<%BANNER%>

Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2008-02-29.

Permanent Link: http://ufdc.ufl.edu/UFE0021088/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2008-02-29.
Physical Description: Book
Language: english
Creator: Shedlock, Daniel
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: Nuclear and Radiological Engineering -- Dissertations, Academic -- UF
Genre: Nuclear Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Daniel Shedlock.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Dugan, Edward T.
Electronic Access: INACCESSIBLE UNTIL 2008-02-29

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021088:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021088/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2008-02-29.
Physical Description: Book
Language: english
Creator: Shedlock, Daniel
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: Nuclear and Radiological Engineering -- Dissertations, Academic -- UF
Genre: Nuclear Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Daniel Shedlock.
Thesis: Thesis (Ph.D.)--University of Florida, 2007.
Local: Adviser: Dugan, Edward T.
Electronic Access: INACCESSIBLE UNTIL 2008-02-29

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021088:00001


This item has the following downloads:


Full Text

PAGE 1

1 X-RAY BACKSCATTER IMAGING FOR RADIOGRAPHY BY SELECTIVE DETECTION AND SNAPSHOT: EVOLUTION, DEVELOPMENT, AND OPTIMIZATION By DANIEL SHEDLOCK A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

2 2007 Daniel Shedlock

PAGE 3

3 For my daughter, Alyssa, and her future

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank God, for the gifts he has gi ven me that allow me to walk this path and complete this journey. I am truly thankful for a ll the gifts He has given me I need to thank my family, Missy and my daughter, Alyssa, for thei r patience, support and sa crifices they have made. I also need to thank my most active advisors Dr. Edward Dugan and Dr. Alan Jacobs for the wisdom of their guidance, and encourag ement along the way. They set the stage and environment for success. I need to thank Dan Ekda hl for his time and excellent work that have contributed to the success of many projects. Thank you, Warren Ussery for the financial funding to push the research far beyond the laboratory and pr ototyping stage. I am thankful for all my colleagues in the research group, th e day-to-day interactions and c onstant exchange of ideas was invaluable. This work has been made possible by a collection of individuals who put forth a tremendous effort in laying the foundations years be fore I even walked onto the project. Thank you everyone who has helped contribute to my gr owth, learning and success each step of the way. To be successful and succeed, one has to be surrounded by great people, and I have worked with some of the very best. Financial acknowledgment: Lockheed Martin Space Systems Co. NASA, Langley Research Center NASA, Marshall Space Flight Center University of Florida, Department of Nuclear and Radiol ogical Engineering

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........8 LIST OF FIGURES................................................................................................................ .........9 ABSTRACT....................................................................................................................... ............12 CHAPTER 1 INTRODUCTION..................................................................................................................14 Background..................................................................................................................... ........14 Pencil Beam Compton S catter X-ray Imaging.......................................................................14 Highly Collimated Techniques........................................................................................14 Uncollimated Backscatter Radiography..........................................................................16 Radiography by Selective Detection...............................................................................17 Lateral Migration Radiography.......................................................................................18 RSD versus LMR............................................................................................................19 Full-Field Illumination Comp ton Backscatter Imaging..........................................................20 Coded Aperture Imaging.................................................................................................20 Optical X-ray Focusing...................................................................................................21 2 SCANNING SYSTEM OVERVIEW.....................................................................................23 Lockheed Martin Prototype System.......................................................................................23 First Industrial Scanning Systems..........................................................................................24 Original Industrial Syst em Component Details......................................................................25 3 ARTIFACT AND CONTRAST GENERATI ON IN SCATTER X-RAY IMAGING..........27 Contrast and Artifacts......................................................................................................... ....27 Compton Backscatter Imag ing Characteristics.......................................................................27 Pencil Beam, Single-Scatter Compton Backscatter Contrast..........................................28 Void Contrast and Shadowing Effects............................................................................30 Scattering and Absorption Contrast, and Shadowing Effects.........................................31 Measurement Results and Discussion....................................................................................37 Application of Contrast Mechanisms.....................................................................................40 Conclusions about Artifact and Contrast Generation.............................................................41

PAGE 6

6 4 OPTIMIZATION AND ANALYS IS OF RSD SCANNING SYSTEM COMPONENTS....42 Detector Testing............................................................................................................... .......42 Comparison of NaI and plas tic (BC404) scintillation.....................................................42 Copper-doped Quartz......................................................................................................43 PIN-Diode Detectors.......................................................................................................44 Universal charge preamp..........................................................................................44 Pin-diode test results................................................................................................45 YSO Detectors.................................................................................................................46 Image Results Comparison for YSO Detectors...............................................................48 RSD Scanning System YSO Det ector and Preamp Upgrades.........................................50 Illumination Beam Aperture Geometry..................................................................................50 Round and Square Aperture Image Analysis..........................................................................52 Detector Modes of Operation.................................................................................................55 Count Mode.....................................................................................................................56 Current Mode...................................................................................................................56 Current Mode versus Counting Mode Detectors Measurements....................................58 Monte Carlo Simulation of Current versus Count Mode for SOFI.................................60 Current mode versus count mode results.................................................................61 NaI performance.......................................................................................................62 5 RSD COMPACT SYSTEM PROTOTYPE...........................................................................65 X-ray Tube Technology..........................................................................................................65 RSD Scanning System Compact Prototype............................................................................66 Compact RSD Scanning System Features..............................................................................67 Rectangular versus Round Shaped YSO Crystals..................................................................68 Illumination Beam Evaluation................................................................................................69 Beam Intensity.................................................................................................................70 Beam Dispersion.............................................................................................................72 6 SNAPSHOT BACKSCA TTER RADIOGRAPHY................................................................77 Image Technology Introduction.............................................................................................77 Snapshot Backscatter Radiography (SBR).............................................................................77 Shadow Aperture Backscatter Radiography (SABR).............................................................79 SABR Nylon Substrate Measurements...........................................................................80 SABR Nylon Substrate Discussion.................................................................................84 SABR Aluminum Substrate Measurements....................................................................85 SABR Aluminum Substrate Discussion..........................................................................87 SBR Radiography Lessons Learned and Failed Attempts......................................................88

PAGE 7

7 7 SUMMARY, CONCLUSIONS AND FUTURE WORK...................................................... 91 Summary and Conclusions.....................................................................................................91 Future Work.................................................................................................................... ........92 LIST OF REFERENCES............................................................................................................. ..95 BIOGRAPHICAL SKETCH.........................................................................................................98

PAGE 8

8 LIST OF TABLES Table page 3-1 Relative grey scale contrast table fo r various RSD imaging modalities based on feature type, and relative location......................................................................................36 4-1 MCNP results comparing relative contra st for count and current mode detector operation...................................................................................................................... ......62 5-1 Effect of illumination beam tube length on x-ray beam intensity.....................................71 5-2 Calculated versus measured illumination spot sizes for different length illumination beam tubes and focal spot sizes.........................................................................................74

PAGE 9

9 LIST OF FIGURES Figure page 1-1 Highly collimated x-ray back scatter imaging, voxel by voxel..........................................15 1-2 Uncollimated pencil beam CBI technique.........................................................................17 1-3 RSD collimated and uncollimated detector with collimation plane..................................18 1-4 LMR collimated regime and uncollimated detector..........................................................19 1-5 Example of a coded aperture using a Modified Uniformly Redundant Array...................21 1-6 Conceptual drawing of lobs ter eye focusing parallel photons...........................................22 1-7 Emerging technology from POC for lobste r eye hand held x-ray imaging device............22 2-1 Prototype RSD scanning device built for Lockheed Martin Space Systems Co...............23 2-2 First commercial RSD sca nning system configuration......................................................25 3-1 Subsurface features above collimation plane.....................................................................33 3-2 Increasing detection solid angle from points along illumination beam.............................34 3-3 Subsurface feature be low collimation plane......................................................................35 3-4 Aluminum sample plate with 10 mm wide, 2 mm high, channels.....................................37 3-5 RSD image results of the aluminum plate.........................................................................39 3-6 Composite material with a void region in the gap filler....................................................40 3-7 RSD scanned image of a composite material with a void region in the gap filler.............41 4-1 RSD image of an aluminum plate with 5 holes.................................................................46 4-2 YSO detector evolution..................................................................................................... .47 4-3 Space shuttle external tank flange bo lting area with stringer and SOFI defect.................48 4-4 YSO, 2.54 cm diameter, detector im age for external tank flange bolt..............................49 4-5 NaI, 5.08 cm diameter, detector im age for external tank flange bolt................................49 4-6 YSO detector with or bit holder and upgrades....................................................................51 4-7 SXI RSD scanning system with YSO upgrade..................................................................52

PAGE 10

10 4-8 Illumination beam spot size analysis.................................................................................52 4-9 Scanning configuration for ramp panel..............................................................................53 4-10 RSD image, 2 mm round aperture ramp panel image ......................................................54 4-11 RSD image, 2 mm roun d aperture ramp panel..................................................................55 4-12 Energy weighted current mode contrast for a single pixel for varying spectral curves.....58 4-13 RSD image of SOFI calibration block...............................................................................59 4-14 MCNP model of SOFI with aluminum substrate...............................................................60 4-15 Average photon energy at detector entrance window........................................................62 4-16 Normalize backscatter spectrum at the detector entrance window c.................................63 4-17 Scattering to absorption ratio as a function of energy for NaI...........................................64 5-1 YXLON.TU 100-D02 x-ray tube for test ing of compact system components..................65 5-2 Prototype compact RSD scanning system.........................................................................66 5-3 Prototype compact RSD s canning system, bottom view...................................................67 5-4 Three-dimensional rendering of the rectangular YSO crys tal holder and PMT................68 5-5 Compact system RSD scan of TPS tile with corrosion......................................................69 5-6 Illumination beam tubes for the compact and Lockheed RSD scanning systems............70 5-7 Illumination spot size on film 6.0 cm from the aperture...................................................72 5-8 Geometric configuratio n to calculate illuminatio n beam spot divergence.......................73 5-9 Horizontal line profile of illumination beam spot size.....................................................75 5-10 Condensed horizontal line profile of illumination beam spot size...................................76 6-1 Snapshot backscatter radiography setup............................................................................78 6-2 Unprocessed snapshot backscatter image..........................................................................78 6-3 Shadow aperture backsca tter radiography illustration.......................................................79 6-4 Shadow aperture examples...............................................................................................80 6-5 Collection of washers a nd lead on a nylon substrate........................................................81

PAGE 11

11 6-6 SABR image of nylon target using 2.54 cm square shadow aperture...............................82 6-7 SABR image of nylon target using va rious dimension shadow apertures.........................82 6-8 Line profile of column 767................................................................................................83 6-9 Line profile of row 917.................................................................................................... ..83 6-10 Collection of washers and l ead on an aluminum substrate...............................................85 6-11 SABR image of FOD on aluminum substrate target.........................................................86 6-12 Line profile of column 1261..............................................................................................86 6-13 Line profile of row 572................................................................................................... ...87 6-14 SBR mask patterns........................................................................................................ ....89 6-15 SBR exposures taken at 70 kVp........................................................................................89 6-16 SABR exposure pattern fo r round illumination apertures................................................90 6-17 SABR exposure pattern fo r line illumination apertures....................................................90

PAGE 12

12 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy X-RAY BACKSCATTER IMAGING FOR RADIOGRAPHY BY SELECTIVE DETECTION AND SNAPSHOT: EVOLUTION, DEVELOPMENT, AND OPTIMIZATION By Daniel Shedlock August 2007 Chair: Edward T. Dugan Major: Nuclear Engineering Sciences Compton backscatter imaging (CBI) is a sing le-sided imaging technique that uses the penetrating power of radiation and unique interaction propertie s of radiation with matter to image subsurface features. CBI has a variety of applications that include non-destructive interrogation, medical imaging, security and mi litary applications. Radiography by selective detection (RSD), lateral migration radiogra phy (LMR) and shadow aperture backscatter radiography (SABR) are different CBI techniques that are being optimized and developed. Radiography by selective detec tion (RSD) is a pencil beam Compton backscatter imaging technique that falls between highly collimated and uncollimated techniques. Radiography by selective detection uses a combination of singl eand multiple-scatter phot ons from a projected area below a collimation plane to generate an imag e. As a result, the image has a combination of firstand multiple-scatter components. RSD tec hniques offer greater subsurface resolution than uncollimated techniques, at speeds at least an or der of magnitude faster than highly collimated techniques. RSD scanning systems have evolved fr om a prototype into near market ready scanning devices for use in a variety of single-sided im aging applications. The design has changed to

PAGE 13

13 incorporate state-of-the-a rt detectors and electronics optimized for backscatter imaging with an emphasis on versatility, efficiency and speed. Th e RSD system has become more stable, about 4 times faster, and 60 % lighter while maintaini ng or improving image quality and contrast over the past 3 years. A new snapshot backscatter radiography (S BR) CBI technique, shadow aperture backscatter radiography (SABR) has been developed from concept and proof of-principle to a functional laboratory prototype. SABR radiography uses digital detection media and shaded aperture configurations to generate near-sur face Compton backscatter images without scanning, similar to how transmissi on radiographs are taken. Finally, a more inclusive theory of the factor s affecting CBI contrast generation has tied together the past work of LMR with the more recen t research in RSD. A variety of factors that induce changes in the backscatter photon field intensity (resulting in contrast changes in images) include: changes in the electr on density field, attenuation cha nges along the entrance and exit paths, changes in the relative geometric positioning of the target, feature, illumination beam, and detectors. Understanding the in terplay of how changes in each of these factors affects image contrast becomes essential to utilizing a nd optimizing RSD for different applications.

PAGE 14

14 CHAPTER 1 INTRODUCTION Background Compton backscatter imaging (CBI) is a sing le-sided imaging technique in which the radiation source and the detection /imaging device are located on the same side of the object. As a result, CBI is a valuable nondestructive evaluation (NDE) tool because of its single-sided nature, penetrating abilities of radiation, and unique interactio n properties of radiation with matter. Changes in the backsc atter photon field intensity (res ulting in contrast changes in images) are caused by differences in absorption a nd scattering cross sections along the path of the scattered photons. Since the inception of CBI, a diverse set of imaging techniques have evolved using both collimated and uncollimated detectors, coded apertures, and hard x-ray optics. Work here at the University of Flor ida focuses on backscatter Radiography by Selective Detection (RSD), Lateral Migration Radiograp hy (LMR) and Shadow Aperture Backscatter Radiography (SABR). Pencil Beam Compton Scatter X-ray Imaging Pencil beam Compton scatter imaging uses a high ly collimated pencil beam of radiation to interrogate objects. The pencil beams may vary in size from microns to centimeters, but usually consists of a near-parallel arra y of photons forming a tight beam. Highly Collimated Techniques As early as 1956, Odeblad and Norhagen1 published results describing the effect of changes in electron density of a localized volum e on Compton scattered photons. Changes in the electron density of the material were measured us ing a highly collimated detector and collimated 60Co gamma source. The volume of material bei ng interrogated is determined by the intersection of the field-of-view (FOV) of a collimated detector and the collimated source forming a small

PAGE 15

15 voxel (Figure 1-1). Highly collimated configur ations are dominated by first scatter components from the voxel because of th e geometric configuration. Figure 1-1. Highly collimated x-ray backscatter imaging, voxel by voxel This first approach lends itself very well to x-ray backscatter to mographic imaging (TI).2 Typically, a highly collimated source and se t of concentric conical collimators3 are used to isolate a volume in the object being interrogated; this volume is then represented as a voxel in the tomographic image. Multiple-scatter components are considered noise and removed with collimation. These TI backscat ter devices then scan the object voxel-by-voxel, to generate a three-dimensional image. Because the design of the collimators collects photons from a very small solid angle, the resulting signal at the de tector is dominated by first-scatter components from the detector FOV. While highly collimated detectors and backscatter tomographic images provide some of the highest possible contrast im ages, some in three-dimensions, the acquisition time can be extremely long. A technique called differential gamma scattering spectroscopy4-5 uses the Compton shift in energy to determine the spatial lo cation of the scattered beam. A lthough this technique is faster, it may require the use of high energy resolution detectors such as High Purity Germanium (HPGe) for some applications. Differential scattering spectroscopy has been revisited by Detector Collimator Noise Signal X-ray generator

PAGE 16

16 Lawrence Livermore National Lab (LLNL)6 using room temperature detectors such as Cadmium Zinc Telluride (CZT) and a brem straulung x-ray spectrum in a n ew technique called virtual collimation. While dealing with a source spectrum complicates the spectral analysis, this is a variation of the gamma-ray techniques used to measure bubble size and dist ribution in two-phase flow.7 Essentially a planar 241Am source illuminated two-phase flow in a pipe. Compton shift in energy was used to determine the annular locatio n and the size of the bubbles in the two-phase flow. Dynamic radiography8 is a type of scatter x-ray imagi ng (SXI) for interrogating objects in motion. The objects can have natural phonon moti on, such as fluid flow, a beating heart, breathing lungs or phonon motion can be mechanical ly provided. This technique, like other highly collimated techniques, uses the intersection volume of a collimated radiation beam and detector. As the material in the illuminated volume oscillates, any significant changes in the cross sections will be measured in the detectors. The periodic motion provides a frequency correlation between the responses of different detectors, which would otherwise have independent responses, because a scattered photon dete cted in one detector can not be detected in the other detectors. Multiple detectors allow for the collection of data in different directions which can be used to generate three-dimensional images. This technique was used to measure the irregular motion in th e beating heart of a dog caused by a decrease in the blood supply or by constriction or obstruction of the blood vessels.9 Uncollimated Backscatter Radiography Because of the relatively long acquisition time required by highly collimated detectors, some CBI techniques use detectors without collimators. Instead of using the intersection of the detector FOV and illumination beam to form a voxel, the image is reconstructed from the assumption that all detected backscatter photons have originated from the illumination beam

PAGE 17

17 spot.10 A variety of commercial equipment is av ailable for many scanning applications which include imaging trucks, cargo, people and luggage.11 These techniques use very large detectors, sometimes on the order of square meters of dete ction area, to collect as many scattered photons as possible. While this uncollimated scanning ap proach is very fast and has a large number of applications, it is limited in subsurface re solution, because the signal is dominated by first-scatter, near surface components (Figure 1-2). Figure 1-2. Uncollimated pencil beam CBI technique Subsurface features that are lo cated more than one mean-fr ee-path (mfp) into the object can be difficult to image because the feature sign al can be masked by near-surface, first-collision components. Collimation can be used to improve contrast and depth resolution, even for near surface features. Radiography by Selective Detection Radiography by selective detec tion (RSD) is a pencil beam Compton backscatter imaging technique that falls between highly collimated and uncollimated techniques. Uncollimated techniques are dominated by first-scatter componen ts from near the surf ace of the interrogated object. While these techniques are very fast, uncollimated techniques lack subsurface resolution at depths beyond a mfp, and collimation often incr eases contrast and depth resolution even at shallow depths. Highly collimated techniques can image at depth, but are usually very slow Detector Noise Signal X-ray generator

PAGE 18

18 because the collimation typical ly only detects single-scatter components from a small voxel formed by an intersecting FOV between the illumi nation beam and the detector. Radiography by selective detection uses a combination of singl eand multiple-scatter phot ons from a projected area below a collimation plane (CP) to generate an image (Figure 1-3). The collimation plane is a plane to which a photon must penetrat e to enter the FOV of the detector. Figure 1-3. RSD collimated and uncollimated detector with collimation plane Lateral Migration Radiography A subset of RSD, Lateral Migration Radiograp hy (LMR), was used for the detection of buried landmines.12-20 The image contrast is dominated by third-order scatter x-ray components and higher. Because typically the illumination x -ray beam penetration barely extends to a depth much beyond the base of the mine, the mine (o r surrounding soil) becomes a diverged scattered x-ray source for the properly collimated dete ctors (Figure 1-4). Wh ile second-order scatter components are still close to the penetrating beam, third and higher order scatter components migrate far enough from the origin al beam, to produce a laterally spread scatter source in the landmine (or surrounding soil). Usually very large detectors (on the order of 0.3 m2) are required to capture these laterally spread, multiple scatter components. The term LMR carried over to NDE experimentation, 21-23 but was later changed to RSD. Detectors Collimator Noise Signal X-ray generator Collimator Plane

PAGE 19

19 Figure 1-4. LMR collimated regime and uncollimated detector RSD versus LMR For RSD, a collimated source beam is used to illuminate an object of interest. Detector collimators are adjusted to prefer entially receive signals from be low a selected depth to remove scatter components which have not traversed th e subsurface feature of interest and would, therefore, only add noise to the image. As with the landmine imaging, uncollimated detector information may be used to remove surface features. However, if the surface does not superimpose a strong signal on the subsurface, th en it may not be necessary to perform such image processing. RSD imaging for smaller features is different from LMR imaging for landmines. In many cases the illumination beam penetrates beyond the feat ure. As a result, the image contrast for the defects tends to be dominated by firstand/or secondscatter components, especially in low Z materials. These second-order scattered phot ons do not migrate very far from the path of first-scattered photons and the te rm RSD is used to described the process by which photons carrying information about the fl aw are detected. This method differs from LMR, because the subsurface feature is relatively small compared to the mfp of the interrogation photons. As a result, multiple-scatter photons traverse the phy sical boundaries of the sm aller feature and move Detectors Collimator Noise Signal X-ray generator Landmine Earth

PAGE 20

20 into the surrounding material. Smaller detectors also favor an RSD regime over an LMR regime, because the scattered photons can easily move outsi de the FOV of the smaller detectors. LMR is still an RSD technique because LMR selectively detects backscatter components that yield the greatest contrast enhancement. Full-Field Illumination Compton Backscatter Imaging Unlike pencil beam CBI techniques, full-fiel d illumination technique s flood the entire FOV at once with x-rays. For these techniques, collecting and/or deci phering the backscatter field may become challenging. These techniques t ypically use a coded aperture or optic lenses for focusing. Coded Aperture Imaging When producing images it is often necessary to focus the photons onto some type of detection array. X-rays with less than 10 keV can be optically reflected, but beyond 10 keV the grazing angle for total external reflection become s very small. At 30 keV the critical grazing angle for gold is 0.153 degrees.24 In order to effectively image higher energy photons from multiple sources, coded aperture imaging uses stra ight-line optics, a coded mask such as shown in Figure 1-5, and a detection array capable of sensing straight-l ine projection patterns.25 If only a single point source is present in the FOV, d ecoding the image based on straight-line optics is trivial. But when multiple source points, such as multiple stars, illuminate the FOV, multiple images are projected on the detector sensor array. These images can be decoded without distortion when the illuminating beams are parallel, which is the case for far-field objects like stars. However, near field coded aperture imaging is more complicated; the source rays are not parallel because the scatter field does not come fr om far-field point sources. This results in artifacts or distortion of the images. One method to reduce the distortion is to obtain two images

PAGE 21

21 with a mask and an anti-mask. When the mask and anti-mask images are summed, they constructively interfere; similarl y, the difference of mask and anti -mask images cancel and leave noise.26 Even with these advances near-field coded aperture imaging does not yet have high resolution, and images may require very large acquisition times (on the order of hours) and very long deconvolution times (hours).27 Figure 1-5. Example of a coded aperture using a Modified Uniformly Redundant Array (MURA) Optical X-ray Focusing Lobster eye optics is an optical x-ray imag ing technique that has its birthplace in astronomy.28 The idea of using the lobster eye to focu s x-rays is conceptualized in Figure 1-6. Until recently, because of the very small critical grazing angle, devices for focusing hard x-rays (greater 10 keV) were very large and impractical.24 But by using highly-polished, micro-tubular channels to focus the incoming x-rays a lobster eye imaging device can be constructed to focus x-rays. Physical Optics Corporation (POC) has proposed a hand-held lobster eye x-ray inspection device (LEXID) listed as an emergi ng technology. Figure 1-7 shows the design for the hand-held product without an x-ray source.

PAGE 22

22 Figure 1-6. Conceptual drawing of lobster eye focusing parallel photons Figure 1-7. Emerging technology from POC for l obster eye hand held x-ray imaging device

PAGE 23

23 CHAPTER 2 SCANNING SYSTEM OVERVIEW Lockheed Martin Prototype System In 2003 spray-on foam insulation (SOFI) from th e external tank of th e space shuttle tore loose during the Columbia launch and punctured th e leading edge of the wing on the orbiter. Because the scatter x-ray imaging (SXI) rese arch group at the University of Florida demonstrated the ability to detect simulated defects in the SOFI as part of a Lockheed Martin Space Systems Co. NDE initiative, funding was gran ted to build a prototype device (Figure 2-1). Figure 2-1. Prototype RSD scanning device built for Lockheed Martin Space Systems Co. The x-ray tube is shielded with lead to redu ce image noise. The shielding prevents leaking x-rays from the tube from reaching the detector An illumination beam tube collimator with a small aperture shapes th e illumination beam into a pencil be am for scanning. Four NaI detectors with finned collimators selectively detect back scattered photons to improve the contrast of subsurface flaws and defects. After prototyp e testing was completed with highly favorable Lead Shielding Finned Collimator Illumination Beam X-ray Tube

PAGE 24

24 results, the design and construction of RSD s canning systems for industrial use began. RSD scanning is one of two new and/or existing NDE technologies that were chosen as favorable for detecting flaws and defects in the SOFI of the space shuttle external tank. First Industrial Scanning Systems Work on this dissertation begins with the firs t industrial RSD scanning systems. The first industrial systems had to be designed and built in a matter months, leaving little time for modeling and simulation to build an optimized sy stem. Off-the-shelve components were chosen based on past experience and deliv ery lead times and then assembled in a configuration based on the success of the prototype. Figure 2-2 shows the first industr ial RSD scanning system which consists of the x-ray generator, an array of dete ctors with their associat ed electronics, a scanning table, and a computer to cont rol data acquisition, motion contro l, and image generation. The array of detectors is fixe d to the x-ray tube and designated as the scanning head. A highly collimated x-ray beam illuminates a single pixel, and a selective backscatter field is measured by the array of detectors. Movabl e collimators allow each of the de tectors to view a unique field. The measured signal from less collimated, or uncollimated detectors is dominated by singlecollision events and contains surface and near surface information. The collimated detector can respond to singleand multiple-scatter photons which have penetrated beyond the collimation plane. When properly collimated, these photons ca rry information about subsurface features. Two-dimensional images are generated using a scanning pattern. For example, the scanning head will sweep from left-to-right, acquiring data and storing a line of pixels. The scanning head will then move to the next line and sweep in the opposite direction from right-to-left, obtaining the next line of data. This pr ocess is repeated, one line at time, until the entire image is completed.

PAGE 25

25 Figure 2-2. First commercial RSD sca nning system configuration Original Industrial System Component Details The yellow cylinder in Figure 2-2 is an Yxl on MXR-160/22 x-ray generator. This is a liquid-cooled x-ray generator with a maximum tube voltage of 160 kV. However, the tube voltage is limited to 100 kV because of the 30 meter long, high-voltage (HV) cable connecting the x-ray tube to the HV power supply for the spec ial application of scanni ng the external tank of the space shuttle. The x-ray tube has a maximum current of 45 mA with a 5.5 mm x-ray tube focal spot (FOC), with a 3000 watt maximum powe r rating. For SOFI applications, the x-ray source is nominally operated at 55 kVp and 25 to 45 mA. The four silver cylinders in Figure 2-2 are the detector assemblies. Each detector assembly includes a 5.08 cm diameter by 5.08 cm long NaI scintillator crystal, a photomultiplier tube and a custom, low-noise pre-amplifier. The collimator assembly at the end of the detector includes an array of lead collimators and the design allows for independent adjust ment of the assembly in different directions. This includes in-and-out movement of the out er, circular (sleeve) collim ator; in-and-out and rotational

PAGE 26

26 movement of the inner collimator (collimator co mponent with the lead fins); and in-and-out movement of the entire assembly. The collimat or design provides the ability to focus the image by the selection of the desired scatter co mponents. Each of the detectors generates a separate image and a cross-correlated image can also be generated from any combination of detector images. The x-ray illumination beam sp ot size used is typi cally 2 mm for first-pass scans, and 1 mm for suspect areas or areas of in terest; the illumination exit beam aperture is located approximately 280 mm for the x-ray tube FO C. The beam spot for the first applications was round but customizable for different applicat ions. The size and shape of the illumination beam is controlled by a lead insert in the bottom of the brass, lead-lined source collimator tube that extends in a direct ion normal to the end of the x-ray tube and is cen tered between the four detectors. The scanning head assembly (x-ray tube, detectors, collimators, and electronics) weigh about 29.5 kg (65 lbs). The fastest linear s can rate for this system is about 50 mm per second. For 2 mm pixels, this tr anslates into a scanning rate of approximately 15 minutes per 0.093 m2 (1 ft2). Traditional NIM-rack component s were used for pulse shaping (Ortec 679 fast filter amp) and noise rejection (Ortec 850 quad SCA) before pa ssing the digital count ra te for each detector channel to the data acquisition software through a BNC 2121 interface and a National Instruments (NI) 6602-PCI counting card. A LabVIE W-based program is used to control the scanner motion, data acquisiti on and image generation.

PAGE 27

27 CHAPTER 3 ARTIFACT AND CONTRAST GENERATI ON IN SCATTER X-RAY IMAGING Contrast and Artifacts The work on artifact and contra st generation is presented next to aid in understanding the analysis of system components in future chapters Some of the initial foundations for this work are based on previous research.29 Under a simplistic CBI model, contrast is determined by changes in a scanned objects electron density and/or scattering-t o-absorption ratio. While this is true for first-scatter models, multiple-scatter photons, detector collimation, and feature geometric locati on play a significant role in determining the relative contrast of the ob ject and its associated fe atures. Features with higher scattering-to-absorption ratios than thei r surrounding media can app ear either dark or bright depending on detector collimation and f eature location with resp ect to the collimation plane and the illumination beam. Bright veils and dark shadow effects can make a subsurface feature appear brighter, darker, or possibly even obscure the feature so that it is no t visible in the image. Geometric location and orientation of the illumination beam, subsurface feature, and detectors can affect relative c ontrast as much as changes in electron density. Understanding how these factors affect contrast is essential to usi ng x-ray backscatter as an imaging technique in any application. Compton Backscatter Imaging Characteristics Regardless of the application, the method in wh ich subsurface features are detected is by changes in contrast with respect to the surrounding materi al. For gray scale images, an increase in detection rate is shown as an area of bright contrast. Likewise a decrease in signal is indicated by an area of dark contrast.

PAGE 28

28 Pencil Beam, Single-Scatter Compton Backscatter Contrast For pencil beam CBI the change in count rate or signal intensity at the detector is directly proportional to the change in contrast in the imag e. To understand the changes in contrast, it is important to understand the factor s that affect the detector in tensity. For this simplified discussion the following assumpti ons apply: coherent scattering is negligible and scattered photons along the incident and ex it path do not significantly contri bute to the detector response. The incident attenuation factor, the probability that a photon reaches a point of interest, is given by e. The attenuation along the incident path for a particle of a given energy is the integral of the photon total macroscopic cross se ction of each material and the incident path length ( ri) to the point of interest or = i i idr r ) (. Once at a point of interest, th e interaction factor is the probability the photon undergoes a scattering event with an electron into the solid an gle of the detector. The interaction probability factor is given by M Z N d d da str KN total 1 where, total is the total macroscopic scattering cross section at the interaction point, KN is the microscopic Klein-Ni shina (KN) scattering cross section, integrated over all directions in the solid angle to the detector, d is the differential scattering angle to the detector, Na is Avogadros number, Z is the number of electrons per nuclide, is the density, and M is the molar mass. The detector solid angle FOV bounds the integral, and density and material change s affect the macroscopic cross section. After the photon is scattered toward the de tector, the exit attenuation factor is the probability that it reaches the detector una ttenuated. The exit factor is given by e, the probability to of reaching the detector unattenua ted. The attenuation along the exit path for a

PAGE 29

29 particle at a given energy is th e integral of the phot on total macroscopic cross section of each material and exit path length ( re) to the detector, = e e edr r ) (. The intensity of the backscattered field is affected by the combination of these three factors, incident attenua tion, interaction, and exit attenuation probabilities. The probability, per source particle, that a photon has a scattering event into the detect or FOV at a point is given by ) ( 1 ) ( e M Z N d d d ea str KN total. Changes in this probability distribution are proportional to detector intensity changes. The contrast of a subsurface featur e is proportional to the ratio of the intensity at the detector when the illumination beam is over the subsurface feature (IDF), to the intensity at the detector when the illumination beam is not over the feature (IDNF): Contrast IDF/IDNF. This formulation implies that the contrast is not only dependent on the scanned objects electron density but highly dependant on the particle path from the illumination to the detector. Changes in contrast are due to a combination of the photon path, and the interactions along that path. Incident attenuation, inte raction, and exit attenua tion factors can increase or decrease the total contrast and the interplay of these factor s determines the change in contrast. Without knowing specific geometric configur ations and material properties, it is difficult to determine which factor will have the most effect on the contrast. The following sections discuss how the contrast of different subsurface features change as the illumination beam moves towards, directly illuminates, and depa rts from the physical feature. These scenarios assume a known geometric configur ation for the feature, beam and detector and that the feature is completely encompassed by homogenous materials. While this set of

PAGE 30

30 scenarios is not exhaustive, it provides an i llustrative foundation of fundamental scattering behavior to determine if a feat ure will be bright or dark with respect to its surroundings. Void Contrast and Shadowing Effects Many subsurface features are voids such as fuse wells in landmines, debonds, delaminations, cracks, and some types of corrosi on have void characteristics. Voids can be defined as areas of substantially lower density than the surrounding material, for example, air may be considered void even in low density insulations (0.03 g/cm3) and nylon may behave as void in tungsten. The first set of void scenar ios includes voids that are above the collimation plane (CP). Voids above the CP usually result in an increase in intensity or bright contrast in an image. As shown in Figure 3-1A, as the illumi nation beam approaches the void region, the exit attenuation to the detectors decreas es resulting in an increase in c ontrast. This yi elds a bright veil that appears between the illu mination beam and the detector. Once the illumination beam is over the defect (Figure 3-1B) the intensity can further increases for several reasons: less attenuation for penetration below the CP, the solid angle to ente r the detector increases with depth below the collimation plane, and there may be re duced exit attenuation if the return path traverses the void. As the illumination beam depa rts the void, the brightness can decrease due to the increased attenuation on the exit path (Figur e 3-1C) and the FOV formed by the intersection of the detector and illumination beam increases with de pth (Figure 3-2). Figure 3-3 shows scenarios where the feature is below the CP. The approach, again, may result in a bright veil due to a decrease in attenuation along the exit path (Figure 3-3A). Once the illumination beam is directly over the void (Figure 3-3B), the feature can appear either bright or dark in relative contrast. The removal of a sca tter source from the detect ion region results in the photons traveling deeper because of the reduced incident attenuation. This increases the exit attenuation and could result in a da rk contrast. However, recall that the detector solid angle from

PAGE 31

31 the illumination beam increases with depth (Figur e 3-2), and the resulting increased interaction rate in the larger FOV of the detector tends to ward a bright contrast. The interplay of these factors determines whether the void region will appear light or dark. This factor is very sensitive to relative depth below the CP. These factors may even combine in such a way to cancel, and result in no change in contrast. As the illumi nation beam begins to de part the physical feature (Figure 3-3C), the void should appear darker because the exit atte nuation is increased. Scattering and Absorption Contrast, and Shadowing Effects Whether a feature is a scattere r or absorber is determined by the surrounding material. Aluminum with respect to lead would be considered a scatteri ng feature, but aluminum with respect to nylon would be considered an absorb er. Scattering features may also be very low density compared to the surrounding materials. For low density scatteri ng features, the void scenarios can play a dominant role in determinin g the feature contrast. As the illumination beam approaches scattering features that are above the CP (Figure 3-1A), the feature will increase attenuation along the exit pa th resulting in a dark shadow be tween the illumination beam and the detectors. When the scattering feature is directly illuminated a nd above the CP (Figure 3-1B), the feature may appear bright or dark. If the de nsity of the scatterer is approximately the same order of magnitude or higher th an the surrounding medium, then this scattering feature may act as an attenuator on both the illumination and exit pa th resulting in a dark image. If the scattering feature has very low density with respect to the surrounding materi al then it may behave as a void region and appear bright. A near perfect scatterer (scattering to total cross section ratio approximately one) may also cause can an increase in contrast with an LMR effect (diverged laterally-migrated scatter sour ce). As the illumination beam departs the physical feature (Figure 3-1C), the feature will appear darker than the surrounding material, unless it is comparatively low density.

PAGE 32

32 Figure 3-3 shows the scattering feature belo w the CP. The approach (Figure 3-3A), usually results in a dark contrast due to an in crease in attenuation along th e exit path. Scattering features are typically bright wh en directly illuminated if the scatterer consequently reduces incident and exit attenuation (Fig ure 3-3B). Because the scatte ring paths do not change as the illumination departs the subsurface feature as show n by the path of shallow scatter (there are 2 scatter paths) in Figure 3-3C, the contrast remain s bright. However, if the scattering feature is very low density, then the void scenarios for co ntrast behavior interp lay with the scattering scenarios. In these situations simulation or experimentation may be required to determine the dominate factors of contrast generation. To predict relative contrast for absorbing featur es is relatively simple. The factors in play for absorbing features will always induce a dark contrast and result in a shadow between the illumination beam and the detector on approach. Each of the scenarios for s cattering, absorbing and void f eatures is summarized in Table 3-1. Detector, collimation, illumination b eam, feature geometry and relative position are extremely important in determining relative contra st for each of these scenarios. Different combinations of contrast factors and scenario s can combine in ways that make the feature contrast intuitively difficult to predict. These factors can even combine such that no change in contrast occurs. Two experiments are discussed in the following sections. The first experiment demonstrates some of the contrast scenarios disc ussed in the ideal situat ions described above. The second experiment demonstrates how inhom ogeneities and complex geometry in an NDE application can make contrast changes difficult to predict.

PAGE 33

33 A B C Figure 3-1. Subsurface features abov e collimation plane A) pre-direct, B) direct illumination, and C) departing illumination Signal Detector Collimator X-ray generator Collimation Plane Subsurface feature Signal Detector Collimator X-ray generator Collimation Plane Subsurface feature Signal Detector Collimator X-ray generator Collimation Plane Subsurface feature

PAGE 34

34 Figure 3-2. Increasing detection solid angle from points along illumination beam with increasing depth from the CP Collimation plane (CP) Detector with collimator Illumination beam (1) (2) (3) Detector solid angle to a point on illumination beam

PAGE 35

35 A B C Figure 3-3. Subsurface feature below collimation pl ane (A) pre-direct illumination (B) direct illumination and (C) departing illumination Signal Detector Collimator Collimation Plane Subsurface feature X-ray generator Signal Detector Collimator X-ray generator Collimation Plane Subsurfacefeature Signal Detector Collimator X-ray generator Collimation Plane Subsurface feature

PAGE 36

36 Table 3-1. Relative grey scal e contrast table for various RSD imaging modalities based on feature type, and relative location Relative Void Figure 3-1A 3-1B 3-1C 3-3A 3-3B 3-3C Above CPa x x x Below CPb x x x Approachc x x Directd x x Departuree x x Relative Contrastf Bright Bright Bright Bright Bright or Dark Dark Relative Scatterer Figure 3-1A 3-1B 3-1C 3-3A 3-3B 3-3C Above CPa x x x Below CPb x x x Approachc x x Directd x x Departuree x x Relative Contrastf Dark Dark or Bright Dark Dark Bright Bright Relative Absorber Figure 3-1A 3-1B 3-1C 3-3A 3-3B 3-3C Above CPa x x x Below CPb x x x Approachc x x Directd x x Departuree x x Relative Contrastf Dark Dark Dark Dark Dark Dark a subsurface feature located above collimation plane, Figure 3-1 b subsurface feature located belo w collimation plane, Figure 3-3 c subsurface feature as illumination beam approaches, Figures 3-1A and 3-3A d subsurface feature di rectly illuminated, Figures 3-1B and 3-3B e illumination beam departing from subsur face feature, Figures 3-1C and 3-3C f relative contrast in grey scale imag e with respect to surrounding material

PAGE 37

37 Measurement Results and Discussion The SXI RSD scanning system was used to scan an aluminum plate to demonstrate some of different contrast mechanisms and artifacts from pencil beam CBI for ideal situations. An aluminum plate (Figure 3-4) has three channe ls that are 10 mm wide, 2 mm high, and run the length of the 150 mm plate. The top of the chan nels are located at dept hs of 3 mm, 5 mm, and 7 mm, from left to right in Fi gure 3-4. The object was scanned with the illumination beam and channels parallel to detectors two and four, and perpendicular to detectors one and three (Figure 3-5). The placement of the images in Figure 3-5 corresponds to the relative detector position during the scan. The x-ray tube voltage was 75 kV, at 40 mA, with an FOC of 5.5 mm. A 1.0 mm beam aperture with 1.0 mm pixels and a dw ell time of 0.1 seconds per pixel were used to acquire the image. Figure 3-4. Aluminum sample plate with 10 mm wi de, 2 mm high, channels at depths of 3mm, 5 mm and 7 mm from left to right The images for detectors one and three ar e presented in Figures 3-5A and 3-5C, respectively. In this geometry the channel wa s orientated perpendicular with respect to the detector and the illumination beam. The void ch annels are below the collimation plane which corresponds to the scenarios in Figure 3-3. As expected a br ight veil appears when the illumination beam is between the channel and the detector due to a reduced exit path (illustrated in Figure 3-3A) of the scattered photons. Note that because of the mirror symmetry of the

PAGE 38

38 channel, illumination beam, and detector arrangement, the bright veil effect is mirrored in the corresponding images in Figures 3-5A and 3-5C. Direct illumination appears dark (Figure 3-3B) because a scatter source was rem oved from the FOV and the photons must now travel deeper and some of the exit paths may not be through the ch annel. Opposite of the bright veil, a dark shadow can be seen cast in the channel. This da rk contrast area in the channel indicates that the exit path of the majority of photons is no longer through the channel, but through solid aluminum. This corresponds to Figure 3-3C and again, th e effect is mirrored between Figure 3-5A and 3-5C. The collimator settings and, therefore, the col limator plane for detector two are identical to collimator settings for detectors one and three. The channel is below the collimation plane. The only difference is that the channe l orientation is along the same line as the illumination beam and detector. There is no bright veil (except possibly in the corners of the imag e) or dark shadow in this image because their production mechanism wa s removed. Decreased incident attenuation to a deeper point leads to an increased interacti on probability (increased FOV to the detector, Figure 3-2) and feature or ientation ensures decreased exit atte nuation from this deeper point both contributing to a bright contrast. Detector four is over collimated, and the channels in the aluminum are above the collimation plane (Figure 3-1). The channel orie ntation is along the same line as the illumination beam and detector, just as for detector two. Because of the over collimation (about 10 mm of collimation depth), the backscatte r signal is on average 40 times lowe r than in the other detectors in this configuration, but the cha nnel still appears bright er in contrast (Figur e 3-5D). This is because of the reduced attenuation along the entr ance illumination path and reduced attenuation along the exit path.

PAGE 39

39 Figure 3-5. RSD image results of th e aluminum plate (Figure 3-4) 1 2 3 4 Illumination beam between four detectors D) Channels above collimator plane. Detector 4 A) Channels below collimator plane. Detector 1 B) Channels below collimator plane. Detector 2 Bright veil Edge shadow Bright veil Edge shadow C) Channels below collimator plane. Detector 3

PAGE 40

40 Application of Contrast Mechanisms Changing contrast mechanisms can be used to find the depth of a feature. As shown in Figure 3-6, a void is located in the gap filler material of a composite sample and the sample is on an aluminum substrate. The purpose of the in spection is to find void locations, but because of inhomogeneities and geometric configuration, the gap filler material (a scatter) controls the contrast factors. As a result, the scattering scen ario logic must be applied, even though the scan is to determine void depth. A SXI RSD scan of the object was taken with the CP set to increasing depths of 3 mm (Figur e 3-7A), 6 mm (Figure 3-7B) a nd 9 mm (Figure 3-7C). The images were obtained with a tube voltage of 60 kV and current of 45 mA, with 1 mm illumination beam and image pixels. When the scatterer (gap filler materials) moves from below the CP (Figure 3-7A), to above the CP (Fi gure 3-7B) the feature changes in contrast from bright to dark, thus indicating the end of filler and the beginning of the void region is located between 3mm and 6 mm into the interrogation object The actual void re gion is located about 4 mm into the object. Figure 3-6. Composite material with a void region in the gap filler Filler Void

PAGE 41

41 A) 3 mm CP depth B) 6 mm CP depth C) 9 mm of CP depth Figure 3-7. RSD scanned image of a composite material with a void region in the gap filler Conclusions about Artifact and Contrast Generation Feature contrast in a pencil beam Compton backscatter imaging system is directly proportional to the change in intensity of the back scatter field measured by the detectors. This formulation for feature contrast implies that the contrast is dependent on the incident attenuation factor, interaction factor and ex it attenuation factor. To what de gree each factor contributes to the change in contrast is a f unction of the photon path, feature pr operties, and relative geometry orientation of the feature, illumination beam and detectors. The dominant factors tend to be the exponential terms along the incident and exit paths. But changes in scattering cross section, and change in solid angle, as well as the detector-, target-, feature-, and beamrelative geometry affect the interaction factor and consequently co ntrast. The interplay of these contrast factors determines the change in contrast in diffe rent pencil beam Compton backscatter imaging techniques such as highly collimated, uncol limated, and Radiography by Selective Detection. An understanding of these contrast factors and th eir interdependencies ca n be used not only to detect features, but to predict parameters such feature depth, size and orientation making Compton backscatter imaging a valu able single-sided imaging tool.

PAGE 42

42 CHAPTER 4 OPTIMIZATION AND ANALYSIS OF RS D SCANNING SYSTEM COMPONENTS Detector Testing The first production radiography by selectiv e detection system (RSD) detectors for Michoud Assembly Facility (MAF) were 5.08 cm di ameter, 5.08 cm thick NaI(Tl) detectors with photomultiplier tubes (PMTs) from Bicron with a 400,000 max count rate custom preamplifier from Inspirion, LLC. One of the very first improvements to the system was to modify the time constant of the pulse preamplifier to count as quickly as possible c onsidering the decay time constant of the NaI(Tl) crystal (230 ns). Th e use of this fast preamp enabled an 800,000 (1 MHz analog bandwidth) max count rate detect or. The immediate impact was to allow the RSD scanning system to acquire data twice as fast, likewise reduc ing the scanning data acquisition time without increasing the statistical error in the counts. These NaI detectors are used in every RSD system with the exception of the new compact prototype system, to be discussed in Chapter 5. NaI is considered the standard scintillator to which most scintillator detectors are compared. Because of NaI detect ors successful results in RSD imaging and flaw detection, it will be used as a standard for detector development benchmarking. Comparison of NaI and plastic (BC404) scintillation Some RSD system NaI detectors are 5.08 cm in diameter and 5.08 cm thick. Although a much thinner NaI crystal (6.35 mm thick) is ad equate and provided the same quality images, 5.08 cm thick crystals were more readily availa ble off-the-self. Plastic (BC404) scintillator detectors were tested in comparison to the NaI detectors, because plastic s have a much faster decay time (1.8 ns), and could in theory measur e higher radiation fields. The plastic detectors have a density of 1.032 g/cm3 and a peak emission of 408 nm.30 The plastic detectors tested also had a 5.08 cm diameter, 5.08 cm thick active dete ction volume, and used about a 1 microsecond

PAGE 43

43 pulse width preamplifier in count mode. Howe ver, plastic detector s were about 30% less efficient than NaI detectors. As a result the c ount rate on the plastic de tectors was about 2/3 of the count rate of the NaI detectors. A lower count rate resulted in images with less contrast in comparison to the NaI detectors and potentially longer image acquisition times. Plastic detectors should be investigated in a curr ent mode configuration. Becaus e one of the inherent problems with pencil beam CBI is the source intensity limitation, work with small area plastic detectors stopped after several image measurements were take n. But, plastic detectors may still play role in SXI because of their large si ze, speed, availability, and ease at which plastics can shaped into optimal geometries. Copper-doped Quartz A copper-doped quartz scintillator material is currently being used for making dosimetry measurements in medical physical.31 This material has physical properties such that it can be optically fused to a fiber optic cable with a hi gh coupling efficiency; light loss at the interface coupling is less than 1%. Most of these dosim etry detectors are about 1 mm in diameter and demonstrate nominally good light output in the energy range for x-ray ba ckscatter (40 80 kVp) for low Z material NDE. The success and efficiency of the these detectors led to the testing of a larger copper-doped quartz crystal, about 2.54 cm in diameter by 0.635 cm thick. It was mounted to a Hamamatsu R6095 PMT and tested. Th e detector response to the x-ray backscatter field was about two orders of ma gnitude less than for the NaI dete ctors, so no further testing was performed. The low response of th e crystal is believed to be rela ted to the large size of the crystal used for the backscatter application. Most of the doping material remains near the surface of the crystals. This is ideal for small detector s but not for the larger area detectors needed for x-ray backscatter imaging, because the crystal may actually attenuate th e scintillation light source.

PAGE 44

44 PIN-Diode Detectors The use of PMTs along with an appropriate sc intillation crystal such as NaI(Tl) or YSO provides for a very fast and high resolution detection and imaging system; however, photomultipliers tend to be large and fragile, and usually require external biasing and shaping/amplification electronics. PIN diode de tectors have the followi ng advantages: require little or no biasing, are very small in size, a nd much more rugged than a PMT. The primary disadvantage of the PIN is that it lacks the signal -to-noise (SNR) of a typi cal PMT type detector. In an attempt to overcome this deficiently Inspirion, LLC worked on developing a new preamp. Universal charge preamp To improve the SNR a new preamp was desi gned. This preamp design is based upon a low noise, room temperature Charge type prea mplifier, designed for pin-diode testing but universally ported over to the YSO dete ctor, which has the following features:32 Higher front-end gain for better overall noise figure Much lower noise front-end amplifier (FET transistor) Independent Mu-metal shield Digital baseline adjust potent iometer for lower micro-phonics 3rd order "T" filter for 12V input power Lower noise 2nd. stage amplifier32 This design also has some very important mech anical advantages. Primarily, the use of a solid state adjust potentiometer reduces the heig ht of the card by 50%. The new preamp design is truly universal, in that it can be used w ith both PMTs and PIN diodes. The card accepts a wide variety of PIN sizes and F ET footprints to match the charac teristics of different detector types and PMTs, as well as time constants.32 In addition to pindiode applications, the performance (gain and SNR) of the universal pr eamp warranted implementation in YSO detector applications.

PAGE 45

45 Pin-diode test results Only one pin-diode detector has yielded test results over the past th ree years. A room temperature PIN from Advanced Photonix, model SD 445-14-21-305, is coupled to a CsI scintillator, measuring 1 cm x 1 cm x 0.1 cm. Th e pin-diode detector operates in current mode (integral of the pulse mode). An NI-PCI6115 A/D card was used to measure the changing voltage from the pin-diode detector. The voltage is sampled over a pixel, and an average voltage is calculated. The average voltage over a pixel is then mapped to a 16 bit integer range from 1 to 65536 and stored in a data array. The data array is a tab-delimited ascii text file, the same format as for the counting mode detectors, to a llow the same LabView coding to handle image processing and generation for both current and code mode detectors for the RSD system. An aluminum plate with five holes machined in the back side was imaged for a comparative evaluation. For a quantitative compar ison, the relative contrast of each flaw is calculated with respect to the surrounding back ground. The relative contrast is defined as: Background Background Signal rast lativeCont ) ( Re A positive contrast would indicate the defect is brighter than background, likewise a negative contrast would indicate the defect is darker than the background. For the holes on the upper left of the image, the NaI has a relative contrast of -2.1% (Figure 4-1A) and the pin-diode a relative contrast of 1.4% (Figure 4-1B). The pin-diode detector has SNR 50% lower than the NaI for the images in Figure 4-1. It should be noted that Figure 4-1B is the negative of Figure 4-1A. The negative imag e is due to a coding error that used a negative slope on the mapping function wh en converting the voltage to a 16-bit integer range. The current mode electronics have also had issues with temperature drift, which should be resolved in the near future.

PAGE 46

46 A B Figure 4-1. RSD image of an aluminum plate with 5 holes for, A) NaI PMT detector, and B) CsI pin-diode detector YSO Detectors Yttrium orthosilicate (YSO Y2SiO5) is a scintillating material with nominal density of 4.45 g/cm3. YSO has a decay time constant of 70 ns and a light output when coupled to a PMT of 1.2 times NaI, with a peak emission of 430 nm for a mono-energetic 511 keV source.33 YSO is not only faster than NaI, with more light out put, but it is not hydroscopic and more rugged. The first YSO detector (Figure 4-2A) used a 2.0 cm diameter, 0.635 cm thick YSO crystal. The crystal is optically coupled to a Hamama tsu R6095 PMT and a BICRON preamplifier. This detector is about 28 cm x 3.5 cm including the wire couplings and was assembled for testing purposes from surplus components. While the PMT was about 10 times noisier than the NaI PMTs, photo diode model 9266B, the relative count rate per unit area of detection surface was about equal to that of the 5.08 cm diameter NaI. The promising count rate led to the developmen t of a second YSO detector (Figure 4-2B). This detector uses a 2.54 cm diameter, 0.254 cm thick YSO scintillator with a Hamamatsu R6094 PMT. This detector used the same electro nic components as the Na I detectors that were being used in the Lockheed RSD systems. Th is detector is about 18.04 cm x 3.5 cm, including the wire connections.

PAGE 47

47 Figure 4-2. YSO detector evol ution, A) First YSO protot ype, B) Second YSO prototype, C) Third YSO prototype The third YSO detector (Figure 4-2C) is a co mpact design. This detector uses a 2.54 cm diameter, 0.635 cm thick YSO scintillator with a Hamamatsu R1924A PMT. This crystal is thicker than the second YSO crystal, to reduce the probability of cracking th e crystal (the thinner crystal was cracked during detect or assembly when the detector was dropped). The thicker crystal is more impact resistant and internal self-shielding from the th icker crystal does not measurably affect the light output for RSD imagi ng. This detector uses the same preamplifier electronic components as the NaI detectors currently used on the RSD systems and is about 12.7 cm x 3.5 cm including the wire coupli ngs and has a mass of 462.3 g including the collimator. In comparison, the NaI detectors currently used on the RSD scanners are about 30.5 cm x 8.26 cm with a mass of 2630 g. A B C

PAGE 48

48 Image Results Comparison for YSO Detectors One of the objects of interests used for compar ing detector results is shown in Figure 4-3. The external tank of the space shuttle is comprised of two tanks, a liquid oxygen and liquid hydrogen tank. These tanks are bolted together at a flange, a portion of which is shown in Figure 4-3. A stiffener stringer is attached to the external tank at each bolting location. The inside of the stringer is then filled with SOFI, covering the bolt. Small defects, approximately 4 mm wide, 4 mm deep, and 10 mm long were machined into the SOFI close to the underside of the bolt as shown in Figure 4-3. This area was th en imaged using the RSD scanning system with both YSO and NaI detectors. The images are shown in Figures 4-4 and 4-5 respectively. Because the defect is difficult to see in print, a line profile of the area with the defect is shown to the right side of each image to further illustrate th at the defect is visible next to the bolt. The relative contrast for the YSO det ector is about 4.7 % and 4.5 % for the NaI detector. It is difficult to determine which detector is actually functioning better, because as demonstrated in Chapter 3, the contrast is dependent on the rela tive geometry, which is always different for different detectors. Figure 4-3. Space shuttle external tank flange bolting area with stringer and SOFI defect Stringer Flange SOFI Defects Bolt

PAGE 49

49 Figure 4-4. YSO, 2.54 cm diameter, detector image for external tank flange bolt near stringer with associated line profile across a known defect near bolt Figure 4-5. NaI, 5.08 cm diameter, detector image fo r external tank flange bolt near stringer with associated line profile across a known defect near bolt 41500 42000 42500 43000 43500 44000 0510 (mm)counts 15500 16000 16500 17000 17500 0510(mm)counts

PAGE 50

50 RSD Scanning System YSO Detector and Preamp Upgrades YSO detectors generate images th at are on par or better than the larger NaI detectors. Over time, the YSO detector was further improved by u tilizing the new universal charge preamp (see section above) to improve th e SNR and a quick-connect LemoTM connector was used to allow the HV, 12V and signal to be attached with one connector. The YS O detectors warranted implementation as an enhancement into existi ng RSD scanning systems at Lockheed Martin Space Systems Co., Marshall Space Flight Center, and the University of Florida. The YSO detectors mount interstitially to the NaI detector s with an orbit holder as shown in Figure 4-6. The orbit holder allows the YSO detector to be placed in any orbit position around the NaI detector, slide axially up and down, and change th e polar angle with respect to the illumination beam. The 12 volt distribution box had to be modifi ed to power up to eight detectors. The high voltage (HV) distribution box reduc es the number of HV cables connected to the bulkhead. Only two HV cables are needed to power eight detector s at two different HV settings. Typically the NaI detectors are operated at 900 volts and th e YSO detectors are operated at 650 volts. Figure 4-7 shows a bottom end-on view of the de tectors of an RSD scanning system with all eight detectors mounted. Addition of the YSO detectors allows the RS D scanning system to acquire data on eight independent channels. The compact size, versatil ity and image quality of the YSO detectors for RSD imaging set the stage for the design and testi ng of a compact prototype system discussed in Chapter 5. Illumination Beam Aperture Geometry Extending downward in a direction normal to the x-ray tube is the illumination beam collimator. The illumination beam size and geom etry can be varied by changing a lead disk aperture where the x-ray beam exits the collimator tube. Originally, circular apertures were used

PAGE 51

51 to illuminate a single pixel (Figure 4-8A). Ho wever, image pixels are square and a square aperture increase s the illumination beam in tensity by a factor of 4/ (1.27 times). The larger beam area and intensity results in a higher count rate, and reduces image acquisition time. Also note, the illumination beam is chosen slightly sm aller than the pixel area (Figure 4-8A). For example, to generate an image with 2 mm pi xels, it is recommended to use a 1.5 mm aperture. This reduces pixel cross illumina tion due to beam dispersion and he lps to improve image quality. Figure 4-8B quantifies the illu mination beam dispersion at 101.6 mm from the beam exit aperture. A pixel size of 2.5 mm for a 2.0 mm ap erture reduces pixel cross illumination to less than 5% for a round aperture and to less than 7% for the square aperture. A scanned image of film exposure used to generate each plot is show n in the legend of Figure 4-8B. To ensure the film is not over exposed, several exposures we re made at 55 kVp. The x-ray current and exposure were reduced until the beam plateau dr opped below 255 (pure white in an 8 bit grey scale image). The final x-ray generator settings for the film exposures were 55 kVp, with a 1 mAs second exposure.34 Figure 4-6. YSO detector with orbit holder, new bulkhead, ne w high voltage distribution box, and modified 12 volt distribution box. Orbit holder YSO detector 12 V distribution HV distribution Bulkhead

PAGE 52

52 Figure 4-7. SXI RSD scanning system with YSO upgrade Figure 4-8. Illumination beam spot size analys is, A) Illumination beam relative aperture geometry for round and square illumina tion beams B) Beam dispersion measure 101.6 mm from aperture Round and Square Aperture Image Analysis Figure 4-9 is a picture of the layout used to scan a ramp pa nel with natural defects and debris embedded in the SOFI. The image pixel size is set at 2 mm, with a pixel dwell time of 0.1 seconds per pixel. The x-ray generator setti ngs were 55 kVp, and 45 mA with a 5.5 mm FOC. The collimators extend a total of 15 mm past the surface of the Na I detectors, and the minimum separation distance (where foam is thickest) between the SOFI and the face of the detectors was 40 mm. The foam varies in thickness from 38.1 mm to 228.6 mm. The aluminum flange runs Image pixel area Round beam Square beam 0 50 100 150 200 250 300012345678Position (mm)Grey Scale Intensity 2.0 mm Round 2.0 mm Square B A

PAGE 53

53 the length of the panel and is used to bolt together the liquid hydrogen and oxygen tanks comprising the shuttle external tank. The stiffe ner-stringers lay perpendicular to the flange located in each position where the flange is bolted together. Figure 4-9. Scanning configuration for ramp panel Figures 4-10 and 4-11 are RSD images of a ramp panel using a 2 mm round and square aperture respectively. The difference in count rate is because of the change in area of the illumination beam aperture as shown in Figure 4-8. The average count rate for the square aperture image is approximately 1.3 times the count rate of the round aperture image as expected. Two metal flanges run vertically in the image at x = 250 mm and are bolted together. The stiffeners are in the x-direction on both sides of the flange located at y = 100 mm, 275 mm and 460 mm. The stiffeners are bolted to the aluminum substrate. Glue lin es can be seen running vertically in the image at x= 125 mm and x = 375 mm. There are five dominate natural defects in the image: defect 1 (x = 60, y = 375) ; defect 2 (x = 90, y = 180) ; defect 3 (x = 450, y = 375) ; defect 4 (x = 150, y = 15) ; and defect 5 (x = 400, y = 25). Ther e are four items of foreign object debris (FOD) easily visible in the image : debris 1, tape (x = 175, y =180) ; debris 2, pencil (x = Glue Lines Tank Flange Stiffener Stringer Origin of scan images (corner behind detectors)

PAGE 54

54 250, y =375); debris 3, nylon washer (x = 350, y =375) ; and debris 4, bolt (x = 360, y = 165). The count rate in the images has a decreasing trend from bottom to top due to the decreasing thickness of the foam. The images in Figures 4-10 and 4-11 were acqui red using round and square apertures equal to the image pixel size (2 mm). This example de monstrates how aperture size can affect image contrast. The absolute signalto-background contrast for natural defect 4 is 5.2 % for the round aperture, and 4.0 % for the square aperture. The reduction in im age contrast can be partially attributed to pixel cross illumination. The illumina tion beam aperture should be slightly smaller than the image pixel size for square aperture s, and smaller or equal size for round apertures.34 Figure 4-10. RSD image, 2 mm round ap erture ramp panel image (55 kVp) Natural defect 1 Natural defect 2 Natural defect 3 Natural defect 4 Debris 3 (nylon washer) Debris 4 (bolt) Debris 2 (brush) Debris 1 (tape) Natural defect 5

PAGE 55

55 Figure 4-11. RSD image, 2 mm roun d aperture ramp panel (55 kVp) Detector Modes of Operation The University of Florida (UF) x-ray backsc atter RSD system currently uses scintillator detectors. Scintillator detect ors detect ionizing radiation thro ugh energy deposition into a crystal resulting in the excitation of electrons. As these electrons return to a gro und state, photons in the visible light spectrum are released. The number of photons created is proportional to the amount of energy deposited in the crystal. The photons are collected by a photocathode resulting in electrons passing into the photo multiplier tube (PMT ). The amplified current of electrons from the PMT can then be measured in either pu lse (counting) or current (integrating) mode. Natural defect 1 Natural defect 2 Natural defect 3 Natural defect 4 Debris 3 (nylon washer) Debris 4 (bolt) Debris 2 (brush) Debris 1 (tape) Natural defect 5

PAGE 56

56 Count Mode For count mode detectors there is a preamp c onnected to the signal from the PMT. This preamp has an RC circuit for collecting the charge from the PMT. The time constant for the RC circuit needs to be long enough to collect the phot ons and resulting charge from an interaction event, but short enough to distingu ish between individual radiation interaction even ts. If the count rate becomes too high, it b ecomes impossible to distinguish between individual radiation interaction events and pulse-p ileup occurs. This saturation can occur anywhere from the scintillator itself, through mo st stages of the electronics. Because the number of photons created in th e scintillator are proportional to the amount of the energy deposited in the crystal, and the amplification and collection process is nearly linear, the resulting height of pulses from the preamp lifier are proportional to the energy deposited. Using a count rate detector allows spectroscopy or acquisition of energy with each count. The current x-ray backscatter RSD system does not an alyze the pulse height, but simply counts. A pixels contrast is proportional to the number of counts received in each pixel. Each count is equally weighted regardless of the energy of the x-ray being detecte d. It is possible to store the energy of each count received for an image us ing a very fast multi-channel analyzer (MCA); however, a 600 cm by 600 cm image with 1 mm pixels and energy data would require approximately 5 GB of disk storage. Current Mode In situations where the detected count rate is high so that pulse pile up occurs, current (or integral) mode can be used. In current mode, pul ses are collected and inte grated over a period of time, the integration response time. The response time of the circuit is large compared to the time between individual events and as a result the ability to distinguish between individual interactions is lost. The an alog voltage/current from the inte grated signal varies with both

PAGE 57

57 detection rate and deposited energy. Typically current mode operation is only used when the radiation field is too high to count, but for RS D imaging, there is an added benefit to current mode. To incorporate spectral energy weighting into th e image contrast, either the pulse height of every individual event could be reco rded in-situ or the detector can be operated in current mode. Recording the pulse height of each individual event in-situ can be costly in terms of hardware and disk storage and even very fast multi-channel analyzers are limited in rate to around 1 Mhz. Current mode offers an alternative solution, because analog voltage from the current mode detector increases with both count rate and ener gy, higher energy x-rays wi ll cause the voltage to increase. If the varying voltage levels are mapped to contrast levels in pixels of an image, the contrast will be weighted toward the higher energy x-rays as s hown in Figure 4-12. Each of the spectral energy curves shown in Figure 4-12 has the same number of total counts under the curve. If the contrast for the pixel was genera ted using count mode, ther e would be no difference in the contrast values for the pixel for the differe nt spectra. However, in current mode, because the contrast is weighted by both energy and count rate, the spectra l curve with the higher energy x-rays has a higher contrast value in the pixel. Higher energy x-rays have a different sca ttering history compared to lower energy particles. In some cases these higher x-rays may have penetrated deeper (spectral hardening with depth penetration) or have a higher probability of interacting with th e subsurface feature of interest. If this is the case, then weighting the contrast based on ener gy can improve the ability to detect certain features.

PAGE 58

58 Figure 4-12. Energy weighted current mode contrast for a single pixel for varying spectral curves Current Mode versus Counting Mode Detectors Measurements Current mode detectors generate an analog volta ge/current output that is proportional to the count rate and energy deposition, while count mode detectors generate a voltage pulse for which the height of the pulse is proportional to de posited energy. However, for the RSD scanning system, image pixel contrast was generated base d on count rate for the counting detectors, and each count was equally weighted rega rdless of pulse height. An image data file simply consists of an array of integers (counts per pixel). The range of integer numbers is then linearly scaled to a 16-bit range (0 to 65,535) where the lowest c ount is mapped to zero and the highest count is scaled to 16-bits minus one. When a detector is operating in current mode, the RSD scanning 0.0 0.5 1.0 1.5 2.0 2.5 3.0 01020304050607080Energy (keV)Normalized Counts 38 39 40 41 42 43 44 4512PixelContrast Value

PAGE 59

59 system measures the analog voltage signal, the 0 to 5 volts signal is then linearly scaled to a 16 bit range. When measuring the an alog signal, the image contrast is affected by the energy of the radiation that is incident on the detector. As a result, the imag e contrast is weighted toward higher energy backscatter x-rays, because higher energy x-rays induce a higher voltage, but not a higher count rate. Figures 4-13A and 4-13B are SOFI calibration block images. There are two small cylindrical voids, 6.35 mm in diameter an d height, and two large cylindrical voids, 12.7 mm in diameter and height. The shallow voids in the bottom of each image are located under 50.8 mm of foam and the deeper voids are near the aluminum substrate beneath 203.2 mm of foam. Absolute percent signal contrast be tween the void and background was calculated for the 12.7 mm voids. For the large shallow flaw, the defect-to-background is -5.7 % for current and -4.0 % for pulse mode. For the large deep flaw the respective current and pulse mode contrast ratios are -3.2 % and -2.0 %. A B Figure 4-13. RSD image of SOFI calibration block, A) current mode, B) count mode

PAGE 60

60 Monte Carlo Simulation of Current versus Count Mode for SOFI The acquired RSD images (Figure 4-13) show a difference in contrast between the current and count mode for the same detector. An MCNP35 simulation was performed to identify the reason for the change in contrast between the tw o detector modes of operation. Figure 4-14 is a yz-slice of the model showing the material distri bution. The NaI detector is 5.08 cm in diameter and 5.08 cm high surrounded by a 0.1 cm thick cylindric al lead collimator. The lead collimator extends 1.27 cm below the face of the detector. The separation distance be tween the edge of the collimator and surface of the SOFI is 5.08 cm. The SOFI is 20.32 cm thick, divided into four, 5.08 cm layers, the top, Layer 1, and the bottom, Layer 4. The aluminum substrate is 0.318 cm thick. Figure 4-14. MCNP model of SO FI with aluminum substrate Three models were used: one with no void, one with the void defect located on the bottom of Layer 4 (Figure 4-14), and one with void defect located on the bottom of Layer 1. The void region was 1.27 cm in diameter and 1.27 cm high ( 0.5 inch right circular cylinder) and filled with air. The composition of the SOFI is proprietary, but was estimat ed from a 50/50 combination of isocyanate and polyol with a CFC-11 blowing ag ent. The polyol also contains a phosporphus Layer 1 Layer 2 Layer 3 Layer 4 Air Detector Collimator SOFI Void Aluminum

PAGE 61

61 flame retardant, bromine flame retardant, silicon surfactant, with an amine, tin, and potassium catalyst. The simulated SOFI was estimated to contain predominantly carbon and hydrogen, with smaller amounts of oxygen, nitrogen, fluorine and chlorine with a nominal density of 0.03 g/cm3. The source spectrum used for the simulation was 55 kVp spectrum generated from an attenuated Kramers spectrum for an electron beam on a tungsten target.13 There is no source of characteristic x-rays peaks in the 55 kVp spectrum for a t ungsten target with a beryllium window. Current mode versus count mode results A positive partial current (J+) tally with 5 keV energy bins was used to determine the backscatter x-ray spectrum ente ring the detector. Surface flagging was used to bin the J+ spectrum by depth penetration into the SOFI. Figure 4-15 shows the average spectrum entering the detector as a function of depth penetrati on. As expected the sp ectrum hardens (becomes weighted toward the higher energy) with incr eased depth penetration. The average spectrum returning to the detector from Layer 1 of the SOFI is 31.0 keV and 34.2 keV from the aluminum substrate. The void region is located just above the aluminum substrate in the bottom of Layer 4 (Figure 4-14). The detectors count and current mode respons e were determined usi ng a J+ and deposited energy tally respectively. Th e relative contrast ((signal background) / background) was determined by comparing the detectors response with and without the void defect. Table 4-1 shows little change in contrast between current and count mode for the shallow void. However, for the deep void on the aluminum substrate, a ch ange in spectrum average energy of 4.2 keV is enough to result in a change in contrast from -1.09 % to -1.27 % between count and current mode. Current mode increases the contrast of the deep void region wi thout compromising the contrast the shallow void. The modeling result s confirm the contrast trends shown in the

PAGE 62

62 experimental results of Figure 4-13. Inhomogeneities, and the s ubjectiveness of selecting the contrast region-of-interest (ROI) for the measur ed image are the major contributing factors to difference between the measured and calculated results. 30.0 30.5 31.0 31.5 32.0 32.5 33.0 33.5 34.0 34.5 13578(inches)(keV) Figure 4-15. Average photon ener gy at detector entrance win dow as a function of depth penetration Table 4-1. MCNP results compar ing relative contrast for count and current mode detector operation Shallow Percent Error Deep Percent Error (%) (1 ) (%) (1 ) Relative Contrast Current Mode -5.88 0.003 -1.27 0.003 Relative Contrast Count Mode -5.82 0.035 -1.09 0.033 NaI performance Sodium-Iodide detectors were selected for th e Lockheed RSD scanning system because of their availability. In addition to examining the effect of current and count mode on image contrast, the performance/effici ency of NaI for x-ray backscatter was also analyzed. The detector absorbed fraction, absorbed photons divi ded by incident photons, is compared to the La y er 1 La y er 2 La y er 3 La y er 4 AL Substrate

PAGE 63

63 average spectrum at the detector in Figure 4-16. The lowest detection e fficiency, 82 %, for NaI is located around the 30 keV backscatter spectral peak. Figure 4-17 is the photon cross section scattering-to-absorption ratio of NaI as a function of energy.36 Scatter in NaI crystal peaks at energies just below the k-edge absorption of iodi ne at 32 keV, severely decreasing the detectors efficiency around this energy. For low energy xray backscatter applications (under 70 kVp), detectors with iodine should proba bly be avoided. Other scintillators such as YSO have been demonstrated to be suitable replacements for NaI. 0.0 0.2 0.4 0.6 0.8 1.0 1.20102030405060Energy (keV)Fractional Total Absorbed Fraction Spectrum Figure 4-16. Normalize backscatte r spectrum at the detector en trance window compared to the fraction of incident particles per un it energy absorbed in the detector

PAGE 64

64 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16051015202530354045505560Energy (keV)Scattering / Absorption Figure 4-17. Scattering to absorption ratio as a function of energy for NaI

PAGE 65

65 CHAPTER 5 RSD COMPACT SYSTEM PROTOTYPE X-ray Tube Technology In mid-2005 YXLON released the YXLON.TU 1 00-D02 x-ray tube (Figure 5-1). This x-ray tube is about 7 cm in diameter, 26.7 cm long and weighs in at about 6 kg. The YXLON/Comet MXR 160/22 tube in comparison is 10 cm in diameter, 27.9 cm long and weighs about 8 kg. Both tubes have a 1 mm (using st andard acc. EN12543) electron focal spot (FOC) with a 640 watt rating. The smaller tube is limite d to 100 kVp, while the la rge tube is rated for 160 kVp. For most low Z applications, 100 kVp is more than adequate. There is also a difference in power rating for the larger FOC on each of the tubes. The 5.5 mm FOC is rated for 3000 W on the 160 kVp tube, and the 3.0 mm FOC on the smaller tube is rated for 1500 W. The compact YXLON.TU 100-D02 tube uses the same HV socket cable as the larger 160 kVp Comet, making the tubes intercha ngeable between systems without having to change, HV cables, HV supply, or x-ray controller units. Figure 5-1. YXLON.TU 100-D02 x-ray tube for testing of compact system components

PAGE 66

66 RSD Scanning System Compact Prototype Figure 5-5 is the assembled compact protot ype system. Although not a final design, the compact system demonstrates the feasibility of a smaller, lighter system that delivers the high quality RSD images. The entire system weighs just over 11.3 kg (25 lbs) and is about 35.5 cm long, 15 cm wide and 15 cm high. Some of the system components are listed below: 1. Bulkhead connection to relieve strain on the signal cables 2. YXLON.TU 100-D02 x-ray tube 100 kVp max output 3. High voltage and 12 volt power distribution box 4. 2.54 cm diameter, 0.635 cm thick cylindrical YS O crystal coupled to a 2.54 cm diameter photo-multiplier tube (PMT) 5. 2.54 cm by 5.08 cm, by 0.635 cm rectangular YSO crystal coupled to a 2.54 cm diameter PMT 6. Preamplifier and detector electronics box 7. Detector mounting/guide track allowing de tectors to slide along the x-ray tube 8. Illumination beam collimator tube and illumination aperture Figure 5-2. Prototype comp act RSD scanning system 1 2 3 6 8 4 5 7 1 diameter YSO detector (1 of 3) 1 x 2 YSO detector Lead collimators YXLON.TU 100D02 x-ray tube

PAGE 67

67 Compact RSD Scanning System Features As shown in Figure 5-3, the illumination beam tube is positioned interstitially between the imaging detectors. The focal spot (FOC) of the x -ray tube is located in the direction normal to the illumination beam tube aperture near the cr oss sectional mid-plane of the x-ray tube as labeled in Figure 5-3 (red dot on the side of the x -ray tube housing). The x-ray tube is bracketed to the scanning head assemble with four bolts, allowing the tube to easily be rotated or removed. Two detectors are connected to each preamp hou sing, which can slide along a slotted positioning plate, allowing the detector separation distan ce from the illumination beam to vary. Figure 5-4 demonstrates how a rectangular detector can be fastened to the PMT housing, increasing the detection area from 5.06 cm2 to 12.9 cm2, 2.6 times. Bulkhead Tube bracket X-ray tube Round YSO crystals Illumination beam tube Water cooling HV and 12V Slotted detector positioning plate Preamp housing FOC Figure 5-3. Prototype compact RSD scanning system, bottom view Rectangular YSO crystal

PAGE 68

68 Figure 5-4. Three-dimensional rend ering of the of compact system drawings for the rectangular YSO crystal holder and PMT Rectangular versus Round Shaped YSO Crystals Both the 2.54 cm diameter, 0.64 cm thick cy lindrical, and 2.54 cm by 5.08 cm rectangular YSO crystals are optically coupled to a 2.54 cm diameter R1924A Hamamatsu PMT. Each crystal is 0.64 cm thick and surrounded by a diffu se reflector on the crys tal surfaces adjacent to the crystal holder surfaces (See Figure 5-4) to increase scintillation collection efficiency. Additionally, the rectangular shaped YSO crystal has a specular reflective material on the crystal surface adjacent to the cover, because the 2.54 cm diameter PMT does not cover the entire contact surface of the rectangular crystal. Figure 5-5 is scan of the ther mal protection tiles (TPS) from the belly of the space shuttle orbital with corrosion spots on the aluminum substrate just beneath the TPS tiles. The experimental setup is shown in Figure 5-2. The RSD scan was taken at 75 kVp, 8.5 mA, 1.0 mm FOC, 1.0 mm aperture 1 mm image pixels, 0.2 s/pixels, 7 cm separation between the detector face and the su rface of the TPS tiles, and 2 cm of collimation beyond the face of the detector. The results fr om the round and rectangular YSO detectors are shown in Figures 5-5A and 5-5B, respectively. The average counts per pixel (cpp) for the round detector is 34626 cpp and 184536 cpp for the rect angular detector. The cpp increased by 5.3 times for several reasons: The detection area of the entrance window of the rectangular detector PMT Crystal Holder Cover YSO

PAGE 69

69 is 2.6 times larger than the cylindrical detector : A significant fraction of the detection area is closer to the illumination beam (higher backscatte r field). Because the collimator sleeve is a fixed length, the portion of the rectangular detector closer to the illumination beam is less collimated than the cylindrical detector. Th e decrease is collimati on is obvious with the comparison of the images in Figures 5-5A a nd B. Figure 5-5B appears noisier than Figure 5-5A because of increased signal contribution from the TPS tile (because of decreased collimation). The TPS tiles are inhomogeneous a nd need to be collimated out in order to increase the signal from the co rrosion spots (Figure 5-5A). A B Figure 5-5. Compact system RSD scan of TPS tile with corrosion on aluminum substrate with 2.54 cm diameter, A) and 2.54 cm x 5.08 cm rectangular, B) YSO crystals Illumination Beam Evaluation As shown in Figure 5-6, the compact illumina tion beam tubes are much shorter (1.9 cm and 3.8 cm) than for the Lockheed units coll imators (22.9 cm). Characterization of the illumination beam is essential to predicting the performance of the compact system. The length

PAGE 70

70 of the illumination beam tube and the x-ray tube electron focal spot (FOC) will have impact on the beam intensity and divergence. Figure 5-6. Illumination beam tubes for th e compact (1.9 cm and 3.8 cm) and Lockheed (22.9 cm) RSD scanning systems Beam Intensity The illumination beam tubes in Figure 5-6 are 22.9 cm, 3.8 cm, and 1.9 cm from top to bottom respectively. The illuminatio n beam tubes are brass, filled with lead with an opening for the illumination beam drilled out. When the illumination beam tube is affixed to the RSD scanning system, the x-ray tube FOC is about 3.6 4 cm from the opening of the beam tube. An aperture is attached to the exit of the illumination beam tube as shown in Figure 5-3. The total illumination beam tube collimator length listed in Table 5-1 is the beam tube length plus the 3.64 cm offset of the FOC. The changing count rate was measured using a 2.54 cm diameter, 0.635 cm thick cylindrical YSO detector. The cen ter of the YSO detector was 6 cm from the illumination beam, in the backscatter field, such th at the detector face was positioned in the same plane as the illumination beam aperture. A 15 cm x 15 cm x 1.3 cm nylon block was located

PAGE 71

71 about 6 cm from the illumination beam tube. The count rate in counts per second (cps) was measured for different illumination beam tube lengths and FOCs. The illumination aperture (1 mm), the x-ray tube HV (55 kV), the tube current (11.6 mA) and detector relative position were held constant so the only change in the m easured intensity was because of the illumination beam tube length and the x-ray tube FOC. There is a small increase in count rate when the FOC is changed from 1 mm to 3 mm as a result of an increase in solid angle between the illumination and exit aperture. As the illumination beam tube length increases the count rate continues to decrease. The count rate does not decrease as 1/R2, because the source is not a point source. The count rate with the 265 mm illumination beam tube is only 11.5 times lower than the count rate measured with the 55 mm tube. As shown in Table 5-1, the 1/R2 approximation can be used to estimate an order of magnitude change in intensity, but may be off by a f actor of 2 or more. It should be noted, this is not a direct beam measurement, but a measurement of the backscatter field. Because the relative geometry of the target, detector and exit remained constant, the backscatter field intensity change is directly proportional to the i llumination beam intensity change. Table 5-1. Effect of illumination beam tube lengt h on x-ray beam intensity with a 1 mm aperture, 55kVp, and 11.6 mA. The intensity drop doe s not follow a point source model (R2) FOC Total Beam Collimator Length Counts R2 R2 Counts (mm) (mm) (cps) (mm2) ratio Ratio* 1 55 15432343025 1.0 1.0 3 55 15585273025 1.0 1.0 1 75 11004415625 1.9 1.4 3 75 11165345625 1.9 1.4 1 265 134002 70225 23.2 11.5 3 265 135432 70225 23.2 11.5 Ratios are calculated by dividi ng smaller length beam tube collimator (55 mm) count values by the larger beam tube values for the respective FOCs

PAGE 72

72 Beam Dispersion Decreasing the illumination beam tube length increases the fiel d intensity, but at the same time increases illumination beam dispersion. In order to measure the beam dispersion, a film was placed 6 cm from the exit aperture on a le ad background to reduce backscatter. An illumination spot size measurement was taken fo r each electron spot size (FOC) with three different length collimators. The illu mination spots are shown in Figure 5-7. A B C D E F Figure 5-7. Illumination spot size on film 6.0 cm from the aperture, A) 1 mm FOC, 55 cm collimator, 1.0 mAs, B) 3 mm FOC, 55 mm collimator, 1.0 mAs, C) 1 mm FOC, 75 mm collimator, 1.5 mAs, D) 3 mm FOC, 75 cm collimator, 1.5 mAs, E) 1 mm FOC, 265 cm collimator, 11.5 mAs, F) 3 mm FOC, 265 cm collimator 11.5 mAs The x-ray tube exposure was increased for the longer illumination beam tube in order to keep the film exposure/dose (number of x-rays h itting the film) approximately constant for each of the measurements. This is essential because th e film measured spot size is proportional to the film exposure. The spot size on film will contin ues to grow as the exposure increases because the penumbra and backscatter will artificially en large the illumination spot. Recall that for RSD imaging the pixel dwell time is usually adjusted so that each pixel ha s at least 10,000 counts to reduce statistical noise. Typical ly, scanning with a lower count rate (less intense illumination field) implies a longer pixel dwe ll time. For RSD scanning, it is t ypical to adjust scan time to maintain the 10,000 cpp, and therefore keep the inte rrogation object exposure re latively constant. In order to make a comparison between illumina tion spot sizes, the total exposure (mAs) was

PAGE 73

73 adjusted to keep the photon flue nce and film dose approximately e qual, similar to how the pixel dwell time is adjusted as the c ount rate changes. The lowest e xposure setting for the x-ray tube and controller is 1 mAs, which is adequate to saturate the film in the illumination spot. The 55 mm illumination beam tube film was exposed at 1 mAs, the 75 mm beam tube at 1.5 mAs, and the 265 mm beam tube at 11.5 mAs. Thes e exposure times are consistent with the illumination field intensity changes measured in Table 5-1. Figure 5-8 is an illustration of the geometric configuration that can be used to calculate the expected maximum divergence of the illuminatio n beam penumbra. Given the geometry shown in Figure 5-8, the illumination spot size can be easily calculated by S = (A/F)(D+F), where F=A/(A+FOC)*(L). Each of the variables is defined as: FOC electron focal spot diam eter on x-tube anode target F optical focus location with respect to th e illumination beam aperture and illumination spot D separation distance between the aperture exit and position where the illumination spot size is measured L length of the illumination beam collimator from the x-ray tube FOC to the exit beam aperture S illumination spot diameter at a distance (D) from the exit aperture A diameter of the illumination beam exit aperture Figure 5-8. Geometric configuration to calculate illumination beam spot divergence FOC Aperture ( A ) Illumination spot (S) Optical focus (F) Illumination beam collimator len g th ( L ) Se p aration distance ( D )

PAGE 74

74 Table 5-2 is a summary of calcu lated and measured illumination beam spot sizes. Some of the spots sizes for the 3 mm FOC show a larg e penumbra, Figure 5-7B and D, indicating nonuniformity in the electron spot on the tungsten target in the x-ray tube. The large penumbra is not only because of the larger FOC diameter, but partially related to the short exposure of 1 mAs. The short exposure does not allo w the FOC to fully develop as HV is ramped up and down again. Table 5-2. Calculated versus measured illumina tion spot sizes for different length illumination beam tubes and focal spot sizes FOC Beam Collimator Length Aperture Film Distance Calculated Measured Exposure (mm) (mm) (mm) (mm) (mm) (mm) (mm) (mAs) 1 55 1 60 3.2 3.0 +/0.25 1.0 3 55 1 60 5.4 5.0 +/1.0 1.0 1 75 1 60 2.6 2.5 +/0.25 1.5 3 75 1 60 4.2 4.0 +/1.0 1.5 1 265 1 60 1.5 1.5 +/0.25 11.5 3 265 1 60 1.9 2.0 +/0.25 11.5 5.5* 270 1 60 2.4 2.5 +/0.25 11.5 Measure on 160 kVp Lockheed system To further analyze the illumination spot si ze diameter, each film was digitized using flatbed scanner with 8 bit grayscale. A hor izontal-line profile acro ss the center of the illumination beam was generated using MATLAB. Figure 5-9 shows the illumination beam spots for the 1 mm and 3 mm FOCs with illumi nation beam tube lengths of 55 mm, 75 mm and 265 mm. Because the 75 mm illumination beam tube collimator with a 1 mm FOC is typically used with the compact RSD scanning system, it is compared to the longer illumination beam tube collimator used in the Lockheed RSD scanni ng systems in Figure 5-10. The horizontal-line profile for 1 mm FOC with the 75 mm illumination beam collimator closely matches the profile obtained with the 3 mm FOC, and the 265 mm long illumination beam collimator. Based on having nearly identical beam dispersion measurem ents at 6 cm from the illumination spot, and

PAGE 75

75 having a count rate 11.5 times higher than the longer 265 mm illumination beam collimator at the same current, the compact RSD scanning system has been able to obtain the same resolution and quality image faster than the previous generati on scanners. It should be noted that while the 75 mm beam tube with the 1 mm FOC compares well with the 3 mm FOC, 265 mm illumination beam tube, the larger FOC can run at a higher cu rrent, increasing the beam intensity. The longer illumination beam tube suffers less divergence as a function of depth. Illumination beam collimator length, aperture and x-ray FOC ar e selected based on application requirements balancing beam spot and scanning speed. 0 50 100 150 200 250 300 012345678 Film Spot Diameter (mm)Grey Scale Value 1FOC-55mm 3FOC-55mm 1FOC-75mm 3FOC-75mm 1FOC-265mm 3FOC-265mm Figure 5-9. Horizontal line profile of illumination beam spot size for various x-ray focal spots (FOC) and illumination beam tube collimat or lengths of 55mm, 75 mm and 265 mm

PAGE 76

76 0 50 100 150 200 250 300 012345678 Film Spot Diameter (mm)Grey Scale Value 1FOC-75mm 1FOC-265mm 3FOC-265mm Figure 5-10. Horizontal line profile of illuminati on beam spot size for various x-ray focal spots (FOC) comparing 75 mm to the 265 mm illumination beam tube collimator

PAGE 77

77 CHAPTER 6 SNAPSHOT BACKSC ATTER RADIOGRAPHY Image Technology Introduction Detector is used in the general sense of a ny media capable of rende ring film-like x-ray images. Computed radiography (CR) plates consis t of a film-like plate w ith a phosphor coating. Electrons are excited to metastable state, and es sentially store the image until they are read by a CR reader (ACR). The ACR scans the CR plate w ith an intense red laser, and measures the output light released from the el ectron transition to a ground stat e with PMTs. CR plates have the advantage of being flexible with handling procedures simila r to film. Digital radiography (DR) utilizes detectors (typically solid stat e or scintillator) that may be coupled to complementary metal-oxide semiconductors (CMO S) arrays and associated electronics for digitizing the radiographic images real-time. DR typically consis t of a plate-like structure, that is not flexible and significantly th icker than CR plates. DR plates are on the order of centimeters thick with resolution limits currently around 100-150 microns, but offer the ability to take realtime motion images. Any detector capable of rend ering film-like x-ray images may be used. Snapshot Backscatter Radiography (SBR) The idea of snapshot backscat ter radiography is to generate a backscatter image without scanning. The detector would be placed over the obj ect that is going to be imaged (Figure 6-1). The exposure would be taken directly through the detector, and the backscatter image could be captured. Knowing the first-pass exposure, it coul d then be digitally subtracted from the image leaving only the backscatter image. This first series of experiments were c onducted using lead letters on a nylon surface (Figure 6-1A). A single x-ray exposure was then taken using a photostimuable phosphor-based image plate or CR plate (Kodak GP Digital Im aging Plate SO-170 with a VMI ACR reader)

PAGE 78

78 (Figure 6-1B). The x-ray generator focus (F OC) was positioned about 58.4 cm above the CR plate. A sample of a raw unprocessed image taken at 50 kVp, with a 2.85 mAs exposure is shown in Figure 6-2. The striping down the center of the image is caused by the film cover sleeve (Figure 6-1B). A B Figure 6-1. Snapshot backscatte r radiography setup, A) lead le tters on nylon, B) single exposure x-ray being taken through digital film Figure 6-2. Unprocessed snap shot backscatter image Striping

PAGE 79

79 While an image is present in Figure 6-2, th e first-pass of illumina tion x-rays through the CR plate creates a strong signal onto which the ba ckscatter signal is su perimposed. While the backscatter signal could be somewhat enhanced with image processing, knowing the first-past distribution, the SNR would still need to be increased to cr eate higher resolution images with more contrast. Shadow aperture back scatter radiography in creases the SNR. Shadow Aperture Backscatter Radiography (SABR) The concept of shadow aperture backscat ter radiography (SABR) is illustrated in Figure 6-3. The entire field is il luminated, similar to SBR, except an aperture is present to shape the illumination field. The aperture allows the illumination field to pe netrate the interrogation object and scatter. The backsc atter components then generate an image on the portion of the detector that is shadowed from the illumination field. Figure 6-3. Shadow aperture back scatter radiogra phy illustration X-ra y tube Illumination beam Shadow Sample ray path Detector Ob j ect Illumination Aperture

PAGE 80

80 The SABR method of SBR more e fficiently uses the dynamic ra nge of the detector. The backscatter signal is no longer superimposed on the illuminati on transmission signal. The illumination field shaping shadow aperture does not have to be in direct contact with the detector, but may be positioned a nywhere in the illumination beam. SABR Nylon Substrate Measurements The x-ray source is a Source Ray SR 115 portable veterinarian x-ray ge nerator that varies in voltage from 30 to 115 kV. The maximum singl e shot exposure is 60 mAs with about a two minute cooling and recharging cycle between shots. The object was placed on a steel table, the CR was placed over the object and the shadow aperture was positioned over the CR. The shadow apertures were cut from 1.06 mm thick sheets of lead. Figure 6-4A shows a shadow aperture which consists of a combination of lead squares and lead strips. The squares vary in size from 0.635 cm to 5.08 cm in 0.635 cm incremen ts (0.25 inch to 2 inch in 0.25 increments). The long rectangular lead strips are about 3 cm in width. Th e shadow aperture illumination separation spacing is about 1 mm or less. Figure 6-4B is an example of a uniform shadow aperture. A B Figure 6-4. Shadow aperture exam ples with A) various size lead shadows B) Uniform shadow aperture grid pattern Lead shadows Illumination aperture opening

PAGE 81

81 Figure 6-5 is a collection of various brass, aluminum and steel washers, nuts, and lead pieces (i.e. foreign object debris (FOD)) on a 6/ 6 nylon substrate. Inte grated exposure in mAs was incremented up in 10 mAs steps until the image on the CR plate reaches an adequate level of exposure in the shadow backscatter region. The amount of exposure required depends on several factors: aperture grid spacing, the size of the shadow area, the kV of the x-ray tube, the illumination objects scatter-to-absorption ratio and the mean-free-path (mfp) of scattered photons in the object. Increasi ng the illumination aperture spaci ng, kV, scattering-to-absorption ratio and mfp have a tendency to reduce the required exposure Figures 6-6 and 6-7 are SABR im ages of the target (Figure 6-5). The exposures were taken at 70 kVp, 120 mAs, and 120 cm from the x -ray tube FOC. Each of SABR images shows most of the FOD. Figure 6-6 s hows the results from a 2.54 cm s quare uniform shadow aperture configuration. Figure 6-7 is a SABR image using a shadow aperture with a variety of different geometries tested simultaneously. Figures 6-8 and 6-9 are line profiles for column 767 and row 917, respectively. Figure 6-5. Collection of washer s and lead on a nylon substrate

PAGE 82

82 Figure 6-6. SABR image of Figure 6-5 target using 2.54 cm squa re shadow aperture, 70 kVp, 120 mAs, 120 cm from x-ray FOC. Figure 6-7. SABR image of Figure 6-5 target using various dime nsion shadow apertures, 70 kVp, 120 mAs, 120 cm from x-ray FOC Row 917, Figure 6-9 Column 767, Figure 6-8 5.08 cm tile washer 1 washer 2 ACR induced lines

PAGE 83

83 Figure 6-8. Line profile of colu mn 767 shown in Figure 6-7. The right side of the image corresponds to the botto m of the line profile shown in Figure 6-7. Figure 6-9. Line profile of row 917 shown in Figure 6-7 5.08 cm tile shadow area Illumination beam entrance washer #2 washer #1 Note Red cross only indicates mouse position in image when line profile was generated

PAGE 84

84 SABR Nylon Substrate Discussion Figure 6-6 is a SABR image using a 2.54 cm unifo rm square shadow aperture grid with 1 mm illumination apertures. While most of th e FOD is visible, the relative FOD contrast compared to Figure 6-7 is discernably lower for many articles. For many of the FOD objects, the relative position with respect to the shadow aperture plays a significant role in the relative contrast. For example, washer #1 has a relative contrast of about -53 % in both figures; however, washer #2 has a relative contrast of -62 % in Figure 6-6 and -73% in Figure 6-7. In Figures 6-6 and 6-7 there is a bright region just to the ri ght of the illumination area. These bright regions in the shadow area are indu ced by the CR reader (ACR). The CR plates are fed into the ACR and read from left to right as shown in the image. The saturated illumination areas leave an afterglow in the reading process. Additional insight into the SABR process is avai lable in the line prof iles of the images. Figure 6-8 is a line profile acro ss column 767of Figure 6-7. Moving from left to right in the figure, the scatter intens ity starts very low. Just before the CR is completely saturated around 2.54 cm (1 in) there is a decrease in signal induced by the wedge-sha ped piece of lead. From the saturation peak, the backscatter signal drops off fi rst as a step function at the shadow aperture edge, then with an exponential pr ofile in the shadow region. A s econd saturation peak is located at 5.08 cm (2 inches). The area between the tw o saturation peaks is a superposition of two decaying exponentials. When the distance and si ze of the illumination aperture and shadow regions approach an optimal setting, the two deca ying exponentials give way to a region where the scatter signal is relatively flat. This large plateau region as shown between 5.08 cm (2 inches) and 10.16 cm (4 inches) is the area where the dynamic range of the film can be utilized to generate a SABR image without any image pro cessing. The average mfp for a 70 kVp x-ray spectrum based on MCNP simulations for nylon 6/ 6 is about 5.5 cm, i ndicating that optimal

PAGE 85

85 shadow aperture spacing for a nylon substrate may be around 1 mfp. It is important to remember this estimate is determined using a homogene ous substrate, with FOD located only on the surface, nearly in contact with the CR plate. Any type of structure or inhomogeneities is expected to change the optimal shadow aperture pattern. Figure 6-9 is a line profile of Figure 6-7 taken from row 917. The backscatter signal response in the shadow regions induced by washer #1and #2 start at 7.62 cm (3 inch es) and 17.8 cm (7 inches), respectively. SABR Aluminum Substrate Measurements Nylon is nearly an ideal substr ate for backscatter experiments in contrast to aluminum. For example, at 35 keV the scatter-to-absorption ration of nylon is 5.1 and aluminum is 0.3. Figure 6-10 is an aluminum substrate consisti ng of two pieces of aluminum. The aluminum plates are about 15.2 cm x 15.2 cm x 1.27 cm. The FOD consists of brass, steel, aluminum and nylon washers along with a few pieces of lead. Figure 6-11 is the resulting SABR image using the shadow aperture from Figure 6-4(A). Th e exposure was taken at 75 kVp, 240 mAs, and 120 cm from the x-ray tube FOC. Figure 6-10. Collection of washers and lead on an aluminum substrate

PAGE 86

86 Figure 6-11. SABR image of FOD on aluminum s ubstrate target (Figure 6-10) using various dimension shadow apertures, 75 kV p, 240 mAs, 120 cm from x-ray FOC Figure 6-12. Line profile of column 1261 shown in Figure 6-11. The right side of the figure corresponds to the bo ttom of line profile shown in Figure 6-11. Row 572 Figure 6-13 Column 1261 Figure 6-12 Small illumination aperture Lead wedge Aluminum plate joint

PAGE 87

87 Figure 6-13. Line profile of row 572 shown in Figure 6-11. SABR Aluminum Subs trate Discussion Figure 6-11 is a SABR image using a multi-dimens ional square shadow aperture grid with most illumination apertures being about1 mm. A line profile of column 1261, Figure 6-12, demonstrates the effect of maki ng the illumination aperture too small. The apertures is nearly closed, ~0.25 mm, (Figure 6-11) and as a result, the illumination intensity does not saturate during the 240 mAs exposure. Figure 6-13 is a line profile of row 572 in Figure 6-11. Visible in both the SABR image and line profile is the joint of the aluminum pl ate. The ability of SABR to locate the joint position indicates this technique may be suitable for imaging cracks. In Figure 6-6 and 6-7 there is a bright region just to the right of the illumination area. These bright regions in the shadow area are indu ced by the CR reader (ACR). The CR plates are Aluminum plate joint Washer

PAGE 88

88 feed into the ACR and read from left to right as shown in the image. The saturated illumination areas leave an afterglow in the reading process. SBR Radiography Lessons Learned and Failed Attempts The first SBR images suffered from having a small backscatter signal superimposed on a large illumination signal (Figure 6-2). Usi ng a mask pattern with 6.35 mm apertures on a 6.35 mm x-y pitch (Figure 6-14A) was the first atte mpt to generate a SABR image. Rectangular illumination apertures were chosen to keep the backscatter field in the shadow region as uniform as possible. However the shadow-to-illuminatio n area ratio was not favor able (Figure 6-15A). Even with the contrast adjustment, the falli ng step function followed by the rapidly decaying exponential change dominates the signal. Figure 6-14B is an attempt to adjust the sh adow-to-illumination area ratio to a more favorable condition. The pattern uses 1 mm circular illumination apertures on a 6.35 mm x-y pitch. The horizontal striping patt ern of increased intensity that is not present in the vertical direction is due to the ACR. However the d ecaying exponential signal around the aperture holes yields only a small annular region where the SABR image can be easily registered without image processing as seen in Figure 6-15B and illustrated in Figure 6-16. The progression of experiments led to the deve lopment of the square and long rectangular shadow aperture designs. The long rectangular sh adow aperture patterns shown on left side of Figure 6-4A creates a large shadow areas, where th e backscatter field is relatively uniform. The illumination and shadow pattern generated with l ong rectangular shadow ap ertures is illustrated in Figure 6-17. Items such as FOD and cracks ca n then induce a backscatter signal change that registers as an image without requi ring significant image processing.

PAGE 89

89 A B Figure 6-14. SBR mask patterns, A) large 6.35 mm aperture, B) small 1 mm aperture A B Figure 6-15. SBR exposures taken at 70 kVp, 30 mAs, 120 cm from FOC, A) Large 6.35 mm SBR image B) 1 mm small aperture SBR image.

PAGE 90

90 Figure 6-16. SABR exposure patter n for round illumination apertures Figure 6-17. SABR exposure patter n for line illumination apertures Optimal CR plate exposure Illumination aperture Large gradient exponential decay Under exposed Optimal CR plate exposure Large gradient exponential decay Illumination hole

PAGE 91

91 CHAPTER 7 SUMMARY, CONCLUSIONS AND FUTURE WORK Summary and Conclusions Radiography by selective detec tion (RSD) is a pencil beam Compton backscatter imaging technique developed at the University of Flor ida that falls between highly collimated and uncollimated techniques. RSD is orders of magnit ude faster than highly collimated techniques, but offers depth resolution that is not available in uncollimated CBI techniques. In addition to depth resolution, RSD allows fo r preferential detection of ba ckscatter compone nts that are responsible for improving image contrast. Lateral Migration Radiography (LMR) is a subs et of RSD where the image contrast is dominated by third order (and higher) scatter x-ray components that was used for landmine detection. Because typically the illumination x-ra y beam penetration barely extends to a depth much beyond the base of the mine, the mine (o r surrounding soil) becomes a diverged scattered x-ray source for the highly collimated detectors. While second order scatter components are still close to the penetrating beam, third and higher order scatter components scatter far enough from the original beam, to produce a laterally spread scatter source in the landmine (or surrounding soil). Typically, very large de tectors (on the order of 0.3 m2) are required to capture these laterally spread, multiple scatter components RSD scanning systems are currently being used by Lockheed Martin Space Systems Co. and NASA for inspection of the sp ray-on-foam-insulation on the sp ace shuttle external tank. For these systems current mode detector operation was found to significantly incr ease the contrast of deep void defects in SOFI without adversely affecting the contra st of shallow void defects. Optimization and analysis of the illumination beam tube a nd detector components of the Lockheed RSD scanning systems has led to the de velopment of a compact scanning system. The

PAGE 92

92 compact scanning system uses YSO detectors in pl ace of NaI detectors. YSO is not only faster than NaI, with more light output, but it is no t hydroscopic and more rugged. Overall the new compact scanning systems are about 60 % lighter and 60 % smaller by volume than the original systems, yet have the ability to maintain a nd/or exceed resolution and scanning speed of the previous generation of scanners. A new technique of Snapshot Backscatte r Radiography (SBR) has been developed. Shadow Aperture Backscatter Radiography (SABR) uses a collection of shadow and illumination areas designed to generate single exposure backscatter images. The SABR technique more efficiently utilizes the dynamic range of the imaging media (film, detector, CR, DR, etc.) by limiting the illumination area, and generating the b ackscatter image on the shadow regions of the imaging media. While further optimization of the technique is required, SABR has from a proofof-principle stage of development to an experimental la boratory procedure. SXI (RSD, LMR, or SABR) is a single-sided imaging technique in which the radiation source and the detection/imaging device are located on the same side of the object. SXI is a valuable non-destructive evaluation (NDE) tool because of its single-sided nature, penetrating abilities of radiation, and unique interaction properties of radiat ion with matter. This technology can be applied in many fields including NDE, me dical, security, and military applications. Future Work The most significant drawback of RSD imagi ng is still image acquisition time. Although the compact system shortened illumination beam tube may increase scanning speed by a factor of three, (beam intensity increases by a factor of 11.2, but x-ray t ube current limit decreases by a factor of 3.86 using a smaller focal spot to mainta in resolution, leaving a net gain of only about three) innovative ways of improving image ac quisition time are needed. Exploration of illumination beam and detector relative geometry could hold many of answers. An example

PAGE 93

93 would be the use of multiple illumination beams, each surrounded with an array of detectors. New image processing algorithms should be tested ; it could be possible to tolerate increased statistical noise (lower pixel dw ell time), if a post processing f ilter algorithm removes the noise. The RSD imaging processing software needs further research. The majority of RSD images are in raw data format. The limited amount of image processing, such as combining different detector images through addition and subt raction, has shown signif icant potential. The landmine application is the perfect example. Al so, there is significant embedded information in each of the detector images that needs to be pr operly correlated to enhance the image. It may even be possible to generate 3-D images with the use of an array of angled detectors, similar to laminography techniques. Current mode detector operation has been found to enhance the contrast of deep voids in SOFI using NaI detectors. This is because th e spectrum hardens with depth of penetration and current mode weights the image contrast by energy and intensity. Because of the importance of energy to deeper feature contrast this phenomenon should be explor ed for other types of defects in various materials using other de tectors such as YSO. It would be very useful to acquire the pulse height spectrum during im age acquisition, pixel by pixel. In current mode, the energy weighting is fixed, but with the acquired spectrum, pixel contrast can then be weighted by energy allowing the user to define or vary the importan ce of different energy groups. The use of energy weighting as a type of collimation should be further explored. Shadow Aperture Backscatter Radiography us ed 1/24 inch thick lead for the shadow aperture. This limits the x-ra y energy to about 70 kVp before the transmission signal (through the lead) begins to significantly reduce the SNR. Thicker lead or other materials such as tungsten should be explored for this applica tion allowing for experimentation with higher

PAGE 94

94 energy. Filtering the illumina tion field hardens and narrows the spectrum which may reduce image distortion and blurring, because as the spec trum hardens and narrows, there should be less variance in the mfp of the scat tered photons. A different x-ra y source other than the SR-115 should be used because it is only capable of 60 mAs exposures and the typical exposure for SABR with an aluminum substrate is about 240 mAs.

PAGE 95

95 LIST OF REFERENCES 1. E. Odeblad and A. Norhagen, Electron De nsity in a Localized Volume by Compton Scattering, Acta Radiologica 45 p. 161-167 (1956) 2. M.A. Kumakhov, A.F. Gamaliy, V.N. Vasiliev, M. Yu. Zaytsev, K.V. Zaytseva Scattered Xrays in medical diagnostics, Proceedings of SPIE, 59 p. 43 (2005) 3. M. Annis, M. Johnson, and R. Mastronardi, Tomographic imaging with concentric conical collimators, United State Patent, 4,825,454 (1989) 4. S. Anghaie, L. Humphries, and N. Diaz, Mat erial characterization and flaw detection, sizing, and location by the diffe rential gamma scattering spec troscopy technique. Part I: development of theoretical basis., Non-Destructive Testing and Evaluation International 28 (3) p. 192 (1995) 5. S. Anghaie, L. Humphries, and N. Diaz, Mat erial characterization and flaw detection, sizing, and location by the diffe rential gamma scattering spec troscopy technique. Part II: experiment, Non-Destructive Testing and Evaluation International 28( 3) p. 192 (1995) 6. A.D. Dougan, http://www.llnl.gov/sensor_technology/STR21.html Lawrence Livermore National Lab (LLNL) (1995) 7. S. Anghaie, Accuracy improvement for gamma-ray techniques in two-phase measurements, Ph.D. Thesis, The Pe nnsylvania State University (1982) 8. E.S. Kenny, and A.M. Jacobs, Dynamic Radiography, United States Patent 3,769,507 (1973) 9. S.H. Nellis, A.J. Liedtke, M.L. Heimer, R.A. Shiroff, A.M. Jacobs et al.,Validation of dynamic radiography in the dog and evaluation of ischemic dyssynergy, The American journal of physiology 238 (1) p. 43-53 (1980) 10. R. H. Bossi; J. L. Cline; K.D. Fridell; J.M. Nelson, One sided radi ographic inspection using backscatter imaging, Non-Destructive Testing an d Evaluation International 28 (3) p.192 (1995) 11. W.J. Baukus, X-ray imaging for on-the-body co ntraband detection Proceedings of SPIE 2932 p. 115-120 (2000) 12. A. Jacobs, and J. Campbell, Landmine detection by scatter radiation radiography, Scientific and Technical Fi nal Report, Contract DAAK 70-86-K-0016, U.S. Army Belvoir Research, Development and Engineering Center, (1987) 13. J. Campbell, and A. Jacobs, Detection of buried land mines by Compton backscatter imaging, Nuclear Science and Engineering 110 p. 417-424 (1992)

PAGE 96

96 14. Y. Watanabe, J. Monroe., S. Keshavmurthy, A. Jacobs, and E. Dugan, computational methods for shape restoration of buried obj ects in Compton backscatter imaging, Nuclear Science and Engineering 122 p. 55-67 (1996) 15. J. Wehlburg, S. Keshavmurthy, E. Dugan, and A. Jacobs, Geometric considerations relating to lateral migration backscat ter radiography (LMBR) as applied to the detection of landmines," Proceeding of SPIE, 3079 p. 384-393 (1997) 16. Z. Su, J. Howley, J. Jacobs, E. Dugan, and A. Jacobs., The discernibil ity of landmines using lateral migration radiography, Proceeding of SPIE, 3392 p. 878-887 (1998) 17. C. Wells, Z. Su, J. Moore, E. Dugan and A. Jacobs, "Lateral migration radiography measured image signatures for the detection and identi fication of buried landmines, Proceeding of SPIE, 3710 p. 906-916 (1999) 18. C. Wells, Z. Su, A. Allard, S. Salazar, E. D ugan and A. Jacobs, Suitability of simulated landmines for detection measurements us ing x-ray lateral migration radiography, Proceeding of SPIE, 4038 p. 578-589 (2000) 19. Z. Su, A. Jacobs, E. Dugan, J. Howley, a nd J. Jacobs, Lateral migration radiography application to land mine detection, confirmation and classification, Optical Engineering 39 (9) p. 2472-2479 (2000) 20. E. Dugan, A. Jacobs, Z. Su, L. Houssay, D. Ekdahl and S. Brygoo, Development and field testing of a mobile backscat ter x-ray lateral migration ra diography land mine detection system, Proceeding of SPIE, 4742 p. 120-131 (2002) 21. E. Dugan, A. Jacobs, S. Keshavmurthy and J. Wehlburg," Lateral migration radiography", Research in Nondestructive Evaluation 10 (2) p. 75-108 (1998) 22. A. Jacobs, E. Dugan, S. Brygoo, D. Ekdahl, L. Houssay and Z. Su, Lateral migration radiography: a new x-ray backscatter im aging technique, Proceeding of SPIE, 4786 p. 1-16 (2002) 23. E. Dugan, A. Jacobs, L. Houssay and D. Ekdahl Detection of flaws an d defects using lateral migration x-ray radiogra phy, Proceeding of SPIE, 5199 p. 47-61 (2004) 24. M. Gertsenshteyn, T. Jannson, and G. Savant, Staring/focusing lobster-eye hard x-ray imaging for non-astronomical objects, Proceeding of SPIE, 5922 p. 107-117 2005) 25. A.P. Hammersley, The reconstr uction of coded-mask data under conditions realistic to x-ray astronomy observations, Ph.D. Disserta tion, University of Birmingham (1986) 26. R. Accorsi and R. Lanza, Near-f ield artifact reduction in pl anar coded aperture imaging, Applied Optics 40 (26) p. 4697-4705 (2001)

PAGE 97

97 27. A.A. Faust, R.E. Rothschild, and W.A. Heindl, Development of a coded aperture backscatter imager using the UC San Die go HEXIS detector, Proceeding of SPIE 5089 p. 95-106 (2003) 28. J.R.P Angel, Lobster eyes as x-ray telescopes, Astrophysical Journal, 233 p. 364-373 (1979) 29. B.T. Addicott, Characterization and optimi zation of radiography by selective detection backscatter x-ray imaging m odality, M.S. Thesis, University of Florida (2006) 30. Glenn F. Knoll, Radiation detection and measurement, 3rd edition, John Wiley & Sons, Inc. (1999) 31. B.L. Justus, P. Falkenstein, A.L. Huston, M. C. Plazas, H. Ning, and R.W. Miller Gated fiber-optic-coupled detector for in vivo real-time radiation dosimetry Applied Optics, 43 (8) p. 1663-1668 (2004) 32. E. Dugan, A. Jacobs, D Ekdahl, C. Meng, N. Sabri, D. Shedlock, Research into the feasibility of utilizing various miniature diode s and x-ray detectors for x-ray backscatter, NASA Final Report, Award Number NNL05AF19P, (2006) 33. A. Saoudi, C.M. Pepin and R. Lecomte, Study of light collection in multi-crystal detectors, IEEE Transactions on Nuclear Science 47 (4) p. 1215-1219 (2000) 34. D. Shedlock, B. Addicott, E. T. Dugan, and A. M. Jacobs, Optimization of a rsd x-ray backscatter system for detecting defects in the space shut tle external tank thermal foam insulation, Proceeding of SPIE, 5923 p. 205-216 (2005) 35. X-5 Monte Carlo Team, MCNP A general mont e carlo, n-particle tr ansport code, version 5, LA-UR-03-1987, Los Alamos National Laboratory (2003) 36. M.J. Berger, J.H. Hubbell, S.M. Seltzer, J. Chang, J.S. Coursey, R. Sukumar, and D.S. Zucker, XCOM: Photon Cross Section Database (version 1.3) http://physics.nist.gov/xcom National Institute of Standards and Technology (2007)

PAGE 98

98 BIOGRAPHICAL SKETCH Daniel Shedlock was born and raised in the sm all city of Wilkes-Barre, Pennsylvania. He completed a Bachelor of Science in Nuclear E ngineering from Penn State University in 1997 and worked as a nuclear engineering consultant pe rforming reactor decommissioning and radioactive waste management for WMG, Inc., before return ing to Penn State University to complete a Master of Science in 2003 in th e area of radiation shielding an d transport theory. After the Columbia space shuttle accident in 2003 he began working on his Ph.D. in Nuclear and Radiological Engineering at the University of Florida in the area of Compton backscatter imaging (CBI). He helped to develop seve ral of the CBI devices based on a new imaging technique called radiography by se lective detection (RSD). Th e RSD scanning systems are still being used for inspection of the spray-on-foam -insulation on the external tank of the space shuttle to reduce the risk of another Columbia accident. In spring of 2007 he was part of the research team that invented a new CBI imag ing technique, shadow aperture backscatter radiography (SABR). SABR allows single-exposure backscatter radiographs to be taken with any film-like radiation detector without any sign ificant image processing He started his own company Advanced Nuclear Services, LLC in 2006 to support the maintena nce and research of the RSD scanning systems. In June of 2007 he jo ined NucSafe, Inc. as a senior scientist and business element manager, to quickly expedite the commercialization of the RSD scanning systems in the areas of non-destructive evaluation.