Opthamology 113 2012 2019 (2006). 58. D. M. Marcus, E. Peskin, M. Maguire, D. Weissgold, J. Alexander, S. Fine and D. Followill, "The age related macular degeneration radiotherapy trial (AMDRT): One year results from a pilot study," Am J. Ophthalmology 106 2239 2247 (2004). 59. J. Hanlon, C. Lee, E. Chell, M. Gertner, S. Hansen, R. W. Howell and W. E. Bolch, "Kilovoltage stereotactic radios urgery for age related macular degeneration: Assessment of optic nerve dose and patient effective dose," Med Phys 36 3671 3681 (2009). 60. J. Hanlon, M. Firpo, E. Chell, D. M. Moshfeghi and W. E. Bolch, "Stereotactic radiosurgery for AMD: A Monte Carlo ba sed assessment of patient specific tissue doses," Invest Ophthalmol Vis Sci in press (2011). 61. C. Lee, E. Chell, M. Gertner, S. Hansen, R. W. Howell, J. Hanlon and W. E. Bolch, "Dosimetry characterization of a multibeam radiotherapy treatment for age rel ated macular degeneration," Med Phys 35 5151 5160 (2008). 62. J. Hanlon, C. Lee, E. Chell, M. Gertner, S. Hansen, R. Howell and W. Bolch, "Kilovoltage stereotactic radiosurgery for age related macular degeneration: Assessment of optic nerve dose and patie nt effective dose," Med Phys 36 3671 3681 (2009). 63. C. Lee, E. Chell, M. Gertner, S. Hansen, R. Howell, J. Hanlon and W. Bolch, "Dosimetry characterization of a multibeam radiotherapy treatment for age related macular degeneration," Med Phys 35 5151 51 60 (2008). 64. S. Senan and E. F. Smit, "Design of clinical trials of radiation combined with antiangiogenic therapy," Oncologist 12 465 477 (2007). 65. L. Archambault, J. Arsenault, J. Gingras, A. S. Beddar and L. Beaulieu, "Plastic scintillation dosimet ry: Optimal selection of scintillating fibers and scintillators," Med Phys 32 2271 2278 (2005). 66. L. Archambault, A. S. Beddar, L. Gingras, R. Roy and L. Beaulieu, "Measurement accuracy and Cerenkov removal for high performance, high spatial resolution scintillation dosimetry," Med Phys 33 128 135 (2006). 67. A. S. Beddar, S. Law, N. Suchowerska and T. R. Mackie, "Plastic scintillation dosimetry: Optimization of light collection efficiency," Phys Med Biol 48 1141 1152 (2003).

PAGE 290

290 68. A. S. Beddar, T. R. Mackie and F. H. Attix, "Water equivalent plastic scintillation detectors for high energy beam dosimetry: I. Physical characteristics and theoretical consideration," Phys Med Biol 37 1883 1900 (1992). 69. M. P. Petric, J. L. Robar and B. G. Clark, "Development and characterization of a tissue equivalent plastic scintillator based dosiemtry system," Med Phys 33 96 105 (2006). 70. A. S. Beddar, "Water equivalent plastic scintillation detectors in radiation therapy," Radiat Prot Dosimetry 120 1 6 (2006). 71. J. F. Williamson, J. F. Dempsey, A. S. Kirov, J. I. Monroe, W. R. Binns and H. Hedtjarn, "Plastic scintillator response to low energy photons," Phys Med Biol 44 857 871 (1999). 72. A. S. Beddar, D. J. Mason and P. F. O'Brien, "Absorb ed dose perturbation caused by diodes for small field photon dosimetry," Med Phys 21 1075 1079 (1994). 73. R. H. S. Carpter, Movements of the Eyes (Pion, London, 1988). 74. S. F. de Boer, A. S. Beddar and J. A. Rawlinson, "Optical filtering and spectral measurement of radiation induced light in plastic scintillator dosimetry," Phys Med Biol 45 (1993). 75. F. Ponisch, L. Archambault, T. M. Briere, N. Sahoo, R. Mohan, A. S. Beddar and M. T. Gillin, "Liquid scintillator for 2D dosimetry for high energy photo n beams," Med Phys 36 1478 1485 (2009). 76. J. Lambert, Y. Yin, D. R. McKenzie, S. Law and N. Suchowerska, "Cerenkov free scintillation dosimetry in external beam radiotherapy with an air core light guide," Phys Med Biol 53 3071 3080 (2008). 77. Saint Go bain Ceramics and Plastics Inc, "Scintillation products: Scintillating optical fibers," (Saint Gobain Crystals, Inc, Hiram, OH, 2008). 78. Mitsubishi Rayon Co. Ltd: Eksa Optical Fiber Division, "Specification Sheet: CK 20 (Mitsubishi Rayon Co., Ltd., Tokyo, Japan, 2001). 79. J. Elsey, D. R. McKenzie, J. Lambert, N. Suchowerska, S. L. Law and S. C. Fleming, "Optimal coupling of light from a cylindrical scintillator into an optical fiber," Appl Opt 46 397 404 (2007). 80. Hamamatsu Corporation, "Specific ation Sheet: E5776 51," (Hamamatsu Corporation, Bridgewater, NJ, 1994).

PAGE 291

291 81. Source Ray Inc., "Specification Sheet: SR 115," (Source Ray Inc., Bohemia, NY, 2005). 82. Florida Department of Health, "Chapter 64E 5, Florida Administrative Code, X Ray in the healing arts," edited by F. D. o. Health (Tallahassee, FL, 2007). 83. D. E. Hyer, C. F. Serago, S. Kim, J. G. Li and D. E. Hintenlang, "An organ and effective dose study of XVI and OBI cone beam CT systems," J App Clin Med Phys 11 181 197 (2010). 84. D. B. Pelowitz, "MCNPX user's manual version 2.5.0," in Los Alamos National Laboratory Report No. LA CP 05 0369, (2005). 85. K. Cranley, B. J. Gilmore, G. W. A. Fogarty and L. Desponds, "Catalogue of diagnostic x ray spectra and other data," in Institute of Physics Report No. 78, (1997). 86. G. Ayotte, L. Archambault, L. Gingras, F. Lacroix, A. S. Beddar and L. Beaulieu, "Surface preparation and coupling in plastic scintillator dosimetry," Med Phys 33 3519 3525 (2006). 87. D. B. Archer, W. M. Amoaku and T. A. Gardiner, "Radiation retinopathy clinical, histopathological, ultrastructural, and experimental correlations," Eye (Lond) 5 239 251 (1991). 88. J. T. Parsons, F. J. Bova, C. R. Fitzgerald, W. M. Mendenhall and R. R. Million, "Radiation retionpathy af ter extrenal beam irradiation: analysis of time dose factors," Int J Radiat Oncol Biol Phys 30 765 773 (1994). 89. J. R. Evans, V. Sivagnanavel and V. Chong, "Radiotherapy for neovascular age realted macular degeneration," Cochrane Database Syst Rev 2010, CD004004 (2010). 90. E. J. Hall and A. J. Giaccia, Radiobiology for the radiologist Sixth edition ed. (Lippincott Williams and Wilkins, Philadelphia, PA, 2006). 91. Hamamatsu Technical Information, "Photon counting: using photomultiplier tubes," (1998). 92. L. Archambault, T. M. Briere, F. Ponisch, L. Beaulieu, D. A. Kuban, A. Lee and A. S. Beddar, "Towards a real time in vivo dosimetry system using plastic scintillation detectors," Int J Radiat Oncol Biol Phys 78 280 287 (2010).

PAGE 292

292 93. A. Tzedakis, J. Dam ilakis, A. Karantanas, S. Karabekios and N. Gourtsoyiannis, "Influence of z overscanning on normalized effective doses calculated for pediatric patients undergoing multidetector CT examinations," Med Phys 34 1163 1175 (2007). 94. A. Tzedakis, J. Damilakis K. Persinakis, J. Stratakis and N. Gourtsoyiannis, "The effect of z overscanning on patient effective dose from multidetector helical computed tomography examinations," Med Phys 32 1621 1629 (2005). 95. A. Tzedakis, K. Persinakis, M. Raissaki and J. Dam ilakis, "The effect of z overscanning on radiation burdens of pediatric patients undergoing head CT with multidetector scanners: A Monte Carlo study," Med Phys 33 2472 2478 (2006). 96. A. J. van der Molen and J. Geleinjins, "Overranging in multisection CT: Quantification of relative contribution to dose comparison of four 16 section CT scanners," Radiology 242 (2007). 97. P. D. Deak, O. Langner, D. Ing, M. Lell and W. A. Kalendar, "Effects of adapti ve section collimation on patient radiation dose in multisection spiral CT," Radiology 252 104 107 (2009). 98. E. Seeram, Computed Tomography: Physical Principles, Clinical Applications, and Quality Control 2nd ed. (Saunders, 2001). 99. R. L. Dixon and A C. Ballard, "Experimental validation of a versatile system of CT dosimetry using a conventional ion chamber: Beyond CTDI100," Med Phys 34 3399 3413 (2007). 100. American Association of Physicists in Medicine, "The measurement, reporting, and management of radiation dose in CT: report of AAPM Task Group 23 of the Diagnostic Imaging Council CT Committee," (College Park, MD: AAPM: Report 96, 2007). 101. D. Cody and M. McNitt Gray, "Optimizing CT dose and image quality for different patient sizes," Med Phys 36 2686 (2009). 102. M. F. McNitt Gray, "AAPM/RSNA physics tutorial for residents: topics in ct: radiation dose in CT," Radiographics 22 1541 1553 (2002). 103. L. Romans, Computed tomography for technologists: exam review (Lippincott Williams and WIlki ns, Baltimore, MD, 2010). 104. C. J. Tien and D. E. Hintenlang, "Poster: Physical calculation of helical computed tomography starting angle dose biasing," Med Phys 37 3351 (2010).

PAGE 293

293 105. C. J. Tien and D. E. Hintenlang, "Physical calculation of helical comp uted tomography starting angle dose biasing," Med Phys (2011). 106. R. J. Staton, C. Lee, M. D. Williams, D. E. Hintenlang, M. M. Arreola, J. L. Williams and W. E. Bolch, "Organ and effective doses in newborn patients during helical multislice computed tom ography examination," Phys Med Biol 51 5151 5166 (2006). 107. R. J. Staton, C. Lee, M. D. Williams, D. E. Hintenlang, M. M. Arreola, J. L. Williams and W. E. Bolch, "Organ and effective doses in newborn patients during helical multislice computed tomograp hy examinations," Phys Med Biol 51 5151 5166 (2006). 108. F. Martini and W. C. Ober, Fundamentals of Anatomy and Physiology 5 ed. (Prentice Hall, Upper Saddle River, 2001). 109. J. N. Althen, "Automatic tube current modulation in CT a comparison betwee n different solutions," Radiation Protection Dosimetry 114 308 312 (2005). 110. AAPM Diagnostic Imaging Council CT Committee Task Group #23, "AAPM Report No. 96: The measurement, reporting, and management of radiation dose in CT," in Med Phys, (American Association of Physicists in Medicine, College Park, MD, 2008). 111. ICRP, "Managing patient dose in computed tomography, Publication 87," (Oxford, Pergamon, 2000). 112. P. Deak, M. vanStraten, P. C. Shrimpton, M. Zankl and W. A. Kalendar, "Validation of a Monte Carlo tool for patient specific dose simulations in multi slice computed tomography," Eur Rad 18 759 772 (2008). 113. R. F. Fisher, Tissue equivalent phantoms for evaluating in plane tube current modulated CT dose and image quality ." (Master's th esis). University of Florida, 2006. 114. Siemens Medical, "SOMATOM Sensation 10/Sensation 16 Application Guide, Software Version B10," (Siemens AG Medical Solutions, Germany, 2006). 115. D. M. Klein, F. Therriault Peroulx, L. Archambault, T. M. Briere, L. Beaulieu and A. S. Beddar, "Determinig regions of interest for CCD camera based fiber optic luminescence dosimetry by examining signal to noise ratio," Med Phys 38 1374 1377 (2011).

PAGE 294

294 116. R. Cirio, E. Garelli, R. Schulte, S. Amerio, A. Boriano, F. Bourha leb, G. Coutrakon, M. Donetti, S. Giordanengo, P. Koss, E. Madon, F. Marchetto, U. Nastasi, C. Peroni, D. Santuari, A. Sardo, G. Scielzo, M. Stasi and E. Trevisiol, "Two dimensional and quasi three dimensional dosimetry of hadron and photon beams with the Magic Cube and the Pixel Ionization Chamber," Phys Med Biol 49 3713 3724 (2004). 117. F. Lacroix, L. Beaulieu, L. Archambault and A. S. Beddar, "Simulation of the precision limits of plastic scintillation detectors using optimal component selection," Med Phys 37 412 418 (2010). 118. F. Lacroix, A. S. Beddar, M. Guillot, L. Beaulieu and L. Gingras, "A design methodology using signal to noise ratio for plastic scintillation detectors design and performance optimization," Med Phys 36 5214 5220 (2009). 119. F. Lacroix, L. Archambault, L. Gingras, M. Guillot, A. S. Beddar and L. Beaulieu, "Clinical prototype of a plastic water equivalent scintillating fiber dosimeter array for QA applications," Med Phys 35 3682 3690 (2008). 120. A. S. Beddar, T. J. Kinsella, A. Ikhlef and C. H. Sibata, "A Miniature "Scintillator Fiberoptic PMT" Detector System for the Dosimetry of Small Fields in Stereotactic Radiosurgery," IEEE Trans Nucl Sci 48 924 928 (2001). 121. J. W. Heemskerk, A. H. Westra P. M. Linotte, K. M. Ligtvoet, W. Zbijewski and F. J. Beekman, "Front illuminated versus back illuminated photon counting CCD based gamma camera: important consequences for spatial resolution and energy resolution," Phys Med Biol 52 N149 N162 (2007). 12 2. T. Yu and J. M. Boone, "Lens coupling efficiency: Derivation and application under differing geometrical assumptions," Med Phys 24 565 570 (1997). 123. W. J. Smith, Modern Lens Design 2nd ed. (McGraw Hill, New York, 2004). 124. S. F. Ray, Applied photographic optics 3rd ed. (Focal Press, Boston, MA, 2002). 125. Polymicro Technologies, "Specification Sheet: FVP400440480," (Phoenix, AZ, 2008). 126. L. Archambault, T. M. Briere and A. S. Beddar, "Transient noise characterization and filtration in CC D cameras exposed to stray radiation from a medical linear acceleration," Med Phys 35 4342 4351 (2008). 127. Thorlabs, "Thorlabs V20 Catalog," Vol. 20, (Thorlabs, Newton, NJ, 2010), pp. 639.

PAGE 295

295 128. Hamamatsu Corporation, "Specification Sheet: C10507 11 025 U," (Hamamatsu Corporation, Bridgewater, NJ, 2009). 129. M. Yokoyama, T. Nobuhara, M. Taguchi, T. Nakaya, T. Murakami, T. Nakadair, K. Yoshimura, K. Kawagoe, Y. Tamura, T. Lijima, Y. Mazuka, K. Miyabayashi, S. Iba, H. Miyata and T. Takeshita, presented at the SNIC Symposium, Palo Alto, CA, USA, 2006 (unpublished). 130. K. Yamamoto, K. Yamamura, K. Sato, T. Ota, H. Suzuki and S. Ohsuka, presented at the IEEE Nuclear Science Symposium San Diego, CA, 2006 (unpublished). 131. M. Taguchi. Development of multi pixel photon counters and readout electronics ." (Master's thesis). Kyoto University, 2007. 132. Hamamatsu Corporation, "Specification Sheet: C11206 0404FB(X) 16ch(44) MPPC array module (Hamamatsu Corporation, Bridgewater, NJ, 2010). 133. J. A. Christn er, V. A. Zavaletta, C. D. Eusemann, A. I. Walz Flannigan and C. H. McCollough, "Dose reduction in helical CT: Dynamically adjustable z axis x ray beam collimation," Am J Roentgenol 194 49 55 (2010).

PAGE 296

296 BIOGRAPHICAL SKETCH Christopher Jason Tie n was born in Cincinnati, Ohio, to Chieh Sheng and I Ching Tien. He is one of two children, along with his younger sister Deborah. He graduated from Troy Athens High School in Troy, Michigan in 2003 He then graduated S umma cum L aude in 2007 from the Univ ersity of Michigan in Ann Arbor with a Bachelor of Science in Engineering degree in nuclear engineering and radiological sciences, with a minor in mathematics. He continued his studies at the University of Michigan and graduated in 2008 with a Master of Sc ience in Engineering degree in nuclear engineering and radiological sciences. Chris then continued his medical physics education at the University of Florida, where he graduated in 2011 with a PhD in the newly established biomedical engineering program wit h a specialization in medical physics He beg a n his medical physics reside ncy at Brown University in 2011.