1 DEVELOPING AND IMPLEMENTING A HIGH PRECISION SETUP SYSTEM By LEE CHENG PENG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010
2 2010 Lee Cheng Peng
3 To my beloved f amily
4 ACKNOWLEDGMENTS I would like to gratefully and sincerely thank my major advisor, Dr. Chihray Liu, for his guidance, support and help His insights and suggestions are always inspiring. I really appreciate his willingness to offer prompt help at any time. His diligence, perseverance and serious attitude on both research and clinic set a good example for me. The person who se active attitude and solid research activities inspired me is my other advisor; Dr. Jonathan Li represented a successful model as the excellent researcher and clinical medical physicist. I really thank him for his continuous encouragement and support. My thanks extend to my committee members, Drs. Robert Amdur, Dietmar Siemann and Sanjiv Samant for their input and many valuable suggestions. I need to express my special gratitude to Dr. Darren Kahler and M r s. Jessica Kirwan who offer greatest efforts for all manuscripts. I need to express my special gratitude to my old colleagues, Dr. Jack Yang and Miss Yie Chen. Im thankful the most amazing friendship in my life with them. They fully support graduate study after working 2.5 years together and help me to transfer UF smo othly. The greatest support made my dream come true. I also wish to thank all physicists and fellow graduate students, Bo Lu, Guanghua Yan, Christopher Fox, Namita Thakur and Chie Kurokawa for friendship, collaboration and numerous suggestions. I would lik e to sincerely thank Dr. Ivan Meir from Vision RT Corporation for his g enerous support, and continuous technique help Most of all, I thank my parents, my sisters (Rita and Mary) brother and nephew (Max) and nieces (Lisa and Jenny) for their great unconditional support, love and encouragement. Give all the glory to the unfailing LOVE from my LORD, KING and JESUS.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................................... 4 LIST OF ABBREVIATIONS ............................................................................................................ 11 ABSTRACT ........................................................................................................................................ 13 CHAPTER 1 INTRODUCTION ....................................................................................................................... 15 High Precision Radiation Therapy (HPRT) ............................................................................... 15 ImagingGuided Stereotactic Positioning System (IGSPS) ..................................................... 16 High Precision Setup System (HPSS) ........................................................................................ 19 Problems in Current HPSSs ................................................................................................ 19 Requirements of Ideal HPSSs ............................................................................................. 20 Daily QA Programs for HPSS .................................................................................................... 21 Dosime tric Evaluation for HPSS ................................................................................................ 22 Study Aims .................................................................................................................................. 25 2 QUALITY ASSESSMENTS OF CURRENT SETUP SYSTEM ........................................... 28 Introduction ................................................................................................................................. 28 Methods and Materials ................................................................................................................ 30 System Descriptions ............................................................................................................ 30 QA Phantom Study .............................................................................................................. 31 Patient Setup Pr ocedures and Image Acquisitions ............................................................ 32 Results .......................................................................................................................................... 34 QA Phantom Tests ............................................................................................................... 34 The Patient Posi tioning Differences between FSA and CBCTs ....................................... 34 Comparisons of bFSA and mLAS Isocenter Corrections Using XVI .............................. 35 Discussions .................................................................................................................................. 35 Conclusions ................................................................................................................................. 39 3 SYSTEM CHARACTERIZ ATION OF NEW POSITIOING SYSTEM ................................ 46 Introduction ................................................................................................................................. 46 Methods and Materials ................................................................................................................ 47 System Des criptions ............................................................................................................ 47 Align RT3C Evaluation ........................................................................................................ 49 Accuracy of system calibration ................................................................................... 49 Ac curacy of registration algorithm ............................................................................. 50 Uncertainties of surface reconstruction ...................................................................... 50 Results and Discussions .............................................................................................................. 52 Accuracy of System Calibration ......................................................................................... 52
6 Accuracy of Registration Algorithm .................................................................................. 53 Uncertainties of Surface Reconstruction ............................................................................ 55 Conclusions ................................................................................................................................. 56 4 CLINICAL EVALUATION OF NEW SETUP SYSTEM ...................................................... 62 Introduction ................................................................................................................................. 62 Methods and Materials ................................................................................................................ 63 Patients, Immobilization and Treatment characteristics .................................................... 63 Imaging, Setup flow and Analysis protocols ..................................................................... 64 System descriptions ...................................................................................................... 65 Setup flow ..................................................................................................................... 66 Results .......................................................................................................................................... 68 Evaluation 1: The Isocenter Positioning Differences between Three IGSPSs at Couch and Gantry 0 ....................................................................................................... 68 Evaluation 2: The Isocetner Positioning Differences between AlignRT3C and FSA at Each Non -coplanar Field ............................................................................................. 69 Evaluation 3: Intra -fractional Motion with AlignRT3C ..................................................... 69 Discussions .................................................................................................................................. 70 Conclusions ................................................................................................................................. 74 5 DAILY QA PROGRAM OF LINAC AND THREE POSITIONING SYSTEMS ................. 84 Introduction ................................................................................................................................. 84 Methods and Material s ................................................................................................................ 86 Description of the DailyQA3TM Device ............................................................................. 86 Description of the Three IGSPSs ........................................................................................ 87 Description of QA Procedures ............................................................................................ 88 Mechanical and imaging .............................................................................................. 88 Dosimetry ..................................................................................................................... 90 Results and Discussions .............................................................................................................. 90 Conclusi ons ................................................................................................................................. 92 6 DOSE EVALUTION OF ROTATIONAL ERRORS ............................................................... 97 Introduction ................................................................................................................................. 97 Methods and Material s ................................................................................................................ 99 Patient Selection .................................................................................................................. 99 Simulation and Target Delineation ................................................................................... 100 Simulation of Patient Rotation with Respect to Isocenters ............................................. 101 Dosimetry Study ................................................................................................................ 103 Results ........................................................................................................................................ 103 Validation of Simulated Images and Contours ................................................................ 103 Target Dose Assessments and Comparisons .................................................................... 104 OAR Do se Assessments and Comparisons ...................................................................... 105 Discussions ................................................................................................................................ 105 Conclusions ............................................................................................................................... 109
7 7 SUMMARY AND CONCLU SIONS ...................................................................................... 116 LIST OF REFERENCES ................................................................................................................. 122 BIOGRAPHICAL SKETCH ........................................................................................................... 127
8 LIST OF TABLES Table page 2 1 List of relevant scanning parameters for the XVI and OBI used as SRT protocols. ......... 41 2 2 Coordinate conversions between FSA and CBCT units obtained with phantom tests. ..... 41 2 3 Comparisons of the bFSA patient isocenter corrections obtained with CBCT. ................. 42 2 4 Comparisons of the isocenter corrections for 8 bFSA and 10 mLAS patients. .................. 42 3 1 Coordinate system displacements between 3 systems explored with phantom tests. ........ 57 3 2 Comparisons of the positioning variances within 20 mm and/or 5 ........................... 57 3 3 Differences between the 3D phantom displacements for full couch rotation. ................... 57 3 4 Differences between inter and intra fractional motion over 2 min. for five volunteers ... 58 3 5 System characteristics of three IGSPSs ................................................................................ 58 4 1 Summary of characteristics of 5 SRT patients. .................................................................... 75 4 2 Comparisons of the isocenter variances at couch and gantry 0 (N=39) ............................ 75 4 3 Differences of the isocenter variances using two surface references at couch/gantry0 ... 76 4 4 Differences of the isocenter variances at all non -coplanar treatment beams (N=635). ..... 76 5 1 Isocenter displacements between systems respect to CBCT explored with DQA3 ........... 93 5 2 Comparisons of positioning variances between systems at couch 0 within 30 mm .... 93 5 3 Comparisons of the accuracy of couch angles between systems at 4 couch angles. ........ 93 5 4 Dosimetry results for daily QA procedures of the LINAC over 8 months ......................... 94 6 1 Summary of characteristics of the patients, tumors, OARs and prescribed treatments ... 111 6 2 Isocenter transforms of rotated CT image set after registration process .......................... 111 6 3 Absolute and percentage differences in volume between rotated and original OARs .... 112 6 4 Effect of rotational setup errors on target coverage in IMRT for SRT ............................. 112 6 5 Relative volume differences between rotated and original plans on sparing of OARs ... 112
9 LIST OF FIGURES Figure page 1 1 Three IGSPSs for HPRT. A) kV -CBCT, B) Frameless Sonarry, C) AlignRT .................. 27 1 2 Flow chart of the dissertation for the high precision setup system (HPSS) ....................... 27 2 1 Examples of registration process: A) Elekta XVI. B) Varian OBI ..................................... 43 2 2 Bite frame attached to QA phantom ..................................................................................... 43 2 3 Two setup for frameless SRT: A) bFSA. B) mLAS ............................................................ 44 2 4 Histogram of isocenter corrections for SRT patients after FSA guidance ......................... 44 2 5 Histogram of isocenter corrections from XVI for the bFSA and mLAS ............................ 45 2 6 Trend of isocenter corrections for 4 weeks for one SRT patient ........................................ 45 3 1 Schematic of the hardware setup for 3 IGSPSs in the treatment room ............................... 59 3 2 Two surface references. A) CT_S. B) ART_S ..................................................................... 59 3 3 Two setups. A) phantom. B) volunteers ............................................................................... 60 3 4 Definition of 8 ROI settings .................................................................................................. 60 3 5 Registration variances of the AlignRT3C during the full gantry rotation ............................ 61 4 1 Schematic of the hardware setup for 3 IGSPSs in the treatment room ............................... 77 4 2 The example of patient setup procedures for frameless SRT. ............................................. 77 4 3 Schematic outline of the imaging, setup flow and analysis protocol. ................................. 78 4 4 The examples of reference surface images. A) ART_S and B) CT_S ................................ 78 4 5 Examples of registration process for AlignRT3C in real -time modulate ............................. 79 4 6 Color display of surfaces at couch angles: reference (pink),and setup (green) .................. 79 4 7 Histogram isocenter corrections using two references at each treatment field .................. 80 4 8 Example of the intra -fractional motion of the AlignRT3C during CBCT .......................... 81 4 9 Example of the intra -fra ctional motion of the AlignRT3C for 6 coplanar beams ............. 81 4 10 Example of the intra -fractional motion of the AlignRT3C at couch 55 ........................... 82
10 4 11 Example of the intra -fractional motion of the AlignRT3C at couch 270 ........................... 82 4 12 Trend of isocenter corrections using XVI during the entire treatment course ................... 8 3 5 1 Schematic of the hardware setup for 3 IGSPSs in the treatment room ............................... 95 5 2 Bite frame attached to D ailyQA3TM device QA phantom ................................................... 95 5 3 Examples of registration process. A) CBCT, B) FSA, and C) AlignRT3C ......................... 96 5 4 Daily QA3 software provides a graphical presentation of data for each template. ........... 96 6 1 Tumor sites and shapes for the 10 cases. Green: CTV, blue: PTV (= CTV+3 mm) ....... 113 6 2 Comparison of the original (upper row) and rotated (lower row) images for case 8 ...... 113 6 3 Changes of tumor coverage in the V100Rx A) all cases, B) each case .............................. 114 6 4 Changes in volume receiving the tolerance dose. A) all OARs. B) brain stem only ....... 115 7 1 Flow chart of the conclusions for this dissertation............................................................. 121
11 LIST OF ABBREVIATIONS 1/2/3/4 D One/two/three/four dimensional AAPM American association of physicists in medicine ART Align RT3C 3D surface camera system aSi Amorphous silicon bFSA Bite plate FSA CBCT Cone beam computed tomography CCD Charge -coupled device CT Computed tomography DOF Degree of freedom DRR Digitally reconstructed radiographs DVH Dose -volume histogram EPI Electronic portal imaging EPID Electronic portal imaging device FOV Field of view FSA Fra meless sonarray system HPRT High precision radiotherapy HPSS High precision setup system IGRT Image guided radiation therapy IGSPS Image -guided stereotactic positioning system IMRT Intensity -modulated radiotherapy kV Kilo -voltage LINAC Linear accelerator m LAS Mask and laser mm Millimeter
12 MU Monitor unit MV Mega -voltage OAR Organs at risk OBI On board imaging ODI Optical distance indicator QA Quality assurance ROI Region of interest SBRT Stereotactic body radiotherapy SRS Stereotactic radiosurgery SRT Fracti onated stereotactic radiotherapy TG Task group TP Thermoplastic TPS Treatment planning system XVI X ray volume imaging
13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DEVELOPING AND IMPLEMENTING A HIGH PRECISION SETUP SYSTEM By Lee Cheng Peng May 2010 Chair: Chihray Liu Cochair: Jonathan G. Li Major: Nuclear Engineering Sciences The demand for high -precision radiot herapy (HPRT) was first implemented in stereotactic radiosurgery using a rigid, invasive stereotactic head frame. Fractionated stereotactic radiotherapy (SRT) with a frameless device was developed along a growing interest in sophisticated treatment with a tight margin and high -dose gradient. This dissertation establishes the complete management for HPRT in the process of frameless SRT, including image -guided localization, immobilization, and dose evaluation. The most ideal and precise positioning system can allow for ease of relocation, real time patient movement assessment, high accuracy, and no additional dose in daily use. A new image guided stereotactic positioning system (IGSPS), the Align RT3C 3D surface camera system (ART, VisionRT), which combines 3D surface images and uses a real time tracking technique, was developed to ensure accurate positioning at the first place. The uncertainties of current optical tracking system, which causes patient discomfort due to additional bite plates using the dental impression technique and external markers, are found. The accuracy and feasibility of ART is validated by comparisons with the optical tracking and cone -beam computed tomography (CBCT) systems.
14 Ad ditionally, an effective daily quality assurance (QA) program for the linear accelerator and multiple IGSPSs is the most important factor to ensure system performance in daily use. Currently, systematic errors from the phantom variety and long measurement time caused by switching phantoms were discovered. We investigated the use of a commercially available daily QA device to improve the efficiency and thoroughness. Reasonable action level has been established by considering dosimetric relevance and clinic f low. As for intricate treatments, the effect of dose deviation caused by setup errors remains uncertain on tumor coverage and toxicity on OARs. The lack of adequate dosimetric simulations based on the true treatment coordinates from the treatment planning system (TPS) has limited adaptive treatments. A reliable and accurate dosimetric simulation using TPS and in -house software in uncorrected errors has been developed. In SRT, the calculated dose deviation is compared to the original treatment dose with the dose -volume histogram to investigate the dose effect of rotational errors. In summary, this work performed a quality assessment to investigate the overall accuracy of current setup systems. To reach the ideal HPRT, the reliable dosimetric simulation, an effective dai ly QA program and effective, precise setup systems were de veloped and validated.
15 CHAPTER 1 INTRODUCTION High Precision Radiation Therapy (HPRT) The feasibility and clinical efficacy of stereotactic radiotherapies as high -precision radiation therapy (HPRT) such as stereotactic radiosurgery (SRS), fractionated stereotactic radiotherapy (SRT), and stereotactic body radiotherapy (SBRT) and proton radi otherapy have been demonstrated. Historically, a large single dose divided into many small fractions has been employed in cancer treatment. This is to maximize the tumor dose and to take advantage of the differential radiobiological responses between tumor and normal tissues. To achieve constant treatment accuracy for each treatment course, various fixation methods have been developed to maintain reproducible positions in patient treatment. However, the treatment with high target localization and irradiatio n accuracies has been developed to achieve HPRT. Most SRS is performed with non -coplanar radiation treatment without invasive open surgery for intracranial diseases. With the use of a reference frame attached to the patient's skull, SRS allows high precisi on for both target localization and treatment. The fixation technique of SRS is invasive and makes it difficult to perform dose fractionation; therefore, a large, single fraction of exposure is currently administered in the SRS procedure. With a single fr action stereotactic treatment, the prescribed dose can be delivered exactly to the target area. However, once the ring is removed from the patient, the relative coordinates disappear and then fractionation becomes impossible. The second fraction of radiati on could be provided by repeating the same invasive localization procedure. Clinically, this is impractical. Therefore, SRS achieves high precision but sacrifices the radiobiological differentiation between tumor and normal tissues. In the treatment of a m alignancy, the therapeutic ratio can be greatly improved by fractionation. Fractionation increases the cellular depopulation of the tumor because
16 of the re -oxygenation while reducing the damage to critical late responding normal tissues. Therefore, treatin g malignant tumors with a single fraction will result in a suboptimal therapeutic gain between local tumor control and the normal tissue late effects even for small tumors. An improved therapeutic ratio is expected with fractionated treatments. To reach H PRT using the fractionated stereotactic technique, the main concern is that the boundaries of the targets are usually close to the organs at risk (OARs). For example, because the spinal cord contains sensory and motor tracts, an overdose of radiation could lead to radiation myelitis or myelopathy. In contrast, an under -dose to the tumor could lead to disease progression. Because of the proximity of the OARs to the target, a high-dose gradient is necessary. In add ition, tumors with an irregular or concave sh ape surrounding the OARs are difficult to cover with three -dimensional (3 D) conformal plans except when practicing intensity modulated radiotherapy (IMRT). IMRT has improved the likelihood of delivering a highly conformal dose to tumors. Recently, however several clinical groups have tried hypofractionated radiotherapy using IMRT to different treatment sites with a significantly higher dose per fraction than in conventional radiotherapy. They concluded that extra -cranial radiosurgery with hypo-fractionat ion might have the ability to improve treatment outcome. Imaging Guided Stereotactic Positioning System (IGSPS) Radiation therapy in cancer management is a complex process that involves many procedures to achieve the best treatment results. However, duri ng the process, inaccuracies will occur due to a variety of procedures, such as localization of tumor volume, treatment setups, and immobilization and repositioning techniques. Routine radiation therapy is usually administered under multi -fractionated treatments. During the treatment process, tumor localization is the most error -prone factors comparing to the prescribed doses, machine output and dose calculation. Errors in localization can lead to reduced doses at the sites of tumors or to excessive doses to
17 normal tissues. Incorrect placement of the radiation field relative to patient anatomy results in the same effects described above. This misalignment primarily comes from patient movement and repositioning errors. The efficacy of the treatment could still be compromised by errors in the setup of the patient with respect to the delivered radiation. Particularly with IMRT, because of the steep dose gradients and tight margins around the tumor there is enhanced risk of excessive doses to OARs due to se tup errors and organ motion. To circumvent this is the advent of image -guided radiation therapy (IGRT), wherein imaging devices are used at treatment delivery to increase the probability that radiation is delivered as closely as possible to the original pl an. In addition, the localization method of stereotactic techniques are potentially useful when ensuring patient positioning reproducibility during each treatment fraction since the reported accuracy is higher than that of any treatment techniques applied in the generalized routine radiation therapy. A number of new image guided stereotactic positioning systems (IGSPSs) have contributed to the improved clinical use of frameless SRT. As defined in the American Association of Physicists in Medicine (AAPM) Tas k Group(TG) report 68,1 intracranial stereotactic positioning systems are used to position patients prior to precise radiation treatment of localized lesions of the brain. The drawbacks and advantages of both minimally invasive and noninvasive intracranial stereotactic positioning systems (ISPSs) are discussed in this report. Many frameless ISPSs employ those IGRT methods as IGSPSs to achieve accurate target localization in SRT.1 Although sub-millimeter accuracy can theoretically be achieved clinically with these systems, it rarely is due to difficulties in patient immobilization, system limitations in image acquisition and registration, and organ motion.
18 The practice of using ima ging treatment fields to localize tumors has been carried out for several decades. Methods such as 2 -dimensional (2D) megavoltage (MV) electronic portal imaging (EPI),2, 3 optical tracking,4 and video surface imaging5 have been developed, and some have been applied to frameless SRT. The merits and limitations of methods have been discussed in several scientific reports.1 MV EPI only uses bony structures to assess the setup deviations in 2D portal images because soft tissues are difficult to visualize in the planar projection X -ray images.6 Nevertheless, computed tomography (CT) images are capable of identifying both bony structures and soft tissues and CT is the standard reference to delineate organs and target in treatment planning system (TPS). Additionally, it is not surprising that volumetric CT registration in 6 degrees of freedom (6DOF) performs more accurately than does bony structure registration based on 2D portal images.6 There is a growing interest in CT based imaging system for 3D volumetric locali zation. The interest in CT based imaging system for 3D volumetric localization is growing. Cone -beam computed tomography (CBCT) provides a higher quality 3D image of the patient anatomy, and daily image guidance using kV CBCT (Figure 11a) has been routine ly implemented as an IGSPS.1,7, 8 However, the CBCT technique is limited to perform at a couch angle of 0 uses ionizing radiation, and cannot acquire images in real time. In contrast, optical tracking techniques (Figure 1 1b), which provide real time trac king of reflective markers affixed to the patient, have been used as an ISPS over the past decade.4 The system can markedly improve patient -positioning precision (1.1 0.3 mm9 ) by providing positioning errors in real time. The techniques are limited to t he reproducibility of the correlation of the internal target and external markers: if a rigid body system is formed between targets and markers, the system is highly accurate. Otherwise, the system accuracy decreases. This technique does not use ionizing radiation, and is not able to visualize internal anatomy.
19 High Precision Setup System (HPSS) HPRT has become an important treatment modality in the management of a wide variety of lesions. Patient setup errors may significantly affect the treatment qualities. However, during the full fractionated treatment course, internal organ movement and deformation as well as patient setup errors introduce various uncertainties to the treatment. This work does not address problems associated with internal organ motion or deformation and only concern the highprecision setup system (HPSS), which combines localization with immobilization systems. Several questions regarding the HPRT process have addressed and classified two main concerns: p roblems in current HPSSs and r eq uirements of i deal HPSSs. Problems in Current HPSSs In clinical experiences, SRT, which combines the dosimetric advantages of stereotactic precision with the wellknown radiobiological benefits of dose fractionation, is the most common example of HPRT. The optical tracking system, frameless sonnary array (FSA), has been used in SRT as an ISPS for over 10 years4 and offers high patient -positioning precision for the HPSSs9 providing positioning errors in real time. However, factors such as the reproducibilit y of bite -plate seating, patient comfort, and obtaining dental impressions can affect positioning accuracy by as much as 3 mm10 and displacement of more than 1 cm compared to MV EPID. In the absence of image guidance, most treatment techniques may be local ized daily by matching the patients skin marks associated with relatively large setup margins. To decrease setup margins, high -quality images of target precision provide an opportunity to adjust the patients position before treatment or monitor the accur acy of dose delivery throughout treatment in a process termed IGRT. CBCT, which can be used to track daily setup errors with only a small contribution to the daily dose, offers precise 3D image visualization coupled with 3D displacement information as an I GSPS.11 Moreover, user variability, region of interest (ROI)
20 selection, and the auto registration process using bony or gray -scale algorithms can lead to registration uncertainties and errors .7, 8,11,12 Negligible variability errors have been reported for automatic registration with a user -defined ROI .12 Zhang et al.13 have shown the importance of defining the correct ROI for automatic image fusion if the patient is not a rigid body. High precision im age guidance based on bony anatomy cannot account for these uncertainties. The CBCT technique is limited to perform at a couch angle of 0 and uses ionizing radiation. It is essential to perform a quality assessment of FSA using CBCT to investigate the ove rall accuracy of setup systems and comprehend the limitations to achieving HPSSs. Requirements of Ideal HPSSs No localization system is perfect or guarantees accuracy in all treatment techniques and all systems have their own limitations. For example, due to ionization radiation, CBCT cannot acquire real -time images to monitor intra -fraction motion and verify target locations at noncoplanar beams. Optical tracking is limited in the correlation among the internal target and external markers, reproducibilit y of bite -plate seating, patient comfort, and dental impression techniques. It is important to develop a new IGSPS for HPRT that applies the strengths of the current systems while eliminating their limitations. A new IGSPS, the Align RT3C 3D surface camera system (ART, Figure 1 1c), which combines the 3D surface images from three camera pods with real time tracking techniques, was developed to ensure accurate positioning at the first place. This system can: (a) be integrated in the treatment planning proces s; (b) perform as a fully automated positioning tool; (c) decrease the number of actions for positioning, such as additional bite frames or image panels; (d) monitor patient positioning in real time during the entire setup and delivery without radiation ri sk; and can (f) perform these tasks within an acceptable time
21 frame. For the greatest precision analysis, performance evaluation, and validation of ART by phantom, volunteer studies compared it to current IGSPSs. The ideal HPSS includes accurate target lo calization and reproducible immobilization. Combining ARTs real -time tracking and CBCTs true 3D imaging of internal images might be the ideal HPSS, ensuring accuracy within 2 mm of the target during the entire treatment. It is essential, however, to opti mize the efficiency and accuracy of applications by combining two systems for the setup flow and to estimate the amount of errors. Daily QA Programs for HPSS The AAPM TG40 recommends that the output of a megavoltage radiation therapy unit be checked every morning before the commencement of treatment.14 The daily check of linear accelerator (LINAC) is one of the main components of any quality assurance (QA) program. Sometimes vendors provide the instructions on how to perform the measurements and the tools found in clinical sites that can be used for performing the measurements. The users should analyze these materials and decide how to proceed or set up their own measurement protocols. Because no standard methods exist, some variations of QA methods exist b etween institutions. It would be ideal to adopt consistent sets of measurement techniques, phantoms, and criteria. The AAPM has accomplished the update to TG 40, specifying new tests and tolerances and produced the TG 142 report.15 It includes recommendati ons for QA parameters as well as their measurement frequency and acceptable criteria. With IMRT in particular creating steep dose gradients and tight margins around the tumor there is a greater risk of dose to surrounding OARs due to setup errors and orga n motion IGRT, wherein imaging devices are used at treatment delivery to increase the probability that radiation is delivered as closely as possible to the original plan, can help avoid these problems. Additionally, IGSPSs employ various methods and demon strate accurate target localization in
22 SRT1 and the results are summarized in the AAPM TG 681 report. The use of this new technology necessitates a comprehensive QA program to maintain and monitor system performance characteristics, which are established at the time of commissioning. QA phantoms and measurement methods may differ with different IGSPSs and institutions. Currently, several AAPM TG efforts15, 16 and studies17, 18 address QA issues associated with radiographic systems. It would be ideal that one QA program could be applied to multiple radiographic and non radiographic IGSPSs. In the standard QA procedure, the radiographic images of the phantom are obtained to verify the systems ability to correctly position objects in the image.17, 18 On the contrary, the non radiographic systems are limited to the correlation of the camera center and radiation delivery center.4, 5 Imaging installations in the treatment room are calibrated by vendor instructions to match the LINAC and room coordinate systems. For daily QA, therapists must routinely perform a dosimetric check of LINAC and geometry accuracy of multiple IGSPSs by repositioning different phantoms provided by vendors. Switching phantoms not only wastes time and effort, but may also cause systematic erro rs between phantoms. It is significant to design t he QA procedures using dedicated tools and phantoms that are validated with specific acceptability criteria and tolerance levels to ensure hardware and software function safely and consistently, and perform as accepted. Dosimetric Evaluation for HPSS Due to tremendous improvements in radiotherapy over the last few decades, therapy plans can be devised to precisely target a tumor, but the efficacy of radiotherapy can still be compromised by errors in the trea tment setup of the patient, which affects the delivered radiation. With IMRT in particular creating steep dose gradients and tight margins around the tumor there is a greater risk of dose to surrounding OARs due to setup errors and organ motion
23 IGRT, whe rein imaging devices are used at treatment delivery to increase the probability that radiation is delivered as closely as possible to the original plan,19 can help avoid these problems. The practice of imaging treatment fields to localize the tumor has bee n carried out for several decades and has been referred to as megavoltage (MV) portal imaging. MV portal imaging has been performed using 2 dimensional (2 D) X -ray detectors to verify orthogonal localization fields. It only uses bony structures to assess t he setup deviations in 2 D portal images because soft tissues are difficult to visualize in the planar -projection X -ray images.6 Nevertheless, CT images are capable of identifying both bony structures and soft tissues and CT is the standard reference to delineate organs and the target in TPS. Additionally, it is not surprising that volumetric CT registration in 6 DOF performs more accurately than bony -structure registration based on 2D portal images.6 There is a growing interest in CT based imaging system for 3D volumetric localization. The most common of the sophisticated image -guided approaches is the cone -beam CT scanner (CBCT) because of its complete set of 3D volume information with a high resolution and accurate image registration.68,20,21 Furthermor e, i t has been shown that the CBCT system sub -millimeter (sub -mm) and rotational setup errors can be correctly determined in 6DOF registrations, which is suitable for high -precision treatments like stereotactic radiosurgery .7,8 Although the standard treatm ent table does not allow rotational corrections, 6 corrections only can be achieved with a robotic couch,22, 23 or by adjusting the collimator and gantry angles.24, 25 The controversial conclusions and guidance on rotational errors management were also disc ussed.26 The post or near real -time verification22, 23 after corrections is essential. Additional verification time, dose, and setup accuracy could be compromised by each other. Therefore,
24 currently, most setup adjustments are applied to the translational direction only, and thus rotational positioning errors still exist throughout patient treatment. M athematically, a 2 rotational deviation induces a translational deviation of about 1 mm at a point located 3 cm from the isocenter. For elongated targets (> 5 cm long) with rotational deviations greater than 2, the translational deviation is more than 1.8 mm, which is significant when a tight margin is needed.20, 27 Guckenberger et al.20 discovered that the maximal rotational errors using kV CBCT were 5 (head and neck), 8 (pelvis), and 6 (thoracic). According to daily MVCT from 3,800 tomotherapy treatments,21 at least 5% of brain patients had more than 3 in roll rotations. Those rotational positioning errors resulted in decreased target coverage and increas ed dose to the OARs. However, m any factors might contribute to the dosimetric effect, including the simulation methods for those uncorrected errors, degree dose -gradient steepness, margin sizes, evaluation parameters, treatment techniques, and treatment sites. The quantitative correlation between the amount of rotational errors and its dosimetric consequence is not obvious and there is a wide range of dose effects.20, 27 32 For example, during simulation, the simplest approach is rotating the gantry for ro ll and couch for yaw.27, 30, 32 Yue et al.24 developed a method to implement all 6DOF corrections, which commonly was used for rotation simulations in 3 axes.28, 29 Additionally, for online corrections, both patient repositioning and plan adjustments have bee n proposed and treatment CBCT images have been used to evaluate dose impacts on setup errors .20 31 At present, it is still a challenging task to accurately delineate the tumor and organs and calculate the dose using CBCT images .33 Thus, the quantitative do simetric effects of rotational uncertainties need to be better understood.
25 Study Aims The key issue in HPRT deals with how to: 1) perform the quality assessments of current setup systems, 2) develop the new IGSPS to replace current systems 3) evaluate in clinical procedures applying the new IGSPS and suitable immobilization s 4) establish the QA program for LINAC and all IGSPSs and 5) simulate the dosimetric effects due to daily setups in TP S This dissertation initially performed a quality assessment of F SA using CBCT to investigate the overall accuracy of setup systems and comprehend the limitations in SRT. It also reports the optimization of the accuracy of applications by combining three IGSPSs. Besides localization systems, modified immobilizations wer e considered suitable for clinical implementation. To avoid removing the phantom variety and to simplify the daily QA procedure, a simple yet comprehensive daily QA program was designed by the commercial device for LINAC, CBCT, FSA and ART systems. Finally a reliable and accurate dosimetric simulation was developed to evaluate dosimetric consequences for those uncorrected errors,. The doses actually delivered to the targets and OARs were determined Standard dose -volume -histogram ( DVH ) was built in TPS. S pecific aim 1 : Quality assessments of current setup system (Chapter 2). The change in displacement reported by CBCT images will be used to evaluate the accuracy of our current frameless SRT setups. Specific aim 2 : System characterization of new positionin g system (Chapter 3). The 3D surface imaging system (ART) was developed to replace the current optical tracking system and validated by comparing to the FSA and kV -x ray volume imaging(XVI) (CBCT). Specific aim 3 : Clinical evaluation of new setup system (C hapter 4). The ideal setup system for HPRT was designed by a combination of ART and XVI as localization and modified mask system as immobilization. The clinical experiences from five SRT test patients were reported.
26 Specific aim 4 : Daily QA program of LIN AC and three positioning systems (Chapter 5). To avoid removing the phantom variety and to simplify the daily QA proce dure, a simple yet comprehensive daily QA program was designed by the commercial device for LINAC, CBCT, FSA and ART systems. Specific ai m 5 : D ose evaluation of rotational errors (Chapter 6). The direct simulation in TPS for systematic rotational setup errors was proposed. Simulated dose distribution with rotational errors was evaluated by DVH.
27 A B C Figure 1 1 Three IGSPSs for HPRT. A) kV CBCT, B) Frameless Sonarry, C) A lignRT Figure 1 2. Flow chart of the dissertation for the high precision setup system (HPSS)
28 CHAPTER 2 QUALITY ASSESSMENT S OF CURRENT SETUP SYSTEM Introduction Modern radiotherapy techniques, such as IMRT, SRT and SRS deliver radiation doses with a high degree of target conformation. Steep gradients occur at the periphery of the target and between the target and OAR. The higher the degree of conformation, the gre ater the setup accuracy that is required to reduce the risk of a target geographic misses with resulting high OAR doses. Modern TPS are able to generate plans with a high level of target conformation and OAR sparing. However, treatment accuracy may be comp romised if the TPS setup margins do not adequately account for uncertainties in the patient setup. ISPSs employ various methods to achieve accurate target localization (SRS / SRT)1 Each has demonstrated an acceptable level of accuracy and precision and the results are summarized in the AAPM TG 681 report. ISPSs are divided into two categories: (1) minimally invasive systems and (2) noninvasive systems. Overall, the minimally invasive systems are considered to be optimal for SRS with better than 1 mm ach ievable accuracy, due to the rigid fixation of the head ring to the patients skull. The noninvasive systems used for SRT combine the dosimetric advantages of stereotactic precision with the well known radiobiological benefits of dose fractionation. Accura te and reproducible daily patient positioning utilizing a noninvasive ISPS is essential for SRT. Positioning accuracies and uncertainties for various ISPSs have been examined using different measurement techniques, s uch as 2D orthogonal films, EPIDs, fluor oscopic video images, repeated CT or MRI scans, optical tracking (with markers and / or 3D surface images), and 3D CBCT. The advantages and disadvantages of these techniques are also discussed in the AAPM TG 68 report.1
29 Immobilization devices used with ISP Ss to reduce inter and intra -fractional motions include thermoplastic (TP) head masks, scotch-cast masks, and bite blocks. The accuracy and precision of patient positioning for stereotactic radiotherapy has been discussed in detail .4,7, 8,34,35 Wide ranges of repositioning accuracies (0.5 mm to 4.0 mm) for combinations of various immobilization and localization schemes have been reported .9,11, 3638 Solberg et al.39 reviewed many frameless technologies and emphasized the importance of user quality assessments in the development of high -precision, hypo -fractionated treatment regimens to study system limitations and to quantify setup error. The FSA system has been used in the treatment of SRS and SRT patients in our clinic for over 10 years.4 The system can mark edly improve patient positioning precision (1.1 0.3 mm9) by providing positioning errors in real time. However, factors such as the reproducibility of bite plate seating, patient comfort, and dental impression technique can affect positioning accuracy by as much as 3mm.10 We have seen displacements of more than 1 cm using conventional 2D MV orthogonal portal verification images after FSA -guided setup on the first day of treatment. CBCT, which can be used to track daily setup errors with only a small contri bution to the daily dose (~0.1cGy40 ), offers the obvious advantage over 2D orthogonal portal imaging of precise 3D image visualization coupled with 3D displacement information .7,11, 12,18 Integration of CBCT into the clinical practice of SRT will greatly b enefit setup accuracy as well as increase the therapists confidence level due to the visual ve rification of internal anatomy. We have used CBCT to perform a quality assessment for SRT patient treatments. Phantom tests were also performed to investigate th e system di fferences between CBCT and FSA.
30 Methods and Materials System Descriptions The FSA is a two 2D CCD infrared camera system (Bova et al. 19974 ). The patients head is immobilized using a mask attached to an adjustable head holder. A custom bite plate with two non reflective and four passive reflective spherical markers is affixed to the patients maxillary dentition to form a rigid system. The non reflective markers are used for the isocenter registration. The patients position is determined through registration of the reflective markers positions with respect to isocenter, tracked in real time. The software calculates the mean registration er ror with respect to the planning CT isocenter. The results are displayed and recorded as three translational and three rotational values. The quality assessment was performed using two CBCT systems: the XVI of the Elekta Synergy (Elekta Oncology Systems, Norcross, GA); and the OnBoard Imager (OBI) of the Varian Trilogy (Varian Medical Systems, Palo Alto, CA). The scanning parameters used for both CBCT systems have been previously reported .18,40,41 Each system consists of an amorphous silicon (aSi) detecto r and a kV xray source that produces diagnostic quality x rays (70 150 kVp for OBI and 40 125 kVp for XVI). The active area of the aSi detector at the nominal detector to focal spot distance is 41 41cm2 (at 155 cm) for XVI and 39.7 29.8 cm2 (at 150 cm ) for OBI. For system calibration, the FSA calibration jig is aligned with the LINAC isocenter using the room lasers, whereas XVI determines the isocenter using eight 10 x 10 cm2 MV portal images for a combination of four gantry and two collimator angles. Field edge detection is applied and the center points for the eight fields are averaged to determine the XVI origin / isocenter. The ma nufacturer specification of kV CBCT and MV isocenter coincidence is A similar procedure was used for the OBI isoc enter calibration.
31 Planning CT images were acquired with a 2 mm slice thickness. The XVI scans were acquired using a 27.6 x 27.6 x 20 cm3 field of view (FOV) and a gantry rotation of 205 over approximately 70 seconds. The XVI registration ROIs, shown in F ig ure 2 1(a), included the entire skull and excluded the neck area. The softwares bony matching algorithm was used to perform the registration. For OBI, the bow -tie filter and full -fan data acquisition geometry were used to reconstruct a 24 x 24 x 15 cm3 FOV for a 360 gantry rotation over approximately 60 seconds. As mentioned above, only soft tissue fusion without registration ROI selection is allowed for the OBI scans. The translational and rotational displacements for each system we re obtained after re gistration. After CBCT imaging, registration with the planning CT is performed. The XVI software (v4.0) supports both bony and soft -tissue fusion and allows for a user -defined ROI fusion volume for the registration. The OBI software (v2.0) offers only soft -tissue fusion with no registration ROI selection. Figure 2 1 displays the image registration screens. We represent the directions of the CBCT registration displacements along patients' left -right, superior -inferior and anterior posterior directions as LA T, LONG and VERT, respectively. The XVI software displays these displacement values to tenths of a millimeter, whereas the OBI software provides only integer millimeter displacements. QA Phantom Study A QA phantom was used to assess the precision, accura cy and system differences for FSA and CBCT. The reproducibility of geometric calibration and registration algorithms have been assessed .4,18, 41 The QA phantom was utilized daily, monthly and bi annually to evaluate the accuracy of each ISPS with respect to the Linac radiation isoce nter. The phantom, shown in Figure 2 2, is a custom polystyrene cubic phantom attached to a rigid frame. It was used to
32 evaluate the FSA determined isocenter position using CBCT as well as to determine the variances in isocenter c orrections. To simulate a typical patient setup, planning CT scans of the phantom were transferred to the FSA and CBCT. The phantom was then positioned using FSA to along all axes. CBCT images were then acquired and registered with the planning CT. This was done ten times to obtain the radiation isocenter displacement values for FSA re lative to CBCT (XVI and OBI). The mean value was used as the offset for patient localization analysis. Patient Setup Procedures and Image Acquisitions The quality assessment was performed for two groups of patients. For the first group, a bite plate and half -face TP mask were used as immobilization devices and FSA was used in real time tracking mode for localization. We termed this group the bite plate F SA (abbreviated bFSA) group. For the second group, a full -face TP mask was utilized for immobilization and room laser alignment was used for localization. We termed this group the mask laser (abbreviated mLAS) group. Patients who could not tolerate the bit e plate comprised the latter group. The setups for t he two groups are shown in Figure 2 3. Two studies were done. The first utilized FSA and both CBCT systems for 15 bFSA patients (8 XVI and 7 OBI) to determine differences in isocenter corrections between the FSA and CBCTs. For FSA, it is assumed that the bite plate and skull form a rigid -body system through the palate and upper teeth. In some cases, a poor dental impression causes bite plate position inconsistencies during the CT simulation and the treatme nts. We have seen greater than 1.5 cm positioning differences between the CT simulation and the treatment. We have addressed this issue by reseating the bite plate ten times before the planning CT images are obtained. We ensure that the reproducibility is within 2 mm.
33 In the second study, we used only XVI for 18 patients, 8 bFSA and 10 mLAS, to assess the positioning accuracy differences between the two patient groups. For the bFSA group, FSA was used for daily localization, and we acquired CBCT images for the first three fractions of bFSA patient positioning to verify that the isocenter correction was less 2 mm / 2 The ROI volumes were verified in the XVI software by the physician before the first CBCT acquisition. Weekly CBCT imaging was done for the rem ainder of the treatment course. The assumption is that if the differences between FSA and CBCT remain stable for the first three fractions, then weekly CBCT verification should be sufficient for the remainder of the treatments. Our standard protocols for S RT CBCT acquisition are listed in Table 2 1. Automatic registrations were used to eliminate user variability. Each patients position was adjusted until the FSA registration error tolerances were met, after which CBCT was performed and the patient was trea ted. In our clinic, registration error tolerances of rotation4, 35 along any axis must be met before treatment can be delivered. Note that CBCT images for these patients were acquired only for verification and quality assess ment purposes. Also note that the residual errors in the patient position accruing during CBCT acquisition and/or treatment delivery were monitored for by the FSA to ensure that the changes in the translational and rotational displacements about any axis remained less than 1 mm and 1 respectively. For the mLAS patients, CBCT alone was used. After registration, the displacement values for all three axes were recorded.
34 Results QA Phantom Tests The isocenter displacement results for the FSA system, with respect to the two CBCT systems are shown in Table 2 2. The mean differences between the XVI and FSA systems were 1.0 mm ( LAT), 0.7 mm ( LONG), and 0.5 mm ( VERT). Similar results between the OBI and FSA systems were observed. The maximum difference be tween the CBCT and FSA systems along any axis was 1.0 0.3 mm. The maximum rotational difference about any axis was less than 1 The system differences are mainly due to the uncertainty in the calibration procedures and reference methods (mechanical for FSA vs. radiation for CBCT). Previously estimated standard deviations for the FSA and CBCT calibrations were on the order of 0.6 mm in each direction .4, 18,35, 41 The slightly larger values found in this study reflect the inherent uncertainty introduced by using the combined systems. The system differences shown in Table 2 2 were subtracted for the patient setup evaluation described below. The Patient Positioning Differences between FSA and CBCTs Figure 2 4 shows a histogram of the distribution of the CBC T isocenter corrections for 15 SRT patients. A total of 65 XVI sessions and 53 OBI sessions were analyzed. The horizontal axis is the difference between FSA and CBCT, and the vertical axis is the distribution of differences as a percentage of total CBCT se ssions. The Gaussian distribution is shown for both CBCT sets with respect to FSA. Approximately 5% of the disagreements were lager than 3 mm. The variances in isocenter corrections are summarized in Table 2 3. The mean and standard deviation of displacements between FSA and CBCT were approximately 1.2 0.7 mm, and were independent of the CBCT system used. A Student t test showed that there was no difference between the distributions for XVI and OBI (p=0.208). The maximum displacement between FSA
35 and XVI was 4.4 mm, and the maximum variance between FSA and OBI was 3.2 mm. The 90th percentile of the rotational differences between FSA and CBCT along the three axes was approximately 1. The maximum rotational difference was approximately 2 for both XVI and OBI. For XVI, the maximum differences between the bony and soft -tissue registration results were under 0.5 mm translation and 0.5 rotation for all axes. Comparisons of bFSA and mLAS Isocenter Corrections Using XVI A summary of isocenter corrections using XVI for 18 bFSA and mLAS patients is shown in Table 2 4 and the histogr am distribution is shown in Figure 2 5. The displacements for 8 bFSA patients (65 CBCTs) were 0.5 0.4 mm ( LAT), 0.8 0.6 mm ( LONG), and 0.5 0.4 mm ( VERT), giving a vector value of 1.2 0.7 mm. The 90th percentile was 1.8 mm. For the 10 mLAS patients (70 CBCTs), the displacements were 2.0 1.6 mm ( LAT), 1.4 1.2 mm ( LONG), 1.3 1.0 mm ( VERT), for a vector value of 3.2 1.5 mm. The 90th percentile was 5 mm. The mean of t he errors were statistically significant (p< 0.005). The median and 90th percentile differences for the two groups were also statistically significant (p~0.001) and the maximum difference in displacement between these two groups was approximately 4 mm (4.4 mm vs. 8.1 mm). The rotational errors were larger for the mLAS group (p~0.001). The 90th percentile of the rotational errors for all distributions was approximately 1 for the bFSA group and approximately 3 for the mLAS group. The maximum rotational error was 5.7 Discussions We previously performed clinical SRT patient localizations using daily FSA guidance and weekly orthogonal MV portal images. Visual comparisons of the portal images, which suffer from relatively poor contrast, with digitally reconstructed radiographs (DRRs) were used to determine the accuracy of isocenter corrections. Using this procedure, the uncertainty can easily
36 be as high as 5 mm and 50 with large user variability. Much more accurate quantitative localization can be achieved with newer CT -based technolo gies such as CT on -rails, and cone beam CT. These technologies allow for image registration using 3D anatomical information. They also possess high quality visual verification tools and give translational and rotational isocenter corrections. We use TP ma sks for patients who are not suitable candidates for bFSA. Baumert et al .42 showed that the repositioning accuracy of SRT is not ideal, with a maximum deviation of up to 7 mm in one direction using sequential CT scanning. We detected approximately the sam e maximum deviation (up to 8mm and 5 ) for mLAS patients. We have developed a positioning technique for SRT that combines FSA, to provide detection of intra -fractional patient motion, with CBCT, which provides internal anatomical information. The system di fferences have been determined using a QA phantom. The differences include calibration methods and image registration uncertainties. The registration error for the FSA system was 0.3 mm4,9, 35 and the system isocenter variation, verified daily with the cal ibration jig, was 0.6 0.2 mm. The uncertainty of the kV and MV isocenter coincidence has been found to be within 0.3 0.2 mm18, 41 and this was verified with our routine QA over a three month period. This defines the relationship of the radiation isocen ter of the LINAC with respect to the imaging isocenter of the CBCT. The system differences between FSA and CBCT are very reproducible and therefore these differences are used as displacements between the two systems. Figure 2 6 shows the trend of isocenter corrections from CBCT sessions for one SRT patient from the first study over the entire treatment period of four weeks. The first three corrections correspond to the first three fractions and the last three corrections are for the weekly
37 CBCT acquisition. The corrections increase from approximately 1 mm for the first three fractions to 4.4 mm for the last weekly CBCT acquisition. These results indicate that the FSA system alone cannot guarantee the repositioning accuracy over the whole course of treatment and that conjunction of FSA with CBCT is needed. However, CBCT usage will add an additional 2 to 4 minutes to the treatment time as well as additional radiation dose to the normal tissues. Also, the patient setup position can be verified only for a 0 co uch angle. The two CBCT systems utilized for our study are widely used, and many studies have reported reproducibility of the calibration and registration algorithms7,18, 41 of up to 0.2 mm with phantom tests. Similar results were found in this study. Moreover, user variability, ROI selection, and the auto registration process using bony or gray -scale algorithms can lead to registration uncertainties and errors .7,8,11, 12 Negligible va riability errors have been reported for automatic registration with a user defined ROI .12 Zhang et al.13 have shown the importance of defining the correct ROI for automatic image fusion if the patient is not a rigid body. Guckenberger et al.12 reported a highly significant correlation (r 0.88) between automatic bony matching and grey scale registration for brain patients whose planning and treatment tumor sizes were not significantly different. However, a decrease in the tumor size from 20 mm to 14 mm a ltered the position of the tumor center by 2.8 mm12 in two patients. High -precision image guidance based on bony anatomy cannot account for these uncertainties, and the discussion of this effect is not in the scope of this work. We selected the whole skul l as the ROI when using XVI bony registration since, in theory, the bite block forms a rigid body with the skull through the upper jaw. XVI bony and gray scale registrations with pre -defined ROIs were processed for each SRT, and the maximum difference
38 betw een the two registrations was less 0.5 mm. Table 2 3 shows that similar corrections (~1mm) were found for both OBI and XVI for both the phantom and patients even though the OBI utilizes only gray scale registration without a predefined ROI selection. Additionally the display of sub-mm isocenter corrections, which is available only with the XVI software, is essential to high precision radiotherapy. The reproducibility of patient positioning using FSA was previously reported to be within 0.4 to 2.3 mm .4,35, 38, 43 The mean corrections obtained in our study is within this range, 1.2 0.7 mm (XVI / OBI). Many studies have been published that evaluate the positional accuracy for different frameless procedures .3,8,11,12,36,37, 39, 42 However, even if the same veri fication method is employed, results can vary greatly, mainly because there are many influencing factors (e.g., different numbers of patients / image sets, mask type, mask making proficiency, patient compliance etc.). Table 2 4 shows the results for the tw o setup techniques. Considerably smaller setup errors were obtained for the bFSA patients (1.1 0.5 mm) than for the mLAS patients (3.5 1.5 mm), with statistical significance (p<0.0001). Masi et al .8 reported setup errors for SRT patients using CBCT and found slightly smaller values (2.9 mm vs. 3.2 mm) for bite block and mask than for mask only without statistical significance. Nevertheless, for the mLAS setups, our errors (3.5 mm) are identical to tho se obtained by Masi et al .8 and slightly smaller than those of Boda -Heggemann et al.11 (4.7 mm) and Guckenberger et al.12 (4.6 mm), who all used XVI to evaluate the setup accuracy of frameless SRT. This reflects the institutional dependence of patient setu p. Rotational displacement can be a significant source of error. In general, patient discomfort leads to more rotational error. Weve estimated that a 2 rotational deviation will induce a translational deviation of about 1 mm at a point located 3 cm away from isocenter. However, for
39 large targets (> 5cm length) with rotational deviations greater than 2, the translational deviation is more than 1.8 mm, which is significant when a small setup error margin is needed. In our study, the mean rotational deviati on in any direction was less than 1 for the bFSA patients, which matches Baumerts et al.42 results for patients with bite block using sequential CT scanning. By contrast, for the mLAS patients, the mean rotational deviation was approximately 1.4 which is slightly larger than the results reported by Jin et al.3 (~0.6 ) obtained using orthogonal portal images with 6D fusion methods. Registration with CBCT is more sensitive in detecting rotational error than 2D / 3D registrations using orthogonal images. H owever, the 90th percentile of the distribution for the mLAS patients was approximately 3 and the maximum rotational error for the mLAS patients was 5.7 This would cause positional displacements of more than 5 mm if the maximum tumor length was greater than 5 cm. The rotational deviation was significant for the mLAS patients receiving SRT with non -coplanar beams. Conclusions A quality assessment of frameless SRT using CBCT was performed. The inherent uncertainties between FSA and CBCT include calibration procedure, reflector tracking versus image acquisition, registration algorithm and dental impression technique. Overall, the system difference between FSA and CBCT using a QA phantom was approximately 1.3 mm, which is due to the different calibration meth ods. The mean and the maximum differences in patient setup error between FSA and CBCT were 1.2 mm and 4.4 mm, respectively. The mean displacements using CBCT were 1.2 mm and 3.2 mm for the bFSA and mLAS groups, respectively. Note that for the bFSA group, t he systematic differences of Table 2 2 were subtracted for the mean displacement analysis. No significant difference in corrections was observed between XVI and
40 OBI. Using the FSA system alone will not guarantee the accuracy of the SRT program even when our criteria of acceptable reproducibility for the first three fractions are met.
41 Table 2 1. List of relevant scanning parameters for the XVI and OBI used as SRT protocols. Table 2 2. Coordinate conversions between FSA and CBCT units obtained with phantom tests. XVI OBI Trans(mm) Rot() Trans(mm) Rot() LAT 1.0 0.3 0.5 0.1 1.0 0.3 0.9 0.3 LONG 0.7 0.2 0.5 0.1 0.8 0.4 0.1 0.1 VERT 0.5 0.2 0.1 0.1 0.8 0.4 0.2 0.2
42 Table 2 3. Comparisons of the bFSA patient isocenter corrections obtained with CBCT. Table 2 4. Comparisons of the isocenter corrections for 8 bFSA and 10 mLAS patients. Translations(mm) XVI OBI Mean S.D. Median 90th % Max Mean S.D. Median 90th % Max LAT 0.5 0.4 0.4 1.0 2.6 0.6 0.6 0.6 1.0 2.7 LONG 0.8 0.6 0.6 1.5 2.8 0.8 0.7 0.6 2.0 3.0 VERT 0.5 0.4 0.3 1.0 2.2 0.4 0.3 0.4 0.8 0.9 Displacement 1.2 0.7 0.9 1.8 4.4 1.3 0.8 1.1 2.4 3.2 Rotations ( ) XVI OBI Mean S.D. Median 90 th % Max Mean S.D. Median 90 th % Max LAT 0.4 0.5 0.3 0.8 2.3 0.6 0.4 0.6 0.9 2.0 LONG 0.3 0.3 0.3 0.7 1.4 0.3 0.5 0.1 0.7 1.9 VERT 0.5 0.5 0.4 1.2 2.3 0.3 0.4 0.2 1.0 1.4 Translations (mm) bFSA mLAS Mean S.D. Median 90 th % Max Mean S.D. Median 90 th % Max LAT 0.5 0.4 0.4 1.0 2.6 2.0 1.6 1.7 4.4 5.8 LONG 0.8 0.6 0.6 1.5 2.8 1.4 1.2 1.0 3.3 5.5 VERT 0.5 0.4 0.3 1.0 2.2 1.3 1.0 1.2 2.1 6.8 Displacement 1.2 0.7 0.9 1.8 4.4 3.2 1.5 3 5 8.1 Displacement Masi et a l.6 2.9 1.3 N/A 4.5 7 3.2 1.5 N/A 5 6.2 Rotations ( ) b FSA m LAS Mean S.D. Median 90 th % Max Mean S.D. Median 90 th % Max LAT 0.4 0.5 0.3 0.8 2.3 1.1 1.2 0.7 2.9 5.7 LONG 0.3 0.3 0.3 0.7 1.4 0.9 0.8 0.6 2.1 3.3 VERT 0.5 0.5 0.4 1.2 2.3 1.5 1.1 1.4 2.8 4.9
43 A B Figure 2 1. Examples of registration process: A) Elekta XVI. B) Varian OBI Figure 2 2. Bite frame attached to QA phantom
44 A B Figure 2 3. Two setup for frameless SRT: A) bFSA. B) mLAS Figure 2 4. Histogram of isocenter corrections for SRT patients after FSA guidance
45 Figure 2 5. Histogram of isocenter corrections from XVI for the bFSA and mLAS Figure 2 6. Trend of isocenter corrections for 4 weeks for one SRT patient
46 CHAPTER 3 SYSTEM CHARACTERIZATION OF NEW POSITIOING SYSTEM Introduction SRS is a single -fraction intracranial treatment technique that uses a rigidly attached frame for stereotactic target localization and patient immobilization. Issues with the technique such as normal tissue dose tolerance led to the development of fractionated SRT which is a conformal treatment technique that is used for a variety of both intracranial and extra cranial lesions. For intracranial SRT, the frame based technology is impractical, and over the last decade image guided (IG) techniques have been develo ped to improve the accuracy and efficiency of target localization to allow for the elimination of the frame entirely .7, 8 Over the past decade, IG techniques have been developed to increase the precision of target localization. These include standard MV EPI ,2, 3 optical tracking,4, 35 video surface imaging,44, 45 and CBCT7,8,17, 46 Some of these techniques have been applied to frameless SRT as IGSPSs The merits and limitations of these methods have been discussed in several reports .1 For MV EPI, translational a nd rotational displacements are obtained from bi -planar x -ray images.2 However, this technique suffers from significant ambiguities due to poor image quality. CBCT provides a higher quality 3 D image of the patient anatomy, and daily im age guidance using kV CBCT as an IGSPS has been routinely implemented.1,7, 8 However, the CBCT technique is limited to a treatment couch angle of zero degree, uses ionizing radiation, and cannot acquire images in real time. Optical tracking techniques, which provide real time tracking of reflective markers affixed to the patient, have been used in many clinics over the past decade.4, 35 The techniques are limited by the reproducibility of the correlation of the internal target with the external markers; if a rigid
47 body system is formed between targets and markers, the system is highly accurate. Otherwise, the system accuracy suffers. This is the first report to discuss the use of the AlignRT3C (ART) system (Vision RT Ltd, London, UK) utilizing three ceiling mounted 3 D camera po ds and real time tracking mode. The purpose of the study was to validate the ART as a new IGSPS by comparing it to both a kV CBCT system and an optical tracking system. The CBCT system used for the study was the XVI system (Elekta Oncology Systems, Norcros s, GA) and the optical tracking system utilized was the FSA (Zmed/Varian, Inc., Ashland, MA). All three systems, ART FSA, and XVI (shown in Figure 3 1 ), are installed in the same treatment room with an Elekta LINAC (Elekta Oncology System, Norcross, GA). Our experiments were done to study: (1) the accuracy of the system calibration (2) the accuracy of the registration algorithm; and (3) uncertainties of surface reconstruction caused by inter and intra -fractional motion, skin tone, room light variances, camera thermal effects, and ROI selection. Methods and Materials System Description s The original Align RT system was introduced by Bert et al.5 The system utilized two side camera pods and was designed to acquire 3 D surface images with a 00 couch angle. Par tial image loss occurred with the system at some gantry angles due to gantry blockage of the cameras. Because the camera blockage could compromise system accuracy, the system was redesigned to include a central camera pod (Align RT3C), as shown in Figure 3 1, to ensure complete image acquisition over the full gantry and couch range. Several stereoscopically arranged charge -coupled -device (CCD) cameras were housed for different purposes, such as computing the 3 D surface image and providing graylevel and dynamic images. Camera calibration is performed using the light field and room lasers. The reproducibility of the
48 calibration is within 0.8 mm.5 Daily calibration verification is recommended by the manufacturer. All acquired images from the 3 camera pods are reconstructed as a 3 D surface with approximately 1 to 3 mm spacing and 10,000 points. Either static or continuous real time surface image displays are supported. Our study was done in the real -time mode, for which the surface image capture frame rate is 0.1 to 0.3 frames per second, depending on the size of the ROI selected for the registration. Additionally, the software (version 4.2) supports two types of surface image references; skin contours reconstructed from p lanning CT images (CT_S) and optical s urface images previously recorded by the ART (ART_S). These are shown in Figure 3 2. The FSA system is a two CCD infra -red camera system.4 A custom bite plate with 2 nonreflective and 4 passive reflective spherical markers is affixed to the patients maxillary dentition to form a rigid system. The non reflective markers are only used for the isocenter registration. The patients position, w hich is determined through registration of the reflective marker positions with respect to the isocenter, is tracked by the system in real time. Calibration of the system is performed using a vendor -supplied calibration apparatus. The spatial accuracy of t he system is within 0.8 mm4, 38 and the mean registration error is approximately 0.3 mm.4, 35 The XVI system46 consists of an aSi detector and a kV x-ray source installed on two retractable arms. The kV x ray beam is orthogonal to the LINAC treatment beam. T he unit produces diagnostic quality x rays from 70 to 150 kVp. The active area of the aSi detector is 41 41 cm2 at the nominal detector to -focal spot distance of 155 cm. The scanning parameters and calibration technique used for the XVI have been previou sly reported.17, 46 The manufacturer specification for calibration accuracy is a kV / MV isocenter coincidence of routinely verify the coincidence to within 0.3 mm. In this study, the XVI scans were acquired using a 27.6 x 27.6 x 20 cm3 FOV an d a gantry rotation range of 205 over approximately 70
49 seconds. The XVI software (version 4.0) allows an ROI fusion volume to be defined on the planning CT image for the registration. After the CBCT images are acquired, registration with the planning CT i s performed using a bony registration method. All three of the systems image registrations yield six degree -of -freedom correction parameters and the translational results are displayed to sub-millimeter precision. Align RT3C Evaluation A modified Rando head andne ck phantom (RHNP), shown in Figure 3 3 (a), was used to assess the accuracy of ART with respect to XVI and FSA. The phantom was modified by affixing a bite plate with attached reflective markers and position adjustment knobs. The adjustment knob s allowed for rotational movement of the phantom, and the FSA system was used to monitor the movement of the phantom via the reflective markers. To simulate a typical patient setup, the planning CT images of the phantom were transferred to both FSA and XVI and the skin contours (CT_S) from the planning CT images were transferred to the ART The treatment plan isocenter was chosen to coincide with the intersection of orthogonal lines that pass through three radiopaque BBs on a single transverse cut of the p hantom image. Accuracy of system calibration The calibration accuracy of all three systems was assessed. Vendor -provided calibration devices were used for the system calibrations, which were done according to vendor guidelines. To determine the inherent v ariances of the isocenter position for each system calibration, the RHNP was placed on the couch at the treatment position, and the FSA system was used to position the isocenter to along all three axes. The CBCT images and the ART surfa ce images were acquired and registered with the planning CT and CT_S references, respectively. This was done ten times to obtain the statistics of the isocenter corrections The mean differences between these values were used as the offset for the phantom localization study
50 Accuracy of registration algorithm The accuracy of the ART registration algorithm was evaluated by using the system to determine rotational and translational displacements of the isocenter using the RHNP, and comparing the results to t hose obtained with the FSA and XVI systems. To simulate pretreatment setup (gantry = 0 and couch = 0 ), the phantom was positioned on the table at the treatment position as determined by the FSA system. The ART_S reference was captured after the phantom wa s localized using XVI (to within 0.3 mm and 0.3 of isocenter) to eliminate the inherent variances between XVI and Align RT3C. The phantom was then randomly positioned at over 70 different locations lying within a range of 20 mm and/or 5 of the isocenter. The displacement at each position was determined with the ART using both the CT_S and ART_S references. A CBCT image was also acquired at each position and the corresponding displacements were calculated using bony registration. The displacements were determined with the FSA system as well. To simulate noncoplanar delivery, thirty additional data sets were acquired over the full range of couch angles ( 90 ) and/or full range of gantry angles (360 ). The displacements were determined using only the FSA and ART systems since the XVI system can only operate with the couch at 00. Uncertainties of surface reconstruction The ART surface reconstruction may be influenced by inter and intra -fractional motion, skin tone, camera thermal effects, room light variation, and ROI selection. Five healthy volunteers differing in a ge, sex, weight, and skin tone were used to validate the accuracy of inter and intra -fractional motion. Only the ART and FSA were used for the volunteers, the XVI could not be utilized because of the hazard of radiation exposure. The
51 standard localization/ immobilization devices were used for each volunteer, such as a bite plate and a custom disposable head support (MoldCare Pillow, Bionix Inc, Toledo, OH). To acquire an adequate surface image, we applied a thermoplastic chin stretch attached to an adjustab le head holder instead of a full face mask, as shown in Figure 3 3 (b). Each volunteer was repositioned 10 times (standing up and lying down each time) to within 0.3 mm and 0.3 of the setup position. Intra -fractional motion was analyzed for 2 minutes each time. The displacements were recorded to illustrate the variances of inter and intra fractional motion between FSA and ART In the ART software, three skintone protocols are available (fair, mid, dark) which use different camera exposure settings to com pensate for the brightness of room light and / or skin tone. To simulate treatment conditions, the phantom and five volunteers with different skin tones were repositioned under different room lighting levels. At each position, we randomly adjusted the room light and tracked the phantom / volunteers position while alternating between the three skin tone protocols. The displacements obtained using the protocols were calculated. If the cameras are left on for long periods of time, the resulting thermal effec t could compromise the long -term stability of the camera performance. The phantom was continuously imaged for over 10 minutes to simulate a typical setup. The phantom was reimaged for 10 minutes every 30 minutes for up to 10 hours to simulate a typical tre atment day. All factors were examined using both CT_S and ART_S references. We analyzed the effect of eight different user -defined ROIs. These regions consisted of no ROI, the whole face, the central face, the default, the default ROI excluding the nose, t he frontal face, the lef t face and the right face (Figure 3 4). The user can assign any ROI as the default ROI. The central and frontal -face ROIs were combined and selected as the default ROI for our
52 study. The ROIs were applied to the reference images for both the RHNP (CT_S, ART_S) and the volunteers (ART_S). For each defined ROI, the displacements were calculated. Results and Discussions Accuracy of System Calibration The system calibration of the ART agrees with the calibrations of the other two syste ms to within 1.3 mm and 0.8 as shown in Table 3 1. The directional differences along the phantoms left -right, superior -inferior, and anterior -posterior axes are LAT, LONG, and VERT, respectively, as shown in Figure 3 3(a). In a previous XVI study a s ystematic difference between the kV and MV defined radiation isocenter of 0.3 0.2 mm was observed10. Chang et al .7 reported good agreement between the results obtained with CBCT with those of the Brown Roberts Wells (BRW) frame. We therefore selected the XVI as our reference. The mean differences between the ART and XVI systems were 0.2 mm ( LAT), 0.4 mm ( LONG), a nd 1.2 mm ( VERT), and there was a difference between the Align RT3C and FSA of 0.6 mm along each axis. The maximum rotational difference about any axis was less than 1 By contrast, the isocenters of the ART and FSA were located closer to one another an d were spaced more uniformly (0.6 mm along each axis), than when either was compared with the XVI. This is due to the similar manufacture -defined calibration methods of FSA and ART which are both calibrated with respect to the mechanical isocenter of the LINAC. Many studies5,9,11 have reported that the registration errors for the XVI, FSA, and ART are all within 0.5 mm of one another. The maximum standard deviation of the registration error for our study was 0.3 mm for all three systems (Table 3 1). Due to the variances in the calibration methods, the software of ART (or FSA) supports the conversion matrix found using the radiation isocenter in order to correct for isocenter displacements.
53 Accuracy of Registration Algorithm The means and standard deviation s of the differences between the ART isocenter displacements with those of the other two systems for the 0 gantry and couch setups are shown in Table 3 2. Note that the systematic calibration differences between the systems were subtracted before doing t he comparison. For the mean values, the ART results obtained using ART_S show better agreement with the FSA (0.3 mm / 0.2 ) than with the XVI (0.9 mm / 0.4 ). Additionally, the XVI presented larger standard deviations than the FSA when compared to the Alig n RT3C (0.5 mm vs. 0.1 mm). The maximum discrepancy between the ART and the XVI was 3 mm, and was 0.4 mm between the ART and FSA when the phantom was located more than 10 mm / 3 from the isocenter. For small displacements (< 3 mm / 3 ), the differen ces between ART and either XVI or FSA were < 0.5 mm / 0.5 Different registration algorithms could cause registration uncertainties. The above results could indicate that the XVI is less accurate in image registration with the combination of larger transl ation and rotational errors. A study by Masi et al .8 similarly concluded that with SRT a second CBCT should be performed if a rotational error of > 2 is obtained with the first CBCT. The accuracy of rotation for the three systems was approximately 1 even for a phantom offset of more than 30 mm / 5 The means and standard deviations of the differences of the ART isocenter displacements with those of the FSA system for couch angles of up to 90 are listed in Table 3 3. For the mean values, the AR T results were within 1.2 mm of the FSA values using the ART_S references. T he rotational differences were all less than 1 The maximum differences were 2.3 mm / 1.2 for couch angles of 90 Because of the locations of the two side cameras, only parti al images can be acquired for non -zero couch angles. The amount of image cutoff increases with increasing couch angle. At a couch angle of 90 only approximately half of the image is
54 acquired, with respect to the image acquired with the couch at 0 Sin ce this may affect the accuracy of registration, a third, centrallylocated camera is important for non -coplanar treatments. Gantry blockage effects are illustrated in Figure 3 5, which shows the trend of displacements for the ART system for a gantry angl e range of 360 The displacement variance is 0.4 mm for the ART_S references. For the CT_S references, the displacement variance lies between 2 mm and 3 mm for gantry angles between 30 and 60 and between 300 and 330 No significant variances in rotat ional displacement differences (Max: 0.7 (CT_S) and 0.3 (ART_S)) were found. ART is feasible for localizing and monitoring non-coplanar SRT treatments for full gantry and couch rotations using ART_S references. However, gantry blockage of the camera pods during gantry rotation may contribute to registration errors. This effect means that a blocked camera is not available for use in the registration. The use of CT_S references might accentuate this effect; however, CT_S is used only for the initial patient setup Tables 3 2 and 3 3 show the registration uncertainty of the ART is slightly larger (Max, 1.3 mm) for CT_S than for ART_S. Average registration variances for isocenter displacements of 0.3 0.2 mm for ART_S and 1.2 0.8 mm for CT_S were obtained, with rotational differences lying within 1 for both reference types. With no gantry or couch rotation, the mean differences between the two references were 0.4 mm and 0.2 (compared to XVI) and 0.6 mm and 0.6 (compared to FSA). With couch rotation, the mean differences between the two reference types were 1.3 mm / 1.0 The values of the standard deviation show that CT_S may have slightly greater uncertainty than ART_S, especially for larger displacements. Som e possible factors contributing to this uncertainty include CT_S construction based on the chosen CT
55 density threshold, and the similar projected spectral pattern between ART_S and the real time captured image during the treatment. Uncertainties of Surface Reconstruction The mean differences between the ART and FSA displacements obtained for the volunteer study are shown in Table 3 4. For inter -fractional motion, the mean and maximum displacements of ART when compared to FSA, were 0.5 mm and 1.0 mm without couch rotation, but the numbers increased to 1.3 mm and 1.5 mm for couch angles of 90 These values are close to those obtained for the phantom tests. The maximum differences for intra -fractional motion were within 0.9 and 1.2 mm for all of the couch angles. No skin tone effect was observed for the volunteers (< 0.5 mm and 0.5 ) and the effect on registration uncertainty for different levels of room lighting was < 0.2 mm and 0.2 For the camera thermal effect stability tests, all displacements were l ess than 0.3 mm and 0.3 for a period of 10 hours. A previous study8 reported the long -term stability in static mode to be within 0.5 mm and 0.01 Figure 3 6 shows the deviations measured using ART_S and CT_S for the 7 ROI settings, with respect to the default ROI. Average registration variances for isocenter displacements of 0.3 0.2 mm for ART_S and 1.2 0.8 mm for CT_S among the ROIs were obtained, with rotational differences lying within 1 for both reference types. Using CT_S, the results showed that the central and frontal face area (the default ROI) was the optimal region to use as the reference since it provided the minimum registration uncertainty. The default ROI was applied to the phantom for the system calibration accuracy and registration algorithm tests.
56 Conclusions This study evaluated the ART as an IGSPS using real time mode for high-precision treatments. A systematic isocenter displacement was observed between the ART FSA, and XVI, but a conversion matrix could be applied with the ART Positioning and tracking performance using a phantom was comparable to that of the CBCT and optical tracking techniques. Inter and intra -fractional motion tests performed using the volunteers demonstrates the equivalency of the system with the FSA. The system characteristics of all three IGSPSs ( ART XVI and FSA) including imaging modality, tracking references, tracking mode, geometry calibration, detection capability, additional dose, attached localization devices, and registration uncertainties are l isted in Table V. The combination of CBCT and ART can be applied to frameless SRT for non-coplanar setups, but with the additional advantages of internal anatomy visualization and the elimination of the uncertainties inherent in the use of patient attached external devices for positioning and tracking.
57 Table 3 1. Coordinate sys tem displacements between 3 systems explored with phantom tests. Table 3 2. Comparisons of the positioning variances within 20 mm and/or 5 Table 3 3. Differences between the 3D phantom displacements for full couch rotation. AlignRT3C v.s. XVI FSA v.s. XVI AlignRT3C v.s. FSA Trans(mm) Rot ( ) Trans(mm) Rot ( ) Trans(mm) Rot ( ) LAT 0.2 0.3 0.2 0.2 0.9 0.3 0.6 0.2 0.6 0.2 0.8 0.1 LONG 0.4 0.1 0.4 0.4 0.4 0.1 0.5 0.2 0.6 0.1 0.1 0.3 VERT 1.2 0.2 0.7 0.3 0.7 0.1 0.0 0.2 0.6 0.1 0.7 0.2 displacements 1.3 0.1 1.2 0.2 1.1 0.1 AlignRT3C v.s. XVI AlignRT3C v.s. FSA ART_S CT_S ART_S CT_S Trans(mm) Rot ( ) Trans(mm) Rot ( ) Trans(mm) Rot ( ) Trans(mm) Rot ( ) LAT 0.5 0.4 ( 2.1 ) 0.2 0.1 ( 0.4 ) 0.6 0.3 ( 0.7 ) 0.3 0.2 ( 0.4 ) 0.2 0.1 ( 0.3 ) 0.2 0.2 ( 0.4 ) 0.3 0.1 ( 0.3 ) 0.4 0.3 ( 0.7 ) LONG 0.7 0.6 ( 2.1 ) 0.4 0.2 ( 0.6 ) 0.8 0.8 ( 2.6 ) 0.6 0.5 ( 1.3 ) 0.1 0.1 ( 0.2 ) 0.1 0.1 ( 0.2 ) 0.8 0.8 ( 1.8 ) 0.4 0.3 ( 0.8 ) VERT 0.4 0.7 ( 2.1 ) 0.3 0.2 ( 0.5 ) 0.9 0.7 ( 1.9 ) 0.5 0.5 ( 1.2 ) 0.2 0.1 ( 0.3 ) 0.1 0.1 ( 0.2 ) 0.2 0.1 ( 0.2 ) 0.7 0.4 ( 0.9 ) displacements 0.9 0.5 ( 3.0 ) 1.3 0.2 ( 3.5 ) 0.3 0.1 ( 0.4 ) 0.9 0.7 ( 1.8 ) ART_S CT_S Trans(mm) Rot ( ) Trans(mm) Rot ( ) LAT 0.3 0.2 ( 0.7 ) 0.7 0.4 ( 1.2 ) 1.3 0.7 ( 2.1 ) 1.7 0.6 ( 2.7 ) LONG 1.0 0.7 ( 2.1 ) 0.3 0.2 ( 0.7 ) 1.5 0.8 ( 2.8 ) 0.7 0.6 ( 1.8 ) VERT 0.7 0.2 ( 1.1 ) 0.3 0.2 ( 0.6 ) 1.5 0.4 ( 2.2 ) 0.5 0.3 ( 1.0 ) displacements 1.2 0.7 ( 2.3 ) 2.5 0.2 ( 3.2 )
58 Table 3 4. Differences between inter and intra fractional motion over 2 min. for five volunteers Table 3 5. System characteristics of three IGSPSs Couch 0 0 Couch 90 0 Interfractions Intra fractions Interfractions Intra fractions Trans(mm) Rot ( ) Trans(mm) Rot ( ) Trans(mm) Rot ( ) Trans(mm) Rot ( ) LAT 0.2 0.2 ( 0.6 ) 0.2 0.1 ( 0.4 ) 0.4 0.3 ( 0.8 ) 0.4 0.1 ( 0.5 ) 0.3 0.1 ( 0.6 ) 0.5 0.4 ( 0.8 ) 0.4 0.2 ( 0.7 ) 0.5 0.1 ( 0.8 ) LONG 0.3 0.3 ( 0.9 ) 0.3 0.2 ( 0.6 ) 0.5 0.2 ( 0.9 ) 0.4 0.2 ( 0.9 ) 1.0 0.7 ( 1.2 ) 0.3 0.2 ( 0.6 ) 0.5 0.2 (1.1) 0.4 0.2 ( 0.8 ) VERT 0.3 0.2 ( 0.7 ) 0.3 0.2 ( 0.8 ) 0.3 0.2 ( 0.6 ) 0.4 0.3 ( 0.9 ) 0.8 0.3 ( 0.5 ) 0.3 0.2 ( 0.8 ) 0.3 0.2 ( 0.6 ) 0.4 0.3 ( 0.7 ) displacements 0.5 0.2 ( 1.0 ) 0.7 0.2 ( 0.9 ) 1.3 0.7 ( 1.5 ) 0.7 0.2 ( 1.2 ) AlignRT3C XVI FSA Imaging Modality Video Imaging X ray Imaging Optical Imaging Tracking References 3D External Surfaces 3D Internal Anatomy Four Reflector Markers Tracking Mode Real Time/ Static Pre Treatment/ Post process Real Time Geometry Calibration Mechanical (Light Field/ Laser) MV Radiation Isocenter Mechanical (Light field/ Laser) Detection Capability Couch, Inter & intra fractional motion Inter fractional motion only Couch, Inter & intra fractional motion Additional Dose No Dose Yes No Dose Attached Localization Devices No No Bite block & sphere markers Registration Uncertainties Camera blockage Misalignment > 1cm/3 Non rigid body
59 Figure 3 1. Schematic of the hardware setup for 3 IGSPSs in the treatment room A B Figure 3 2. Two surface references. A) CT_S. B) ART_S
60 A B Figure 3 3. Two setups. A) phantom. B) volunteers Figure 3 4. Definition of 8 ROI settings
61 Figure 3 5. Registration variances of the AlignRT3C during the full gantry rotation Figure 3 6. Registration variances for isocenter displacements for 7 ROI settings measured
62 CHAPTER 4 CLINICAL EVALUATION OF NEW SETUP SYSTEM Introduction Frame -based SRS, with a rig id ring fixed to the skull providing a frame of reference for targeting and rigid immobilization for treatment, has been used for >50 years in the treatment of intracranial lesions. Issues with dose tolerance of normal tissue lead to the development of fra ctionated SRT In recent years, linear accelerators that h ave integrated MV and kV imaging systems have provided the capability to deliver image -guided alignment to eliminate entirely the frame based technology for SRT. With these advances, interest has in creased in using image alignment with non rigid immobilization as frameless SRS/SRT. A number of new IGSPSs have contributed to the improved clinical use of frameless SRT. The drawbacks and advantages of IGSPSs are discussed in the AAPM TG 68.1 The potential sources of errors in image -guided targeting include the calibration of the imaging system and errors in the reconstructed radiograph registration caused by limits of system geometry, pixel size and slice thickness. Although sub-mm accuracy c an theoretically be achieved in phantom, it rarely is due to difficulties in the patient immobilization, system limitations, and organ motion. Using X ray imaging, targeting of the intracranial tumor is then adjusted to the internal anatomy at treatment, thus removing the potential error associated with targeting that is determined using external landmarks. Besides X ray imaging, optical tracking techniques, which provide real time tracking of external markers affixed to the patient, have been used in many clinics over the past decade .4 The techniques are limited by the reproducibility of the correlation of the internal target with the external markers. Therefore, a few reports describing the clinical outcomes for several commercial frameless IGSPSs such as Sonarray (optical imaging) ,4 CBCT
63 (x ray imaging ),7 Novalis51 and CyberKnife22 (the combination of xray and optical imaging) systems have been published. Without the need of external markers and additional radiation exposure, the complete 3D surf ace mode ls using video cameras for the patient setup were developed52 and applied in many clinical applications53, 54 as a commercial surface imaging system, ART (Vision RT Ltd, London, UK). The original system was introduced by Bert et al.5 and utilized two side camera pods to acquire static 3 D surface images at 0 couch angle. The system was redesigned to include a central camera pod and provide the real -time tracking, as A RT, to ensure high precision guidance and monitor intra -fraction motion with non -coplanar beams. At our institution, all three commercial IGSPSs, ART FSA (Zmed/Varian, Inc., Ashland, MA), and kV XVI (Elekta Oncology System s, Norcross, GA) (shown in Figure 4 1), are installed in the same treatment room with the same LINAC (Elekta Oncology System, Norcross, GA). In preclinical investigations to quantify the accuracy of target localization using ART we confirmed its accuracy in phantom and healthy volunteers wa s comparable to that of other two systems.55 We now report on the clinical experiences of patients who underwent SRT treatment with those IGSPSs. For this work, we analyzed 5 patients to quantify setup accuracy and focused on (1) the comparison of three systems in pre -treatment target alignment at zero gantry and couch angle, (2) the quantification of inter -fractional motion using ART and FSA in real time tracking at each non -coplanar field, and (3) the analysis of intra -fractional motion using ART during r adiation delivery and CBCT acquisitions. Methods and Materials Patients, Immobilization and Treatment characteristics We selected 5 patients treated with IMRT for intracranial cancer whose targets were located at different areas with different ages, skin tones, tumor sizes and beam arrangements to
6 4 be able to study the setup errors in these situations. Table 4 1 lists the characteristics of the patients, the tumors, the prescribed and delivery treatments. All plans were designed with 6 MV as a minimum of 95% of the planning target volume (PTV) received 100% or more of the prescribed dose on a Philips Pinacle3 treatment planning system ( Philips Medical Systems, Madison, WI) (V 8.1) when optimizing the IMRT treatment. Five to 8 non -coplanar beams yielded a result that was optimal, combined with a realistic treatment time (approximately 10 minutes). The range of dose prescription was 4560 Gy with conventional 1.22.0 Gy fractions delivered daily 5 days a week for 5 to 7 weeks. The standard localization/immobilization devices were applied to each patient, such as a bite plate and a custom disposable head support (MoldCare Pillow, Bionix Inc, Toledo, OH). A custom bite plate with two nonre flective and four passive reflective spherical markers is affixed to the patients maxillary dentition to form a rigid system. The non -reflective markers are used for the isocenter registration. To acquire an adequate surface image, we modified the bite pl ate by reversing marker arrangements and applied a thermoplastic chin stretch attached to an adjustable head holder instead of a full face mask, as shown in Figure 4 2. A planning CT scan was made with a 2 -mm slice width using a Philips Brilliance 16 -slice scanner (Philips Healthcare, Andover, MA) from the bottom of the bite plate through the whole head. ). The patients position is determined through registration of the reflective markers positions with respect to isocenter, tracked by FSA in real time. Imaging, Setup flow and Analysis protocols The complete prot ocol is shown in Figure 4 3, and the system description, setup flow and evaluation analysis used for patient setup are described in the following sections.
65 System descriptions The AlignRT3Csystem utilized 3 camera pods to ensure complete 3D surface image acquisition over the full gantry and couch range. Several stereoscopically arranged CCD w ere housed inside each pod. Camera calibration is performed using the light field and room lasers. The repro ducibility of the calibration is within 0.8 mm5 on the daily basis. All acquired images are reconstructed as a 3 D surface by triangular tiles with approximately 1 to 3 mm spacing between intersections and 10,000 points. Two types of surface image referenc es; skin contours reconstructed from p lanning CT images (CT_S) and optical surface images previously recorded by the ART (ART_S) are shown in Figure 4 4 (a -b). Texture is achieved with the appropriate part of a bitmap image as Figure 4 4 (c). Our study was done in the real time mode, for which the surface image capture frame rate is 1 to 2.0 frames per second, depending on the size of the region of interest (ROI) selected for the registration. The regi stration screen is shown in Figure 4 5 which displayed t he displacements in 6 DOF, the ROI and the couch angles. The FSA system is a CCD infra red camera system and can markedly improve patient positioning precision by providing positioning errors in real time.4 The patients head is immobilized attached to an adjustable head holder. The patients position, which is determined through registration of the reflective marker positions with respect to the isocenter. However, factors such as the reproducibility of bite plate seating, patient comfort, and dental impre ssion technique can affect positioning accuracy by as much as 4 mm.48 The XVI system consists of an aSi detector and a kV xray source installed on two retractable arms. The unit produces diagnostic -quality x rays from 70 to 150 kVp. In this study, the XVI scans were acquired using a 27.6 x 27.6 x 20 cm3 FOV and a gantry rotation range of 205 over approximately 70 seconds. The XVI software (version 4.0) allows an ROI fusion
66 volume to be defined on the planning CT image for the registration. After the CBCT images are acquired, registration with the planning CT is performed using a bony registration method. All three of the systems image registrations yield six DOF correction parameters and the results are displayed to sub -mm and sub-degree precision. Setup flow The accuracy of the ART to determine rotational and translational displacements of the isocenter by comparing the results to those obtained with the FSA and XVI systems was evaluated. Patient setup at each field was first done using FSA for daily localization. After the initial FSA setup, The ART images was acquired prior to treatment at each treatment beam for daily and the CBCT images for the first three fract ions and weekly thereafter. Figure 4 3 shows the setup flow schematically. To evaluate pre-treatment setup (gantry and couch = 0 ), the patient was positioned on the table at the treatment position as determined by the FSA system to .4 along all three axes as the setup tolerances. After the FSA guidance, simultaneously, the ART_S reference in the first fraction and the setup surface images were captured (SFSA_ISO). The CBCT images were also acquired and registered with the planning CT. The ROI volumes were verified in the XVI software by the physician before the first CBCT acquisi tion Note that CBCT images for these patients were acquired only for verification and quality assessment purposes. Also note that the residual errors in the patient position accruing during CBCT acquisition and/or treatment delivery were monitored for by the FSA to ensure that the changes in the translational and rotational displacements about any axis remained less than 1 mm and 1 respectively. Therefore, after the CBCT acquisition, another setup surface images were captured (SXVI_ISO) to eliminate the residual errors and intra -fractional motion before the comparisons.
67 To evaluate inter -fractional motion at each non -coplanar beam, each patients position was re adjusted until the FSA setup tolerances were met. The surface references were rotated to treat ment couch angles by selecting treatment couch angles in the software, shown in Figure 4 6 The reference and setup surfaces were shown with pink and green colors, respectively. Additionally, to evaluate the infra -fractional motion, during the beam deliver y, the displacements were monitored in real time by FSA to Note that maximum tracking time is 2 mins for each noncoplanar beam ( case #1,3,and 4) because therapists repositioned at each couch angle but 14 mins for coplanar beams (case # 2,and 5) while the gantry is rotated. The setup surface images were acquired (SFSA_Inter and SFSA_Intra) and displacements with ART were recorded to illustrate the variances of inter -fractional and intra -fractional motion between FSA and ART In our study, according to the study of system characteristics,55 we selected the ROI covering the central and frontal head in the ART software, shown in Figure 44 and 4 6 as the default ROI. All setup surface images were registered both references with the default ROI We randomly adjusted the room light and tracked the patients position while alternating between the three skin -tone protocols (fair, mid, dark). Note that the ROI and skin -tone settings can be adjusted by the user. Following the described positioning pr ocedure, we were able to analyze the translational and rotational errors along the tree major axes. We represent the directions of the registration displacements along patients' left right, superior inferior and anterior -posterior directions as LAT, LONG and VERT, respectively. The mean, standard deviation, 90th percentile (%), and maximum values were calculated along three axes and for the module of the vector, considering the totality of the data obtained for all patients and all fractions.
68 Results Ev aluation 1: The I socenter P ositioning D ifferences between Three IGSPSs at C ouch and G antry 0 Table 4 2 shows the isocenter positioning differences of pre treatment verification between 3 ISPSs at couch and gantry 0 A total of 39 SXVI_ISO, SFSA_ISO and CBCTs, (range, 7 11 scans/patient, mean, 7.8) in first 3 fractions and on weekly basis for 5 patients were analyzed. For the mean and standard deviations (SD) values, the ART results obtained using ART_S show better agreement with the XVI (0.2 0.3 m m) than with the FSA (0.5 1.1 mm). We also analyzed whether the positioning differences were affected by bite -tray with teeth impression and reflector balls. The mean SD vector difference between FSA and XVI were approximately 1.3 1.2 mm. The 90th % differences between the ART and XVI systems were < 1.0 mm in all directions. Comparing the positioning between the ART and FSA, those values were approximately 1.0 mm in the lateral and longitudinal directions and 2.4 mm in the vertical direction. The maxi mum rotational difference between ART and XVI (FSA) was 1.3 (1.1 ) about any axis and 2.4 between FSA and XVI. Table 4 3 show s the isocenter positioning differences between the ART (SXVI_ISO) and XVI at couch and gantry 0 using two surface references. Note that the systematic differences between systems found using CT_S in the first fraction were subtracted before doing the comparison. The results show the mean registration uncertainty of the ART at couch and gantry 0 using ART_S ( 0. 2 mm/ 0. 2 ) is simil ar to CT_S (0.3 mm/ 0. 1 ). Maximum translational and rotational differences for both reference types were within 2.1 mm and 1.4 respectively.
69 Evaluation 2: The I socetner P ositioning D ifferences between AlignRT3C and FSA at E ach N on -coplanar F ield The isocenter positioning differences of the ART with those of the FSA system for all treatment fields are shown in Table 4 4 and Figure 4 7 A total of 635 SFSA_Inter (range, 5 8 beams/ patient, 25 50 fractions/patient) on daily basis for 5 patients were analyzed. For the mean SD values, the ART results using both references were within 0.8 1.5 mm of the FSA values. Overall using ART_S, 71 % and 89 % of the cases had displacement differences < 1 mm/1 and < 2 mm/1 respectively, in any direction. In contrast, using CT_S, 90%, 67% and 37 % of cases had displacement differences < 3 mm, < 2 mm and <1mm, respectively, and 95% and 72 % of cases were < 2 and < 1 respectively. The maximum differences using ART_S (CT_S) of 3.9 mm / 2.6 (6.9 mm/3.4 ) were found at couch angles of 90 Evaluation 3: Intra -fractional M otion with AlignRT3C The Intra -fracti on motion is illustrated in Figure 4 8 which shows the trend of displacements for the ART system during one CBCT acquisition (a) and treatment delivery with different treatment beams (couch 0 couch 20 and gantry 90 ( c20g90) and c270 g300 sh own in Figure 4 8 (b) (d)).Note that during the intra -fractional time period, treatment tolerance for FSA is 1 mm/1 and the data represent with ART only since the FSA can't record position values. Also note for coplanar beam, due to longer treatment time (~10 mins), therapists need to walk in to re -positioned when the displacement is larger than 1 mm/1 O verall, the displacement variance of intra-fraction motion with SFSA_Intra (N=635 sessions) is approximately 0.5mm/0.5 < 1mm/0.5 and 1.5 mm/1.5 at couch 0 and 90 These values are close to those obtained for the volunteer tests.10. Figure 4 8 ( b ) show the intra -fraction motion was tracked for 6 co -planar beams (couch 0 and gantry 120 90 45 315 270 and 240 ) over 14 mins
70 without any interruption. The range of intra -fraction motion is within 0.4mm/0.4 and approximately 1 mm/1 during the gantry rotations. In addition, the displacement variance lies between 2 mm and 3 mm for gantry angles between 20 and 50 during the CBCT acquisitions (range, 255 110 in clockwise direction) with extendable panel and the X ray tube bl ockage. Discussion s The goal of IGSPS is to ensure the high precision guidance for the efficacy of SRT treatment when frameless setup is applied. Three IGSPSs were discussed. For FSA, a poor dental impression and reflector balls mainly causes bite plate po sition inconsistencies during the CT simulation and the treatments. Currently, w e performed clinical frameless SRT patient localizations using FSA guidance and kV CBCT images as the quality assessment and reported our clinical experience.48 The FSA system alone cannot guarantee the repositioning accuracy over the whole course of treatment and the conjunction of FSA with CBCT is needed. Additionally, the bite -plate with dental impression is not suitable for all patients. ART is developed to replace FSA syste m without any external frame/markers attached to patients and validated as the feasible IGSPS. The inherent uncertainties between three systems include calibration procedure, surface localization versus reflector tracking versus image acquisition, and registration algorithm. Overall, according to the phantom study, the system calibration of the ART agrees with the calibrations of the other two systems to within 1.3 mm and 0.8 which is due to the different calibration methods and the registration error .55 The system differences between systems are very reproducible and therefore t he software of ART (or FSA) supports the coordinate transformation matrix found to correct for isocenter displacements.
71 For the pre -treatment verification for the 0 gantry and couch, SXVI_ISO has better agreement with XVI (0.2 0.3 mm) than SFSA_ISO with FSA (0.5 1.1 mm) although SXVI_ISO included the residual errors and intra -fractional motion during the CBCT acquisition. We suggested that the main source of uncertainties for SFSA_ISO was from the inconsistency of bite -tray. Figure 4 9 shows the trend of isocenter correc tions from 8 CBCT sessions for case #2 over the entire treatment period of 5 weeks. The first 4 corrections correspond to the first 4 fractions and the last 4 corrections are for the weekly CBCT acquisition. The corrections increase from approximately 2 mm for the first 4 fractions to 4.6 mm for the last weekly CBCT acquisition. Therefore, in table 4 2 the displacement between FSA and XVI was 1.3 mm for those 5 patients in our study which can match our previously study (1.2 mm) for 15 SRT patients using two CBCT systems to demonstrate the uncertainties of FSA.48 The results in patient study demonstrated larger differences at non-coplanar treatment fields between FSA and ART using the ART_S references than in the phantom study.55 The data presents t he mean displacement was 0.6 1.0 mm but still 11 % and 8 % of the cases had displacement differences > 2 mm and >1 respectively. In contrast to, in phantom study ,55 the mean difference is within 1.2 mm for the full test region ( for 20 mm and/or 5 of the isocenter for couch angles of up to 90 ). The maximum differences were 2.3 mm / 1.2 for couch angles of 90 with larger translations (>1cm) and rotations (>3 ). Note that the, for those patients, we used ART_S reference after the FSA guidance to 0.3m m/0.3 not CBCT. Therefore, the inconsistency of bite tray in FSA may contribute the same amount uncertainties to the noncoplanar treatment beams as well 0 gantry and couch in patients not phantom. Additionally, the systematic coordinate differences bet ween ART and FSA could be considered at large couch angles ( 90 ). Because of the gantry (couch) blockage effec ts, partial surface images (Figure 4 6
72 (c), (e) (f)) mainly were acquired by one central camera and one side camera at non zero gantry and couch angles. The amount of image cutoff increases with increasing couch angle. Since this may affect the accuracy of registration, a third, centrallylocated camera is important for non coplanar treatments. Furthermore, the extendable panel and X -ray tube can block the camera during the CBCT acquisitions when gantry rotated between 20 and 50 and the registration errors were increased up to approximately 3 4mm (Figure 4 8 (a)). Note that the range of gantry blockage and register errors can change with different panel positions. Two types of surface references were applied for all setup surface images. At the 0 gantry and couch angle, the displacement differences between ART and XVI using ART_S are similar (<1mm/1 ) to CT_S. At the non-zero gantry and couch angles, results show the registration uncertainty of the ART is slightly larger (range: 1 3 mm, 1 2 ) for CT_S than for ART_S. Some possible factors contributing to this uncertainty include CT_S construction based on the chosen CT density threshold, and the similar projected spectral pattern between ART_S and the real time captured image during the treatment. The use of CT_S references might accentuate this effect, especially for gantry blockage effect at couch angles; therefore, CT_S is suggested to use onl y for the initial patient setup in the first fraction. A critical concern with clinical adoption of IGSPSs which rely on mask -based immobilization is that of intra -fraction motion. For the studied treatments a minimum of 5 no coplanar beams is sued, the ta ble is rotated at least 3 times and the treatment duration is not short (approximately 15 min); intra -fractional positional changes could be serious and real -time tracking could be essential. In our clinical experience,29 ,f or CNS and head and neck IMRT pa tients using half TP facemask and real -time optical tracking (ExacTrac, BrainLab), patient displacement less than 1.5 mm for 95% of treatment time. In other studies ,8 56 the residual errors
73 and intra -fractional motion both at the beginning and end of treat ment using full TP facemask and CBCT and combination of kV X -ray imaging with optical tracking were approximately 1.5 mm and 1.0 0.5 mm, respectively. Therefore, the displacement of 1.0mm/1.0 is set as our tolerances for intra -fractional motion and tracked in r eal time in present study. Figure 4 8 (a) (d) demonstrated the intra -fraction motion with ART were 1.0 mm/1.0 and simultaneously monitored by FSA Due to the limitation of frame rate ( 1 1.7 frame/sec) in ART and auto recording in FSA, we cant correlate to displacements difference between two systems. However, the ideal patient immobilization not only can allow relatively fast and reproducible setup, but also help to maintain patient po sition during radiation. Engelsman et al.37 report on the 95 % probability that the displacement differences between immobilization devices for intra fractional accuracy is approximately 5 mm (range,2 7 mm).In present study, we applied TP chin stretch only but alternative devices will be designed to acquire enough surface im ages in the further study. At our institution, the first focus to data has been on use of 3D surface imaging for clinical evaluation to replace the current frameless system. However, for high precision radiotherapy, the system accuracy is essential to be validated in the wide variety of treatment conditions such as ROI selections, patient skin -tone and skin deformation, treatment room light. In the present study, the external ROI in ART w e applied demonstrated the reliable regions to correlate the internal ROI of in XVI. However, the hair, nose, ear, and muscle regions might the source of system uncertainties. Additionally, no skin-tone and deformation (<1 mm/1 ) was found. The skin deform ation can happen on patients who lose/gain weight or take certain medicines. Currently, we suggested CT_S uses for first set up only because of larger uncertainties. CT_S might be the ideal references to daily detect the skin deformation in the future stud ies. Typically,
74 we rely on lasers as initial setup so we turn off the room light in most cases. We randomly tested the room light in the present study with current skin -tone protocols settings. However, the settings would be changed by users for the clinic al setup flow. Conclusions We evaluated the accuracy of three frameless methods for SRT and validated ART as an IGSPS using real time mode for high-precision treatments. Using ART_S, sub -mm and sub degree accuracy for pre treatment setup at zero gantry and couch angle was observed between three systems. Because of the uncertainties of bite -tray, < 2mm/1 and 1mm/1 accuracy for inter and intra -fractional motion delivering by non -coplanar beams, respectively, with Align RT3C was reached. CT_S is suggested to use as the initial setup in the first fraction. The combination of CBCT and ART is feasible for framele ss SRT in the future studies, but with the additional advantages of internal anatomy visualization and the elimination of the external skin uncertainties.
75 Table 4 1. Summary of characterist ics of 5 SRT patients. Table 4 2. Comparisons of the isocenter variances at couch and gantry 0 (N=39) Translation(mm) ART3C v.s. XVI ART3C v.s. FSA FSA v.s. XVI Mean SD 90th % (Max.) Mean SD 90th % (Max.) Mean SD 90th % (Max.) LAT 0.0 0.4 0.6 (1.2) 0.0 0.9 1.3 (2.7) 0.5 1.2 1.8 (4.6) LONG 0.0 0.7 1.0 (2.1) 0.1 0.9 1.0 (2.5) 0.2 1.3 1.8 (3.5) VERT 0.2 0.3 0.6 (0.9) 0.5 1.2 2.4 (2.9) 1.2 1.2 2.8 (3.5) VE CTOR 0.2 0.3 0.5 1.1 1.3 1.2 Rotation() ART3C v.s. XVI ART3C v.s. FSA FSA v.s. XVI Mean SD 90th % (Max.) Mean SD 90th % (Max.) Mean SD 90th % (Max.) LAT 0.2 0.6 1.2 (1.3) 0.1 0.4 0.6 (1.1) 0.5 0.8 0.2 (2.3) LONG 0.1 0.5 0.6 (1.1) 0.2 0.4 0.3 (0.9) 1.0 0.6 0.2 (2.4) VERT 0.1 0.5 0.6 (1.1) 0.0 0.3 0.4 (1.1) 0.0 0.4 0.4 (1.0) Pt. no. Gender Age Diagnosis Site PTV volume (cm3) Prescribed dose /fraction dose Beam arrangements 1 M 17 Oligodendroglioma R. Frontal 111 60 /1.2 Gy c20g90,c55g120,c340g270, c305g240,c270g300 2 M 58 Glioblastoma L. Parieto occipital 269 60 /2.0 Gy c20g90,c55g120,c340g270, c305g240,c270g300 3 M 58 Brain Metastasis Whole brain 1777 45 /1.8 Gy g120,g80,g40,g320,g280, g240,c270g345,c270g300 4 M 25 Astrocytoma Midbrain brainstem 175 55.2 /1.2 Gy c20g90,c55g120,c340g270, c305g240,c270g280 5 F 44 Brain Metastasis Whole brain 1380 45 /1.8 Gy g120,g90,g45,g315,g270,g 240,c270g320,c270g280 Abbreviations: Pt. no: patient number; PTV: Planning target volume; c: couch; g: gantry; c20g90: couch 20 gantry90
76 Table 4 3 Differences of the isocenter variances using two surface references at couch/ gantry0 Translation(mm) ART_S CT_S Mean SD 90th % (Max.) Mean SD 90th % (Max.) LAT 0.0 0.4 0.6 (1.2) 0.0 0.7 0.5 (2.1) LONG 0.0 0.7 1.0 (2.1) 0.1 0.9 1.0 (1.8) VERT 0.2 0.3 0.6 (0.9) 0.2 0.5 0.9 (1.4) VE CTOR 0.2 0.3 0.3 0.6 Rotation() ART_S CT_S Mean SD 90 th % (Max.) Mean SD 90 th % (Max.) LAT 0.2 0.6 1.2 (1.3) 0.1 0.6 0.7 (1.4) LONG 0.1 0.5 0.6 (1.1) 0.0 0.6 0.7 (1.4) VERT 0.1 0.5 0.6 (1.1) 0.0 0.5 0.7 (1.0) Table 4 4 Differences of the isocenter variances at all non -coplanar treatment beams (N=635). Translation(mm) ART_S CT_S Mean SD 90th % (Max.) Mean SD 90th % (Max.) LAT 0.4 1.0 1.7 (3.9) 0.4 2.0 2.7 (6.9) LONG 0.2 1.1 1.1 (3.7) 0.1 1.4 1.6 (4.6) VERT 0.4 0.9 0.7 (3.3) 0.7 1.3 1.0 (4.5) VE CTOR 0.6 1.0 0.8 1.5 Rotation() ART_S CT_S Mean SD 90 th % (Max.) Mean SD 90 th % (Max.) LAT 0.0 0.6 0.8 (1.7) 0.2 0.8 0.3 (2.8) LONG 0.3 0.5 0.3 (2.2) 0.2 0.9 0.2 (3.4) VERT 0.2 0.6 0.5 (2.6) 0.2 0.7 0.2 (2.6)
77 Figure 4 1 Schematic of the hardware setup for 3 IGSPSs in the treatment room Figure 4 2 The example of patient setup procedures for frameless SRT
78 Figure 4 3 Schematic outline of the imaging, setup flow and analysis protocol. A CB BB Figure 4 4 The examp les of reference surface images A ) ART_S and B ) CT_S
79 Figure 4 5 Examples of registration process for AlignRT3C in real -time modulate Figure 4 6 Color display of surfaces at couch angles: reference (pink), and setup (green)
80 Figure 4 7 Histogram isocenter corrections using two references at each treatment field
81 Figure 4 8. Example of the intra -fractional motion of the AlignRT3C during CBCT Figure 4 9. Example of the intra -fractional motion of the AlignRT3C for 6 coplanar beams
82 Figure 4 10. Example of the intra -fractional motion of the AlignRT3C at couch 55 Figure 4 11. Example of the intra -fractional motion of the AlignRT3C at couch 270
83 Figure 4 12. Trend of isocenter corrections using XVI during the entire treatment course
84 CHAPTER 5 DAILY QA PROGRAM OF LINAC AND THREE POSITIONING SYSTEMS Introduction The AAPM TG 4014 recommends that a megavoltage radiation therapy units output be monitored every morning before commencing of treatment.60 A daily check of the linac is one of the main components of any QA program. The QA procedures involve periodic measurements of specified parameters using dedicated tools and phantoms that are validated with s pecific acceptability criteria and tolerance levels to ensure that hardware and software function safely and reliably and perform as expected. Vendors may provide instructions concerning how to perform the measurements and use the tools in clinical sites. Users analyze these materials and decide how to proceed or set up their own measurement protocols; therefore, some variations of QA methods may exist between institutions. A more ideal method is to adopt consistent sets of measurement techniques, phantoms, and criteria. To this end, the AAPM has updated the TG 40, which specifies new tests and tolerances, and produced the TG 142 report .61 It includes recommendations for QA parameters as well as their measurement frequency and acceptable criteria. With IMRT creating steep dose gradients and tight margins around tumors the potential for set up errors and organ motion leads to a greater chance of dose to surrounding OARs IGRT, wherein imaging devices are used at treatment delivery to increase the probability that radiation is delivered as closely as possible to the original plan, can help avoid these problems. Additionally, IGSPSs employ various methods and demonstrate accurate target localization in SRT ,1 the results are summarized in the AAPM TG 681 report. The use of this new technology necessitates a comprehensive QA program to maintain and monitor system performance characteristics, which were established at the time of commission. Currently, QA phantoms and
85 measurement methods may differ with different IG SPSs and institutions; however, several recent AAPM TG efforts61,62 and studies42,48 address these QA issues associated with radiographic systems. Ideally, one QA program could be applied to multiple radiograph ic and non radiographic IGSPSs. In the standar d QA procedure, radiographic images of the phantom are obtained to verify the systems ability to correctly position objects in the image.42,48 However, nonradiographic systems are limited to the correlation of the camera center and radiation delivery center.9,12 Currently, 3 IGSPSs, the AlignRT 3 D system with three camera pods as ART (Vision RT Limited, London, UK); the FSA (Zmed/Varian, Inc, Ashland, MA); and the kV CBCT system (Elekta Oncology Systems, Norcross, GA), were installed in the same treatment room with an Elekta LINAC (Elekta Oncology System, Norcross, GA) ( Figure 5 1). The imaging installations in the treatment room were calibrated by vendor instructions to match the LINAC and room coordinate systems. For daily QA, therapists routinely pe rformed dosimetric checks of the LINAC and geometric accuracy tests of the 3 IGSPSs by repositioning different phantoms provided by the vendors. Switching phantoms not only wastes time and effort but may also cause systematic errors. To avoid removing the variety of phantom and to simplify and improve the daily QA proce dure, we designed a simple yet comprehensive daily QA program using the commercial DailyQA3TM (DQA3) (Sun Nuclear Inc, Melbourne, FL) phantom for LINAC and the 3 IGSPSs The QA program encom passes 3 components: dosimetry, mechanical (couch and laser), and imaging geometry. In this paper we describe the procedures of test items included in the QA program and present the results of measurements over extended periods.
86 Methods and Materials Description of the DailyQA3TM Device With the increasing number of complex treatments there is a pressing need for QA tools that can help improve QA efficiency and comprehensiveness. The commercially available DQA3 can perform dosimetry characteristics of LINAC for daily QA procedures using multi detectors.47 It consists of 25 detectors in a 20x20 cm2 detection area shown in Figure 5 2: one cylindrical chamber at the central axis (CAX) and 4 rectangular ion chamber at 8 cm from CAX for the flatness, symmetry, an d output check, 4 curved ionization chambers located along the corners of 16 cm square for photon energy check; 4 circular chambers with inherent attenuators fo r electron energy verification; 4 sets of three diodes with 5 mm spacing in the cent ral for light radiation coincidence check. This device supports automatic temperature and pressure correction and real -time measuremen ts without additional buildup. In order to remove differences among the readings of individual detectors, array calibration should be performed under the same conditions as the measurements that will be made. The array was calibrated using a 2020 cm2 field size with a 100 cm source skin distance (SSD) to determine a relative sensitivity of the detectors, stored as individual correction factors to be applied to the raw measurements. In order to perform absolute dose measurements, a dose calibration has to be applied to the raw data. The calibration was performed with a 1010 cm2 field size, positioning the device at the calibr ated depth and relating the reading of the central axis detector for 100 monitor units (MU) with the dose of 100 cGy. The absolute dose calibration factor is then applied to each array's detector, in addition to the sensitivity correction factors obtained during the array calibration.
87 Description of the Three IGSPSs The Align RT3C system was designed to acquire 3 D surface images utilizing two side and one central camera pods. Several stereoscopically arranged CCD cameras were housed for computing the surfa ce image and providing gray level and dynamic images. All acquired images from the 3 camera pods are reconstructed as a 3 D surface with approximately 1 to 3 mm spacing and 10,000 points. Either static or continuous real -time surface image displays are su pported. All measurement in our study was done in the real time mode, for which the surface image capture frame rate is approximately 1.5 frames per second. Additionally, the software (version 4.2) supports two types of surface image references; skin conto urs reconstructed from p lanning CT images and optical surface images previously recorded. Camera calibration is performed with the light field and room lasers on the daily and monthly basic recommended by the manufacturer. The reproducibility of the calibration is within 0.8 mm .12 The FSA system is a two CCD infra -red camera system .48 A custom bite plate with 2 nonreflective and 4 passive reflective spherical markers is affixed to the patients maxillary dentition to form a rigid system. The non refl ective markers are only used for the isocenter registration. The patients position, which is determined through registration of the reflective marker positions with respect to the isocenter, is tracked by the system in real time. Calibration of the system is performed using a vendor -supplied calibration apparatus. The spatial accuracy of the system is within 0.8 mm9 and the mean registration error is approximately 0.3 mm .9 The kV CBCT system consists of an aSi detector and a kV x ray source installed on tw o retractable arms. The kV x ray beam is orthogonal to the LINAC treatment beam. The unit produces diagnostic quality x rays from 70 to 150 kVp. The active area of the aSi detector is 41 41 cm2 at the nominal detector to -focal spot distance of 155 cm. The scanning parameters and
88 calibration technique used for the CBCT have been previously reported.64 The manufacturer specification for calibration accuracy is a kV / MV isocenter coincidence of 1517 We routinely verify the coincidence to within 0. 3 mm. In this study, the CBCT scans were acquired using a 27.6 x 27.6 x 20 cm3 FOV and a gantry rotation range of 205 over approximately 70 seconds. Description of QA Procedures Mechanical and imaging Many QA procedures of mechanical and imaging items we re recommended by the TG 4060 ,TG14261 and TG 10462 reports. For examples, on daily basis, the laser localization, optical distance indicator (ODI) at isocetner, imaging and treatment coordinate coincidence and positioning/repositioning were suggested. The indicators of couch position (shifts and angles) were verified in monthly frequency. Because three IGSPSs image registrations yield six degree of -freedom parameters and the results are displayed to sub-mm and sub -degree precision, accuracy of those mecha nical and imaging QA items can be cross -verified. The modified DQA3, shown in Figure 5 2, was used to assess the accuracy of LINAC and three IGSPSs. The device was modified by affixing a bite plate with attached reflective markers. The adjustment knobs und er the bottom of device were allowed for rotational movement of the device and the bubble level is embedded in the device. The reference planning CT images of the device were transferred to both FSA and CBCT. The treatment plan isocenter was chosen to coin cide with the intersection of orthogonal lines that pass through three radiopaque markers on a single transverse cut of the device image. The coincidence of imaging and treatment coordinates between systems was assessed. Vendor -provided calibration devices were used for the installations. To determine the inherent variances of the isocenter position, the DQA3 was placed on the couch at the t reatment position,
89 and the FSA system was used to position the isocenter to along all three axes in real time. The reference surface images was captured after the phantom was localized using CBCT (to within 0.3 mm and 0.3 of isocenter) to eliminate the inherent variances between CBCT and ART. The CBCT images and the ART surface images were acquired and registered with the planning CT and reference surfaces, respectively. The CBCT software (v4.0) supports both bony and soft tissue fusion and allows for a u ser defined ROI fusion volume for the registration. The ART software (v4.3) offers 3D surface image fusion in real time with a u ser defined ROI selection. Figure 5 3 displays the image registration screens in two systems. This was done t en times to obtain the statistics of the isocenter position (LINAC, CBCT, FSA and ART ), laser localization, ODI The mean differences between these values presented the coordinate differences between systems were also used as the offset for the repositioni ng study. The accuracy of the repositioning in couch or any IGSPS was evaluated to determine rotational and translational displacements of the isocenter using the DQA3 and comparing the results to those obtained between systems. The phantom was then randomly positioned at over 2 different locations lying within a range of 30 mm of the isocenter. The displacement at each position initially was determined by the ART A CBCT image was acquired at each position and the corresponding displacements were calcul ated using grey-scale registration. The displacements were determined with the FSA system and couch digits number as well. To verify the couch angles, additional 4 data sets were acquired at four couch angles ( 45 and 90 ). The displacements were determined by couch digits and the FSA and ART systems only because the CBCT system can only operate with the couch at 00. This was repeated ten times on ten different days within the span of one month.
90 Dosimetry According to the recommendations from TG 4060 and TG142,61 the daily QA procedures is often operated by a therapist with time constraints. The main items of daily QA procedures are xray and electron output constancy. In addition to checking the output constancy, the use of the DQA3 can help in deter mining other static beam factors which recommended the monthly QA procedures, such as field size indicators, light/radiation coincidences, beam profile and depth dose variations. Measurements were done setting the incident beam normal to the device surface at a SSD of 100 cm and irradiating with 100 MU. In this orientation, the cylindrical chamber located at CAX for output check. The rectangular ion chambers were for the flatness and symmetry check. The curved and circular ionization chambers were for photo n and electron energy verification, respectively. From May 2009 until January 2010, the DQA3 device was used for daily QA procedures in our clinical practices. Results and Discussions Table I shows the isocenter displacement of LINAC and three IGSPSs with the total of 10 measurements. The directional differences along the devices left right, superior inferior, and anterior posterior axes are LAT, LONG, and VERT, respectively, as shown in Figure 5 1. In a previous CBCT study a systematic difference betwe en the kV and MV defined radiation isocenter of 0.3 0.2 mm was observed .16, 17 We therefore selected the XVI as our reference. In TG142 report, the tolerances of imaging and treatment coordinate were suggested IMRT and vely.61 The isocetner displacement of the CBCT presently agrees with those of the other three systems (LINAC, ART and FSA) to within 0.7 mm and 0.6 which demonstrate the coincidences of isocenters between radiation, LINAC and imaging systems. Because it i s not a 6D couch, the displacement of couch displays in translations only. In
91 addition to, due to the similar calibration methods, we found that the isocenters were located uniformly between ART and FSA than CBCT. At the each isocenter location guiding wit h those imaging systems, we inspected the laser alignment which perfectly localized at the central radiopaque markers and slightly below (approximately 0.5 mm) the two side markers within suggested tolerances (1mm for SRS15). The ODI were shown 990 0.5 m m which passed the tolerances (2 mm). Many studies4,5, 16 have reported that the registration errors for the CBCT, FSA, and ART are all within 0.5 mm of one another. The maximum standard deviation of the registration error for our study was 0.4 mm. The means and standard deviations of the repositioning accuracy respect to isocenter between systems with the total of 20 measurements (2 random positions/each measurement) for the 0 gantry and couch setups are shown in Table II. Note that the systematic isocenter differences between the systems were subtracted before doing the comparison. For the mean values, the results show similar results between systems within 0.4 mm which are within the suggested tolerances in 14261 for repositioning accuracy Because of the small rotations (< 1 ), the rotational differences between systems were 0.1 0.3 The means and standard deviations of the accuracy of the couch angles at four couch angles ( 45 90 ,135 and 270 ) are listed in Table III. The rotatio ns in the vertical directions which present the couch angels were within 0.3 0.1 for ART and 0.0 0.1 for FSA The values are within the TG 142 recommendation of 1 for the table angle in IMRT treatment.15 T he rotational variances in longitudinal an d lateral direction were all less than 0.5 in both systems which could be caused by the mechanical isocener of couch. The dosimetry results following the recommended daily QA procedures of LINAC are tabulated in Table IV including output, energy, symmetry and flatness of dose profiles, field size
92 and light radiation coincidences. The software of this device supports to re cord automatically each measurement and to view the long term stability of dosi metry in LINAC as shown in Figure 5 4. Users have the option to choose which data parameters to view, and over what period to view them. A single measurement instance can be exa mined by clicking on a data point which corresponds to a particular day. For over 8 months, we can verify that the output is within 2%, photon and electron energy are within 2% and the symmetry and flatness are within 1%. The QA device is able to detect 1 mm field size error on a daily basis. The data showed that field size is within 1mm with approximately 1mm standard deviation and light radiation coincidence is within 1 mm. In summary, the readings are within the AAPM recommendations .14, 15 In our clinical practices, the total measurement time to finish the daily QA tasks is within 20 minutes including setting up device (< 3 minutes) to isocetner, verification of isocenter coordinates between LINAC and three ISPSs (< 5 minutes), couch repositioning (two random in translation and two angles) (< 10 minutes) and dosimetry of LINAC (photon and two electron energy) (< 2 minutes).We design the template which lists above QA items for therapists. Conclusions The DQA3 can be used reliably not only for daily output an d energy constancy, as well as profile field size indicator, lightradiation coincidences of LINAC but for mechanical couch and imaging geometry (coordinates, positioning and repositioning). Applied this device, the daily QA procedure can become efficie nt, simple and extensive.
93 Table 5 1. Isocenter displacements between systems respect to CBCT explored with D QA 3 Table 5 2. Comparisons of positioning variances between systems at couch 0 within 30 mm Table 5 3. Comparisons of the accuracy of couch angles between systems at 4 couch angles. AlignRT3C v.s. XVI FSA v.s. XVI LINAC v.s. XVI Trans(mm) Rot ( ) Trans(mm) Rot ( ) Trans(mm) Rot ( ) LAT 0.2 0.4 0.6 0.1 0.1 0.2 0.6 0.1 0.1 0.3 LONG 0.2 0.2 0.5 0.1 0.0 0.4 0.6 0.2 0.4 0.2 VERT 0.5 0.4 0.2 0.1 0.5 0.2 0.3 0.2 0.6 0.3 Vectors 0.6 0.3 0.5 0.2 0.7 0.3 AlignRT3C v.s. XVI FSA v.s. XVI Couch v.s. XVI Trans(mm) Rot ( ) Trans(mm) Rot ( ) Trans(mm) Rot ( ) LAT 0.3 0.5 0.1 0.2 0.0 0.2 0.0 0.2 0.0 0.5 LONG 0.2 0.4 0.0 0.1 0.1 0.4 0.0 0.2 0.1 0.5 VERT 0.2 0.5 0.1 0.3 0.2 0.5 0.1 0.3 0.3 0.7 Vectors 0.4 0.5 0.2 0.4 0.3 0.7 Couch 450 Couch 1350 AlignRT3C FSA Couch AlignRT3C FSA Couch LAT 0.0 0.1 0.5 0.1 0.0 0.2 0.2 0.3 0.1 0.0 LONG 0.3 0.2 0.5 0.4 0.0 0.1 0.3 0.1 VERT 0.3 0.1 0.0 0.0 0.0 0.2 0.0 0.0 Couch 900 Couch 2700 AlignRT3C FSA Couch AlignRT3C FSA Couch LAT 0.0 0.2 0.5 0.1 0.0 0.2 0.2 0.1 0.2 0.0 LONG 0.3 0.1 0.5 0.4 0.2 0.1 0.1 0.2 VERT 0.2 0.1 0.0 0.1 0.0 0.1 0.0 0.1
94 Table 5 4. Dosimetry results for daily QA procedures of the LINAC over 8 months
95 Figure 5 1. Schematic of the hardware setup for 3 IGSPSs in the treatment room Figure 5 2. Bite frame attached to DailyQA3TM device QA phantom
96 A B C Figure 5 3. Examples of registration process A) CBCT, B) FSA, and C) AlignRT3C Figure 5 4. Daily QA3 s oftware provides a graphical presentation of data for each template.
97 CHAPTER 6 DOS E EVALUTION OF ROTATIONAL ERRORS Introduction Due to tremendous improvements in radiotherapy over the last few decades, therapy plans can be devised to precisely target a tumor, but the efficacy of radiotherapy can still be compromised by errors in the treatment set up of the patient which affect the delivered radiation. With IMRT in particular creating steep dose gradients and tight margins around the tumor there is a greater risk of dose to surrounding OARs due to set up errors and organ motion IGRT, w herein imaging devices are used at treatment de livery to increase the probability that radiation is delivered as closely as possible to the original plan ,19 can help avoid these problems. The practice of imaging treatment fields to localize the tumor has been carried out for several decades and has bee n referred to as MV portal imaging. MV portal imagine has been performed using 2D X ray detectors to verify orthogonal localization fields. It only uses bony structures to a ssess the setup deviations in 2D portal images because soft tissues are difficult t o visualize in the planar projection X -ray images.6 Nevertheless, computed tomography (CT) images are capable of identifying both bony structures and soft tissues and CT is the standard reference to delineate organs and the target in TPSs. Additionally, it is not surprising that volumetric CT registration in six DOF performs more accurately than bony-structure registration based on 2D portal images.6 There is a growing interest in CT based imaging system for 3D volumetric localization. The most common of t he sophisticated image guided approaches is the CBCT scanner because of its complete set of 3D volume information with a high resolution and accurate image registration.6 8,12, 21 Furthermore, i t has been shown that the CBCT system sub-mm and rotational setup errors can be correctly determined in six DOF registrations, which is suitable for high precision treatments like stereotactic radiosurgery .7, 8 Although the standard treatment table does
98 not allow rotational corrections, 6 corrections only can be achieved with a robotic couch,22, 23 or by adjusting the collimator and gantry angles.25, 26 The controversial conclusions and guidance on rotational errors management were also discussed .28 The post or near real time verification 22, 23 after corrections is essential. Additional verification time, dose, and setup accuracy could be compromised by each other. Therefore, currently, most setup adjustments are applied to the translational directi on only, and thus rotational positioning errors still exist throughout the patient's treatment. M athematically, a 2 rotational deviation induces a translational deviation of about 1 mm at a point located 3 cm from the isocenter. For elongated targets (> 5 cm long), with rotational deviations greater than 2, the translational deviation is more than 1.8 mm, which might be significant when a tight margin is needed.12, 27 Guckenberger et al .12 discovered that the maximal rotational errors using kV CBCT were 5 (head and neck), 8 (pelvis), and 6 (thoracic). According to daily MVCT from 3,800 tomotherapy treatments21, at least 5% of brain patients had more than 3 in roll rotations. Those rotational positioning errors resulted in decreased target coverage and increased dose to the OARs. However, m any factors might contribute to the dosimetric effect, including the simulation methods for those uncorrected errors, degree dose gradient steepness, margin sizes, evaluation parameters, treatment techniques, and treat ment sites. The quantitative correlation between the amount of rotational errors and its dosimetric consequence is not obvious and there is a wide range of dose effects.12,27 32 For example, during simulation, the simplest approach is rotating the gantry f or roll and couch for yaw.27 30 32 Yue et al .24 developed a method to implement all 6 DOF corrections, which commonly was used for rotation simulations in 3 axes.28 29Additionally, for online corrections, both patient repositioning and plan adjustments have been proposed and treatment CBCT images were used to evaluate
99 dose impacts on setup errors .12 31 At present, it is still a challenging task to accurately delineate t he tumor and organs and calculate the dose using CBCT images .33 The purpose of this study was to propose direct simulation in a TPS to evaluate the dose consequences for systematic rotational setup errors as the standard dose evaluations in fractionated SR T with IMRT. The simple and direct simulation method was achieved by directly rotating the original CT image set and contours of targets and OARs with the same coordinates. We examined the accuracy of this method by registering the image sets and calculati ng the volume differences from the contours. Therefore, all dose distributions were done in the TPS by using those rotated image sets and contours. For a group of SRT patients, the simulated dose distributions with rotational errors were retrospectively ca lculated and compared with the original plans using DVH parameters to investigate the dosimetric effects. Methods and Materials Patient Selection Ten SRT patients with intracranial tumors who had undergone IMRT between August 2008 and April 2009 were selected for the present study. We selected patients whose targets were located at different areas with different sizes, shapes, and diameters to be able to study the dosimetric impacts of rotational setup errors in these situations. Table I shows the char acteristics of the patients, the tumors, and the prescribed treatments. Since we wanted to focus on larger targets, we selected patients with a glioblastoma, oligoendroglioma, astrocytoma, and meningioma as well as a gross target volume (GTV) larger than 1 5 cm3. Histologies and staging were determined by clinical evaluation. CT or magnetic resonance imaging (MRI) of the brain was obtain ed as clinically indicated. Figure 6 1 displays the cropped CT images of the tumors and adjacent OARs for each patient. As shown in Figure 6 1, the shape and size of the targets varied dramatically from patient to patient. To study the effect of the planning target volume
100 (PTV) shape on the criteria for plan inter -comparison, we classified the target shapes into 3 categories: sphere, ellipse, and irregular. All glioblastmoas were shaped regularly or elliptically. The base -of -skull meningiomas were all irregularly shaped, mostly because of extensions in the cavernous sinus and through foramina, therefore the targets were irregul ar. The mean PTV volume and diameter were 237 cm3 (range, 90 cm3 377cm3) and 9.3 cm (range, 7.2cm 12.2cm), respectively, to study the effect of the target size and shape. Simulation and Target Delineation Each patient was scanned by CT and laid with a custom disposable head support (MoldCare Pillow, Bionix Inc, Toledo, OH) Two immobilization devices were used for SRT scanning protocols; one was a bite block using teeth impression combined with a half -face thermal plastic (TP) mask and the other was a full -face TP mask only for patients who could not tolerate the bite block. A planning CT scan for each patient was made with a 2 -mm slice thickness using a Philips Brilliance 16 -slice scanner (Philips Healthcare, Andover, MA). All patients also had an MRI scan (voxel size 1.1 mm3 x 1.1 mm3 x 1.3 mm3; T1 -weighted with intravenous gadolinium and T2 -weighted) of the brain. The CT isocenter was marked on all patients with 3 tattoos outside of the mask system. Co registration of CT and MRI images and all contouring were done on Syntegra software in the Philips Pinacle3 TPS ( Philips Medical Systems, Madison, WI) In glioblastoma patients, the clinical target volume (CTV) was formed by expanding the GTV by xx mm, except in regions where natural boundaries precluded microscopic tumor spread. In meningioma or oligoendroglioma patients, we did not use any GTV -to -CTV margin, assuming no microscopic spread beyond the visible tumor. Once the CTV was created, areas of overlap with bone or uninvolved OARs were excluded fr om the volume. Uninvolved lymph nodes were not included in the CTV. The CTV was then uniformly expanded 1 cm to create the PTV. For patients in
101 whom the CTV was in close contact with OARs, the PTV margin was adapted manually to avoid overlapping of the PTV and these OARs. However, in our situation, the positioning accuracy of SRT setup was assessed by CBCT to allow the use of 3 -mm setup margins between the CTV and PTV.4 8 To impartially evaluate the dose effects using the sam e margin for each patient, CTV wa s defined in this study by the uniform subtraction of a 3-mm margin from the PTV. OARs included brain stem, chiasm, optical nerves, retina, cochlea, and lens. After contouring the tumor volumes and OARs, an IMRT plan was designed on a Pinacle3 (V 8.1) TPS Manually optimized beam directions for the IMRT plans were used only when it was not possible to keep the volume of brain stem receiving 50 Gy to < 0.1cm3 using a 3D conformal radiotherapy plan. Four to 9 non-coplanar beams yielded a result that was opti mal, combined with a realistic treatment time. The dose prescription in all cases was 61.2 Gy with conventional 1.8 Gy fractions delivered daily 5 days a week for 6 to 7 weeks. Because of the high conformity of the IMRT plans, an evaluation using the presc ription dose only was inappropriate; therefore, the 3 tumor -volume (CTV and PTV) coverage at all prescription dose levels (volume receiving 100Rx], 93% [V93Rx]. and 110% [V110Rx] prescription dose levels, respectively) were evaluated. The PTV coverages aimed for were V100 93 and V110 cm3: V55Gy (volume receiving 45Gy for the retina and cochlea; and V12Gy for the lens. Calculations of the DVHs were done with a grid size of 2 mm. Simulation of Patient Rotation with Respect to Isocenters The dosimetric effect of systematic rotational setup errors on each patients treatment plan was evaluated by simulating multiple rotational setup errors and re -computing the dose distributions in the TPS. Dose consequences were analyzed with a conservative assumption that
102 systematic errors existed continuously throughout the treatment. Systematic anatomy rotations and residual translations were simulated using a software program developed in MATLAB language. The patients anatomy was represented by a colle ction of points in a coordinate frame. The group of anatomical points in the original CT image set and contours by the coordinate matrix were designated as [O]. The software supported the commercial TPS (ADAC Pinacle3) format without coordinate conversions The coordinates of the same set of anatomical points in the rotated image set and contours (designated as [OR]) were created after software simulation. For the rigid body, one assumes that the [O] is congruent with the [OR]. Ignoring anatomical deformati on in the intracranial treatment site, the difference between the [O] and [OR] can be completely represented by a 3 x 3 rotation matrix [R] and a translation vector T, as shown Eqs.1 through 3, to simulate patient treatments in which a certain extent of po sitioning error existed. [R] shows the combination of rotations in pitch, roll and yaw. [ ] = [ ] [ ] + ( 1 ) [ ] = [ ] ( 2 ) = 1 0 0 0 0 = 0 0 1 0 0 = 0 0 0 0 1 ( 3 ) All simulated rotations were about the beam central axis and the isocenter, shown in Figure 6 2. In patient treatments, the maximum rotational errors (range, 5 8) reported .50 [O] were simulated in 3 axes by combining rotations ( 1 3 5 and 7 ) and translations (1 mm), which were used to present residual errors after CBCT corrections. After simulating the process in the software, [OR] were re imported into TPS to replace [O]. For each patient chosen for this study, the original plan was regenerated 8 times using the [OR] and the same fluence, sequences, and isocenters of each IMRT field as the rotated plan. The dose distributions calculated on
103 each rotated plan using the same total monitor unit represent the distributions with the rotational setup errors. Dosimetry Study A total of 90 plans were analyzed for the 10 patients (1 original and 8 rotated plans for each patient). Before calculating the dose, to validate the [OR], rotated images were registered with the original images using Syntegra in the TPS and volume differences between rotated and original contours were calculated. In each plan, the dose distribution was calculated using a collapsed -cone convolution algo rithm on a 2 -mm grid. The DVHs of the rotated plans for the t arget volumes (CTV and PTV) and 6 OARs were evaluated with respect to the planning objectives outlined above. These data were compared with those obtained from the original plan. Results Validati on of Simulated Images and Contours The accuracy of simulating rotational setup errors was validated by all [OR]. For the rotated image sets, Table II shows the isocenter transformation of each image set after registering with the original image set using Syntegra registration software in the TPS. The mean simulation errors were 0.3 0.4 mm and 0.1 0.2 in any axis and the maximum errors were 0.8 mm/0.4 for all 80 image sets (8 rotated image sets per patient). For rotated contours, the volume difference s by comparing to the original contours were used to validate the accuracy of the simulation method. Table III shows the details of the volume differences in absolute and percentage values for all targets and OARs. Eight rotated contours were examined per target/OAR. The overall volume of contours and the maximum error in the absolute differences was 2.2 cm3, which happens in targets with larger volumes than OARs. By comparison, the percent volume differences were approximately 0.5% for all 3 targets, and < 2% for OARs with volume size > 1 cm3. The maximum percentage errors were < 5% when the volume was < 1 cm3.
104 Target Dose Assessments and Comparisons All 10 patients were treated to 61.2 Gy in 34 fractions of a 180 cGy fraction dose to the PTV coverage > 95% and analyzed for dose variation. Table IV lists the absolute coverage and coverage changes (relative) from V100Rx and V93Rx for the 3 targets in the original (0 ) and rotated (1 ~7 ) plans. V100Rx showed greater discrepancies than V93Rx because of th e high dose gradient steepness. Therefore, rotational errors caused insignificant coverage variations in V93Rx for all cases (within 3%, except for the PTV with 7 errors). For patient 1 with an irregular shape of target, significant losses would have res ulted in the V100Rx and V93Rx (up to 19% and 5%, respectively) for the CTV if the ro tations were not corrected. Figure 6 3(a) illustrates how the V100Rx changed for 3 tumors with rotations. The amount of mean coverage decreased, but SD increased with incre asing rotational angle. The rotations adversely affect the coverage of PTV. The mean coverage (V100Rx) to the PTV in the original and rotated plans were 96.91.1% (original), 94.62.6% (1 ), 90.35.3% (3 ), 82.49.4% (5 ), and 71.115.5% (7 ), resp ecti vely. The coverage of CTV was less than 95% when rotational errors were more than 3 The coverage of PTV in V100Rx was much less than the CTV with 3 -mm setup margins if rotational errors went uncorrected (up to 71% for the PTV versus 81% for the CTV ). Th e m ean coverage change in the CTV for those rotated plans were 0.8% (1 ), 3.5% (3 ), 9.3% (5 ), and 17.8% (7 ), respectively. Fig ure 6 3(b) shows the coverage loss (V100Rx) of CTV for each patient. Within 3 rotations, the insignificant impact (withi n 5%) on coverage loss of the CTV was found in most cases. However, a substantial amount of coverage change in the CTV was observed in patients 2, 3, and 4 (up to 9% with 3 21% with 5 and 38% with 7 ). The data in Figure 6 3(b) reveals that the unlikely results consequences in tumor coverage happen between clockwise and counter -clockwise rotations. For example, for Patient 4, the decreased
105 coverage of CTV only happens in the clockwise direction (25% [+7 ] versus 0.3% [ 7 ]), but the decreased occurred in the counter -clockwise direction for Patient 8 (0.4% [+7 ] verse 5% [ 7 ]). OAR Dose Assessments and Comparisons Table V and Figure 6 4(a) illustrates the changes in volume receiving more than the tolerance dose for the OARs i n the rotated plans. In general, all OARs had no change (< 0.1 cm3) in most cases except for the brain stem. The SD of the volume changes was > 0.1 cm3 only when patients had more than 7 rotations. The brain stem had the most dramatic mean change (range, 0.20.3 cm3 [1 ] 0.41.3cm3 [7 ]). The original plan mean volume of organs receiving the tolerance was less than 0.1 cm3. Figure 6 4(b) shows the V55Gy changes for the brain stem for each patient. Within 3 rotations, the insignificant impact (within 0.1 cm3) was discovered in most cases. Furthermore, the wide range of changes between those 10 patients was up to approximately 4 cm3. Two patients (Patients 6 and 7) experienced a significant increase in the V55Gy with all rotation angles (up to 0.9 cm3 with 1 2.0 cm3 with 3 2.8 cm3 with 5 and 3.7 cm3 with 7 ). Additionally, the positive V55Gy values (3.7 cm3 [+7 ] versus 1.4 cm3 [ 7 ]) were found for Patient 6 when patients rotations were clockwise, but for Patient 7 when they were in the counter -clockwise directions ( 1.4 cm3 [+7 ] verse 3.0 cm3 [ 7 ]). Discussions For intracranial -cancer patients treated with IMRT, the above results have demonstrated that dosimetric deviations increase with increases in rotational setup errors. For syst ematic rotational setup errors up to 3 ty of coverage loss to the CTV > 5% and of increased volumes receiving the tolerance dose to most OARs > 0.1 cm3 is small. While the results showed that the rotational setup error indeed degraded the d osimetric coverage in SRT, the degradation caused by a 3 setup error was not significant for IMRT and 3 -mm margin sizes.
106 It is generally accepted that systematic rotations less than 3 do not lead to clinically significant dose consequences. Therefore, ou r institutions practice applied 3 as the setup action level for rotational corrections. Overall, our data illustrated the influence of rotational setup errors alone for SRT and the greater dosimetric effects from clockwise (+) versus counter -clockwise ( ) rotational displacements. The effect appeared to be associated with the relative location of the tumor and OARs, the shape and size of the tumor and OARs, and the relative location of the isocenter in the tumor OAR geometry. Larger targets and those wit h irregular shapes might be more susceptible to rotational setup errors .54 The more significant dosimetric effect of patients rotations was expected in 5 of 10 patients, because of the irregular and/or ellipse tumor shapes ( case 1 3) which were located in parietal area of the brain. Furthermore, for case 3, the combination of larger tumor size, elliptical shape, tumor in the parietal area, and isocenter away from the center of the PTV contributed to the most consequential dosimetric impacts on rotational e rrors regardless of directions and angles (V100Rx for CTV decreases by approximately 5% with 1 rotations). It is best to control the smallest systematic rotations in clinical patients who have those characteristics. Our definition of the CTV included the GTV and a few margins (or no margin) which covered the malignant tissue, depending on the histologies. Institutional definitions of CTV could vary and, in our study, case 2 didnt even have a CTV. Since the range of PTV is added to setup margins from the CTV depending on localization and immobilization, the CTV can be a more critical concern than the PTV. In our clinical practice, we performed quality assessments using CBCT and the mean setup errors were within 3.2 mm for the SRT patients.48 Therefore, to have an objective dose e valuation, we studied the CTV excluding 3 -mm uniform margins from
107 the PTV rather than the CTV. In addition, we also found that the volume ratio of PTV to CTV was 2.1 for case 5 and 13.3 for case 8. The results of 2.2 and 10.4, for case 5 and 8, respectively, were shown in the similar V100Rx ratio of PTV to CTV when they had 7 rotations. However, the volume and V100Rx ratio for PTV to CTV became equal (approximately 1.4 -1.6) in those 2 patients. We choose CTV to evaluate the dose effects. Because of 3 mm setup margins, the setup rotational errors could have a different effect on the dose to the target and OARs than what was seen in our study. For all 6 OARs analyzed, the deviations in the volume receiving the tolerance dose were within 0.1 cm3 for all rotated plans. The only exception was the brain stem because of its large volume. Between 10 patients in Fig ure 6 4(b), the V55Gy values of the brain stem were changed dramatically with increasing rotational error s in cases 6 and 7. In addition, D0.1 (the dose to 0.1cm3) was observed in the original plan for more than 60 Gy when the PTV was adjacent to structures for those 2 patients. Both patients also had elongated tumors, so a part of the OAR could have been eas ily pushed into a higher dose region. Fig ure 6 4(c) displays fewer influences in the brain stem for cases 5 and 10 because D0.1 was approximately 54.7 Gy and the tumor was spherical. Therefore, D0.1 for the brain stem could be the effective check point to setup the higher action level for rotational corrections when patients experience systematic rotational errors. The direct simulation we presently proposed was derived with the assumption that patients were rigid bodies. We had validated [OR] using image r egistration and the volume size of the contours in the TPS. For rotated images, Table II shows greater SDs in translations (0.5 mm) than rotations (0.2 ). Because we simulated only a 1 -mm translation but up to 7 in rotations, the pixel size of the CT ima ges was approximately 0.8 mm, which contributes more to the registration errors in 1 -mm translations than rotations. In contrast, the voxel size of the contours
108 was approximately 0.6 mm in the TPS. The maximum percentage differences in the cochlea were approximately 4.9% because the contour volume was less than 0.1 cm3, although those differences were <2% in other OARs. The pixel (voxel) size was the factor that contributed to the uncertainties during the simulation process. While the inherent simulation uncertainties were negligible, this rigid body assumption was a first -order approximation to the problem; therefore, the results may deviate from the reality since patient bodies are deformable. In our study, we performed the simulation by rotating CT images and contours for the systematic rotational setup errors, which provided the most accurate dose results and true treatment geometries between the tumors and OARs. Rotating the gantry f or the roll and the couch for th e yaw is the simplest method, but no pit ch direction is the limitation for yaw .27 30 32 Therefore, this method cannot simultaneously perform 3 rotations; further, the inherent coordinate differences between the treatment machines and patient anatomies could be the main uncertainties in the dose evaluations. To eliminate the limitation, most studies used the method developed by Yue et al.24 with the rigid body assumption to simulate any patient setup error. This method involved 3 matrix transformations and the isocetner shift, and the gantry, couch, and collimator were adjusted to implement th e full 6 DOF corrections. However, it is an indirect method by applying transform matrices for the coordination conversion. The beam reestablishment error from the matrices, machine operation limits, and collision constraints were reported.2 4 It is essential that the phantom study with a set of fiducial markers was used to verify for the accuracy of beam reestablishment. In contrast, this direct simulation method in our study provides the simple verification and spatial visualization for the patient and tum or -OARs geometry. The dose impacts on rotational setup errors using this direct simulation in a TPS could be the most reliable method and become the gold standard. There is only one limitation: once the
109 significant dose influences caused by rotational erro rs were found in patients, those errors only could be corrected using a robotic couch,13, 49 not Yues methods.24 The situation might vary for different treatment sites and techniques. For head and neck sites, the body contour may change considerably from one point to another, and the targets are large and irregular, which might make the patient more susceptible to rotational setup errors. Recent studies found that rotational errors contribute to treatment uncertainties for head and neck IMRT patients .13 50 Kim et al.27 studied the dose impacts of head roll setup errors during head and -neck IMRT treatments to the spinal cord and found dose change were 3.1% and 6.4% for 3 and 5 respectively. Furthermore, even in prostate treatments when tumors were spherical, the impact of rotational setup errors may be dissimilar if the treatment technique, margins, and simulation methods are different. Sejpal et al .32 evaluated the dose changes in prostate cancer patients treated with proton therapy by rotating the gantry and couch to simulate the roll and yaw directions. They found no significant dose changes to targets and OARs when patient rotational movements were less than 5 Fu et al.28 studied the same topics using Yues method24 for patients receiving 3 -phase sequential boost IMRT treatments. They reported that, for systematic rotational setup errors up to 3 2% is small and less than 30%. For seminal vesicles, slightly larger influences were observ ed. In our study, future studies will focus on different treatment sties and techniques using the direct simulation. Conclusions The primary goal of this work was to identify rotational tolerances for SRT patients treated with IMRT. The direct simulation method in a TPS was applied to these patients and we found excellent agreements with the original images and contours for the isocenter transform and
110 volume sizes, respectively. The results show that rotations are problematic for cases with irregular, elli ptical targets, and a tolerance of 3.0 should be set for patients with 3 -mm setup margins. This study reveals that rotational tolerances are suggested and that the dosimetric implications using direct simulation in a TPS are of clinical importance.
111 Tab le 6 1. Summary of chara cteristics of the patients, tumors, OARs and prescribed treatments Table 6 2. Isocenter transforms of rotated CT image set after registration process Abbreviations: Trans.=Translations, Rots.= Rotations
112 Table 6 3. Absolute and percentage differences in volume between rotated and original OARs Abbreviations: Mean V.=Mean volume of original OARs Table 6 4. Effect of rotational setup errors on target coverage in IMRT for SRT Abbreviations: Abs.=absolute coverage. Rel.=relative coverage respect to original plans.V100Rx, V93Rx=minimal volume received by 100 %, 93% of prescribed dose, respectively. Table 6 5. Relative volume differences between rotated and original plans on sparing of OARs V55Gy V45Gy V12Gy Brain Stem Chiasm Opt.N Retina Cochlea Lens Mean V. 27.63.4 0.60.4 1.00.5 5.50.9 0.080.03 0.30.2 1 0.2 0.3 0.0 0.1 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 3 0.2 0.6 0.0 0.1 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 5 0.3 0.9 0.0 0.1 0.0 0.1 0.0 0.1 0.0 0.0 0.0 0.1 7 0.4 1.3 0.0 0.1 0.0 0.1 0.1 0.3 0.0 0.0 0.1 0.2 Abbreviations: V55Gy, V45Gy, V12Gy =volume received by 55Gy, 45Gy and 12Gy, respectively.
113 Figure 6 1. Tumor sites and shapes for the 10 cases. Green: CTV, blue: PTV (= CTV+3 mm) Figure 6 2. Comparison of the original (upper row) and rotated (lower row) images for case 8
114 A B Figure 6 3. Changes of tumor coverage in the V100 Rx A) all cases B) each case 94.1 96.3 98.1 98.7 99.3 98.4 95.8 90.0 81.4 88.1 91.3 94.4 95.9 96.9 94.6 90.3 82.4 71.7 50 55 60 65 70 75 80 85 90 95 1007 5 3 1 0 1 3 5 7V100Rx(%) Rotational angle ( ) CTV (V = 162.6 85.4 cm3) PTV (V=232.5 103.8 cm3) 40 35 30 25 20 15 10 5 0 7 5 3 1 0 1 3 5 7 V100Rx Rotational angel ( ) case#1 case#2 case#3 case#4 case#5 case#6 case#7 case#8 case#9 case#10
115 A B Figure 6 4. Changes in volume receiving the tolerance dose. A) all OARs. B) brain stem only 0.00 0.10 0.20 0.30 0.40 7 5 3 1 0 1 3 5 7 VDiff. (cm3) Rotational Angles ( ) Brain Stem_V55Gy (V=27.6 3.4 cm3) Chaism_V55Gy(V=0.6 0.4 cm3) Optical Nerve_V55Gy (V=1.0 0.5 cm3) Retina_V45Gy (V=5.5 0.9 cm3) Cochlea_V45Gy( V=0.08 0.03 cm3) 2.0 1.0 0.0 1.0 2.0 3.0 4.0 7 5 3 1 0 1 3 5 7 V55Gy(cm3)) Rotational Angle ( ) 1Moore 2Coleman 3Janes 4Holt 5Allen 6Thortsen 7Hiler 8Cobb
116 CHAPTER 7 SUMMARY AND CONCLUSIONS The five key issue in HPRT deals with how to: 1) perform the quality assessments of current setup systems, 2) develop the new IGSPS to replace current systems 3) evaluate in clinical procedures applying the new IGSPS and suitable immobilization s, 4) establish the QA program for LINAC and all IGSPSs and 5) simulate the dosimetric effects due to daily setups in TP S A quality assessment of the FSA system for intracranial SRT was performed using CBCT. FSA setup errors for two groups of patients were analyzed : (1) FSA -positioned patients immobilized with a bite plate and TP mask (the bFSA group);(2)Room laser -positioned patients immobilized using a TP mask (the mLAS group). A QA phantom was used to study the system differences between FSA and CBCT. The quality assessment was performed using the Elekta XVI and the Varian OBI for 25 patients. For the first 3 fractions, and weekly thereafter, the FSA system was used for patient positioning, after which CBCT was performed to obtain setup errors. 1) Phantom Tests: The mean difference in the isocenter displacements for the two systems was 1.2 0.7 mm. No significant variances were seen between XVI and OBI (p~0.208); 2)Patient Tests: (a) The mean of the displacements between FSA and CBCT were independent of the CBCT system used; (b)Mean setup errors for the bFSA group were smaller ( 1.2 mm) than those of the mLAS group (3.2 mm)(p< 0.005). For the mLAS patients, the 90th percentile and the maximum rotational displacements were 3 and 5 respectively. A 4 mm drift in setup accuracy occurred over the treatment course for 1 bFSA patient. System differences of >1 mm between CBCT and FSA were seen. Error regression was observed for the bFSA patients using CBCT (up to 4mm) during the treatment course. For the mLAS group, daily CBCT imaging was needed to obtain acceptable setup accuracies.
117 The ART system is an IGSPS that provides real tim e target localization. The study involves the first use of this system with three camera pods. I ts localization accuracy and tracking ability has been evaluated using a CBCT system and an optical tracking system. A modified Rando head and neck phantom and five volunteers were used to evaluate the calibration, registration, and position -tracking accuracies of the ART system and to study surface reconstruction uncertainties, including the effects due to inter and intra -fractional motion, skin tone, room light level, camera temperature, and ROI selection. Results for two different reference image sets; CT_S and ART_S are reported. System accuracy was validated through c omparison with the XVI and the FSA. The system origin displacements for the ART and XVI syste ms agreed to within 1.3 mm and 0.7 Similar results were seen for ART versus FSA. For the phantom displacements having couch angles of 0 those that utilized ART_S references resulted in a mean difference of 0.9 mm / 0.4 with respect to XVI, and 0.3 mm / 0.20 with respect to FSA. For phantom displacements of more than 10 mm and 3 the maximum discrepancies between ART and the XVI and FSA systems were 3.0 mm and 0.4 mm, respectively. For couch angles up to 90 the mean (Max.) difference between the ART and FSA was 1.2 (2.3) mm / 0.7 (1.2 ). For all tests, the mean registration errors obtained using the CT_S references were approximately 1.3 mm / 1.0 larger than those obtained using the ART_S references. The mean inter and intra -fractional registration differences between ART and FSA were within 1.3 mm / 0.5 ART can be used as a nonionizing IGSPS with accuracy comparable to current image/marker -based systems. ART and CBCT can be combined for high -precis ion positioning without the need for patient attached localization devices. To evaluate ART in the clinical study as a feasible IGSPS for frameless SRT with non coplanar fields, it provides accurate daily target localization in real -time without additiona l dose
118 and compares results with a CBCT system and an optical tracking system. Five intracranial cancer patients undergoing intensity-modulated radiotherapy were studied. The system accuracy was perform ed using the ART XVI and FSA For each session, the patient was positioned using thermoplastic chin stretch for immobilization and FSA in real time tracking for target localization with setup tolerances (0.3mm/0.3 ). CBCT images were acquired for the first 3 fractions and weekly thereafter. Setup surface ima ging data sets for evaluation of pre -treatment, inter -fractional and intra -fractional motion were obtained daily to compare XVI and FSA The positioning differences variations in the translational and rotational directions among three systems were analyzed Results for two different reference image sets (CT_S, and ART_S are reported. The mean differences in the pre treatment displacements for the ART and XVI systems was 0.3 mm/0.2 but 1.3 mm/1 for the FSA and XVI systems. For the displacements of the inte r -fractional motion having non -coplanar beams with respect to FSA, those that utilized ART_S references resulted in a 90th percentiles of difference within 2.1 mm / 0.8 using ART_S and 3.3 mm/0.3 using CT_S. For the displacements of the intra -fractional motion over maximum 14 mins tracking time, the accuracy is within 1mm/1 Bite tray uncertainties over the treatment course for one patient up to 3 mm/2 were occurred. The combination of gantry, panel and tube blockage effects during the CBCT acquisition caused approximately 3 mm registration errors. No skin tone and deformation effects were found. ART can be used as a non -ionizing IGSPS with accuracy comparable to current image/marker -based system s. ART and CBCT can be combined for high-precision positioning without the need for patient attached localization devices. To develop a simple yet comprehensive effective daily QA program for LINAC and 3 IGSPSs and to summarize the results of these QA test s over extended periods. The daily output
119 check on a linac unit is commonly performed utilizing a chamber embedded in the phantom. Currently, 3 IGSPSs, the ART surface system; the FSA and the Elekta kV CBCT were installed in the same treatment room with an Elekta LINAC For daily QA, therapists routinely performed dosimetric checks of the linac and geometric accuracy tests of the 3 IGSPSs by repositioning different phantoms provided by the vendors. We investigated the use of a commercially available DQA3 ph antom to perform this task. The QA program encompasses 3 components: dosimetry, mechanical (couch and laser), and imaging geometry. T he coincidence of the linac coordinate system for three ISPSs was within 0.7 mm/0.6 of that of the CBCT system due to diff erent calibration methods. Guiding with those imaging systems, we inspected the laser alignment, which perfectly localized within 0.5 mm. The ODI were 999 0.5 mm. For gantry and table angles of 0 the mean displacement vectors were within 0.4 mm and 0.1 in an axis between systems. The differences of the couch angles using ART and FSA was within 0.3 For over 8 months, we were able to verify that the output was within 2%, the photon and electron energy was within 2%, and the symmetry and flatness were w ithin 1%. The data showed that the field size and light radiation coincidence are within 1mm 1mm. The total measurement time of all tasks in one section takes less than 20 minutes. The use of DQA3 simplifies the task of daily dosimetric consistency check s, shortens the set up time, and eliminates the systematic errors by switching phantoms. Clinical use of this multi-detector device has improved the efficiency and thoroughness of daily QA programs for linacs and IGSPSs. By considering dosimetric relevance and clinical flow, we have established a reasonable action level. To determine dose -delivery errors resulting from systematic rotational setup errors for fractionated SRT using direct simulation in a TPS. Ten patients who received IMRT of brain tumors wer e studied. Simulation of systematic rotation was done by applying a 3x3 rotation
120 matrix to the original images and contours. Combined rotational errors of 1 3 5 and 7 and residual translation errors of 1 mm along each axis were simulated. The rot ated images were re imported into the TPS and the dose effects were evaluated by re -computing the dose distributions. Target V100Rx, V93Rx, and V110Rx of the prescription dose level and the volumes of the 6 OARs receiving more than the tolerance dose levels were evaluated. The accuracy of the image rotation algorithm was validated using the image registration package within the TPS and was found to be within 0.5 mm/0.2 in translation/rotation transformation, res pectively, and the volumes of the rotated contours are within 2% of the original contours. A systematic reduction in CTV coverage was observed, with the mean V100Rx of the CTV to be 99.30.5% (original), 98.41.2% (1 ), 95.83.0% (3 ), 90.07.3% (5 ), a nd 81.413.1% (7 ). In most cases, OAR doses fell within the tolerance levels. However, for two cases where D0.1cc of the brain stem received close to the dose limit in the original plan, even a 1 rotations had made D0.1cc > 55Gy, which exceeded our dos e limit. We conclude that for rotations less than 3 no action level is required; however, cases presenting extreme rotations and high doses in the brain stem should be investigated as routine practice. This procedure allows easy evaluation of dosimetric consequences when systematic rotational errors are present in patient setup. In summary as shown in Figure 7 1 t his dissertation initially performed a quality assessment of FSA using CBCT to investigate the overall accuracy of setup systems and comprehe nd the limitations in SRT. It also reports the optimization of the accuracy of applications by combining three IGSPSs Besides localization systems, modified immobilizations were considered suitable for clinical implementation. To avoid removing the phanto m variety and to simplify the daily QA procedure, a simple yet comprehensive daily QA program was designed by the commercial device for LINAC, CBCT, FSA and Align RT3C
121 systems. Finally, a reliable and accurate dosimetric simulation was developed to evaluat e dosimetric consequences for those uncorrected errors, The doses actually delivered to the targets and OARs were determined Standard DVH was built in TPS. Figure 7 1. Flow chart of the conclusions for this dissertation
122 LIST OF REFERENCES 1. A. Lightstone, S. Benedict, F. Bova, T. Solberg and R. Stern, "Intracranial stereotactic positioning systems: Report of the American Association of Physicists in Medicine Radiation Therapy Committee Task Group no. 68.," Me d Phys 32, 23802398 (2005). 2. N. Linthout, D. Verellen, K. Tournel and G. Storme, "Six dimensional analysis with daily stereoscopic x ray imaging of intrafraction patient motion in head and neck treatments using five points fixation masks.," Med Phys 33, 504513 (2006). 3. J. Jin, S. Ryu, K. Faber, T. Mikkelsen, Q. Chen, S. Li and B. Movsas, "2D/3D image fusion for accurate target localization and evaluation of a mask based stereotactic system in fractionated stereotactic radiotherapy of cranial lesions.," Med Phys 33, 45574566 (2006). 4. F. Bova, J. Buatti, W. Friedman, W. Mendenhall, C. Yang and C. Liu, "The University of Florida frameless high -precision stereotactic radiotherapy system.," Int J Radiat Oncol Biol Phys 38, 875882 (1997). 5. C. Bert, K. Metheany, K. Doppk e and G. Chen, "A phantom evaluation of a stereo -vision surface imaging system for radiotherapy patient setup.," Med Phys 32, 2753 2762 (2005). 6. H. Li, X. Zhu, L. Zhang, L. Dong, S. Tung, A. Ahamad, K. Chao, W. Morrison, D. Rosenthal, D. Schwartz, R. Mo han and A. Garden, "Comparison of 2D radiographic images and 3D cone beam computed tomography for positioning headandneck radiotherapy patients.," Int J Radiat Oncol Biol Phys 71, 916925 (2008). 7. J. Chang, K. Yenice, A. Narayana and P. Gutin, "Accura cy and feasibility of cone -beam computed tomography for stereotactic radiosurgery setup.," Med Phys 34, 20772084 (2007). 8. L. Masi, F. Casamassima, C. Polli, C. Menichelli, I. Bonucci and C. Cavedon, "Cone Beam CT Image Guidance for Intracranial Stereotactic Treatments: Comparison With a Frame Guided Set -Up.," Int J Radiat Oncol Biol Phys 71, 926933 (2008). 9. S. Meeks, F. B ova, T. Wagner, J. Buatti, W. Friedman and K. Foote, "Image localization for frameless stereotactic radiotherapy.," Int J Radiat Oncol Biol Phys 46, 1291-1299 (2000). 10. C. Herbert, M. Ebert, D. Barclay, D. Whittall, D. Joseph, C. Harper and N. Spry, "Ef fect of bite tray impression technique on relocation accuracy in frameless stereotactic radiotherapy.," Med Dosim 28, 27 30 (2003). 11. J. Boda Heggemann, C. Walter, A. Rahn, H. Wertz, I. Loeb, F. Lohr and F. Wenz, "Repositioning accuracy of two different mask systems 3D revisited: comparison using true 3D/3D matching with cone -beam CT.," Int J Radiat Oncol Biol Phys 66, 15681575 (2006).
123 12. M. Guckenberger, K. Baier, I. Guenther, A. Richter, J. Wilbert, O. Sauer, D. Vordermark and M. Flentje, "Reliabili ty of the bony anatomy in image -guided stereotactic radiotherapy of brain metastases.," Int J Radiat Oncol Biol Phys 69, 294301 (2007). 13. L. Zhang, A. Garden, J. Lo, K. Ang, A. Ahamad, W. Morrison, D. Rosenthal, M. Chambers, X. Zhu, R. Mohan and L. Don g, "Multiple regions -of interest analysis of setup uncertainties for head and -neck cancer radiotherapy.," Int J Radiat Oncol Biol Phys 64, 15591569 (2006). 14. G. Kutcher, L. Coia, M. Gillin, W. Hanson, S. Leibel, R. Morton, J. Palta, J. Purdy, L. Reinst ein and G. Svensson, "Comprehensive QA for radiation oncology: report of AAPM Radiation Therapy Committee Task Group 40.," Med Phys 21, 581618 (1994). 15. E. Klein, J. Hanley, J. Bayouth, F. Yin, W. Simon, S. Dresser, C. Serago, F. Aguirre, L. Ma, B. Arj omandy, C. Liu, C. Sandin and T. Holmes, "Task Group 142 report: quality assurance of medical accelerators.," Med Phys 36, 41974212 (2009). 16. F. F. Yin, J. Wong, J. Balter, S. Benedict, J. P. Bissonnette, T. Craig, L. Dong, D. Jaffray, S. Jiang, S. Kim C. Ma, M. Murphy, P. Munro, T. Solberg and J. Wu, "The role of inroom kV X ray Imaging for patient setup and target localizationReport of Task Group 104 of the Therapy Imaging Committee (American Association of Physicists in Medicine, 2009). 17. J. Bissonnette, D. Moseley and D. Jaffray, "A quality assurance program for image quality of cone -beam CT guidance in radiation therapy.," Med Phys 35, 18071815 (2008). 18. S. Yoo, G. Kim, R. Hammoud, E. Elder, T. Pawlicki, H. Guan, T. Fox, G. Luxton, F. Yin and P. Munro, "A quality assurance program for the on -board imagers.," Med Phys 33, 44314447 (2006). 19. L. Dawson and D. Jaffray, "Advances in image -guided radiation therapy.," J Clin Oncol 25, 938946 (2007). 20. M. Guckenberger, J. Meyer, D. Vor dermark, K. Baier, J. Wilbert and M. Flentje, "Magnitude and clinical relevance of translational and rotational patient setup errors: a cone -beam CT study.," Int J Radiat Oncol Biol Phys 65, 934942 (2006). 21. L. Schubert, D. Westerly, W. Tom, M. Mehta, E. Soisson, T. Mackie, M. Ritter, D. Khuntia, P. Harari and B. Paliwal, "A comprehensive assessment by tumor site of patient setup using daily MVCT imaging from more than 3,800 helical tomotherapy treatments.," Int J Radiat Oncol Biol Phys 73, 12601269 ( 2009). 22. A. Muacevic, M. Staehler, C. Drexler, B. Wowra, M. Reiser and J. Tonn, "Technical description, phantom accuracy, and clinical feasibility for fiducial -free frameless real time image -guided spinal radiosurgery.," J Neurosurg Spine 5 303 312 (2006). 23. J. Ma, Z. Chang, Z. Wang, Q. Jackie Wu, J. Kirkpatrick and F. Yin, "ExacTrac X ray 6 degree -of -freedom image guidance for intracranial noninvasive stereotactic radiotherapy: Comparison with kilo voltage cone -beam CT.," Radiother Oncol (2009).
124 24. N. Yue, J. Knisely, H. Song and R. Nath, "A method to implement full six degree target shift corrections for rigid body in image guided radiotherapy.," Med Phys 33, 21 -31 (2006). 25. S. Boswell, R. Jeraj, K. Ruchala, G. Olivera, H. Jaradat, J. James, A Gutierrez, D. Pearson, G. Frank and T. Mackie, "A novel method to correct for pitch and yaw patient setup errors in helical tomotherapy.," Med Phys 32, 16301639 (2005). 26. M. Murphy, "Image guided patient positioning: if one cannot correct for rotatio nal offsets in external -beam radiotherapy setup, how should rotational offsets be managed?," Med Phys 34, 18801883 (2007). 27. G. Kim, T. Pawlicki, Q. Le and G. Luxton, "Linac -based on-board imaging feasibility and the dosimetric consequences of head rol l in headand -neck IMRT plans.," Med Dosim 33, 93 99 (2008). 28. W. Fu, Y. Yang, X. Li, D. Heron, M. Huq and N. Yue, "Dosimetric effects of patient rotational setup errors on prostate IMRT treatments.," Phys Med Biol 51, 53215331 (2006). 29. S. Kim, H. Jin, H. Yang and R. Amdur, "A study on target positioning error and its impact on dose variation in image -guided stereotactic body radiotherapy for the spine.," Int J Radiat Oncol Biol Phys 73, 15741579 (2009). 30. H. Wang, A. Shiu, C. Wang, J O'Daniel, A. Mahajan, S. Woo, P. Liengsawangwong, R. Mohan and E. Chang, "Dosimetric effect of translational and rotational errors for patients undergoing image guided stereotactic body radiotherapy for spinal metastases.," Int J Radiat Oncol Biol Phys 7 1 12611271 (2008). 31. O. Gutfeld, A. Kretzler, R. Kashani, D. Tatro and J. Balter, "Influence of rotations on dose distributions in spinal stereotactic body radiotherapy (SBRT)." Int J Radiat Oncol Biol Phys 73, 15961601 (2009). 32. S. Sejpal, R. Amos, J. Bluett, L. Levy, R. Kudchadker, J. Johnson, S. Choi and A. Lee, "Dosimetric changes resulting from patient rotational setup errors in proton therapy prostate plans.," Int J Radiat Oncol Biol Phys 75, 40 48 (2009). 33. N. Mail, D. Mo seley, J. Siewerdsen and D. Jaffray, "The influence of bowtie filtration on cone -beam CT image quality.," Med Phys 36, 22 32 (2009). 34. S. Kumar, K. Burke, C. Nalder, P. Jarrett, C. Mubata, R. A'hern, M. Humphreys, M. Bidmead and M. Brada, "Treatment acc uracy of fractionated stereotactic radiotherapy.," Radiother Oncol 74, 53 59 (2005). 35. J. Buatti, F. Bova, W. Friedman, S. Meeks, R. J. Marcus, J. Mickle, T. Ellis and W. Mendenhall, "Preliminary experience with frameless stereotactic radiotherapy.," In t J Radiat Oncol Biol Phys 42, 591599 (1998).
125 36. J. van Santvoort, R. Wiggenraad and P. Bos, "Positioning accuracy in stereotactic radiotherapy using a mask system with added vacuum mouth piece and stereoscopic xray positioning.," Int J Radiat Oncol Bi ol Phys 72, 261267 (2008). 37. M. Engelsman, S. Rosenthal, S. Michaud, J. Adams, R. Schneider, S. Bradley, J. Flanz and H. Kooy, "Intra and interfractional patient motion for a variety of immobilization devices.," Med Phys 32, 34683474 (2005). 38. T. Wagner, S. Meeks, F. Bova, W. Friedman, T. Willoughby, P. Kupelian and W. Tome, "Optical tracking technology in stereotactic radiation therapy.," Med Dosim 32, 111120 (2007). 39. T. Solberg, P. Medin, J. Mullins and S. Li, "Quality assurance of immobiliz ation and target localization systems for frameless stereotactic cranial and extracranial hypofractionated radiotherapy.," Int J Radiat Oncol Biol Phys 71, S131135 (2008). 40. W. Song, S. Kamath, S. Ozawa, S. Ani, A. Chvetsov, N. Bhandare, J. Palta, C. L iu and J. Li, "A dose comparison study between XVI and OBI CBCT systems.," Med Phys 35, 480486 (2008). 41. J. Bissonnette, D. Moseley, E. White, M. Sharpe, T. Purdie and D. Jaffray, "Quality assurance for the geometric accuracy of cone beam CT guidance in radiation therapy.," Int J Radiat Oncol Biol Phys 71, S57 61 (2008). 42. B. Baumert, P. Egli, S. Studer, C Dehing and J. Davis, "Repositioning accuracy of fractionated stereotactic irradiation: assessment of isocentre alignment for different dental fixations by using sequential CT scanning.," Radiother Oncol 74, 61 66 (2005). 43. S. Meeks, W. Tom, T. Willou ghby, P. Kupelian, T. Wagner, J. Buatti and F. Bova, "Optically guided patient positioning techniques.," Semin Radiat Oncol 15, 192201 (2005). 44. T. Ryken, S. Meeks, E. Pennington, P. Hitchon, V. Traynelis, N. Mayr, F. Bova, W. Friedman and J. Buatti, Initial clinical experience with frameless stereotactic radiosurgery: analysis of accuracy and feasibility.," Int J Radiat Oncol Biol Phys 51, 11521158 (2001). 45. P. Schffel, W. Harms, G. Sroka -Perez, W. Schlegel and C. Karger, "Accuracy of a commercia l optical 3D surface imaging system for realignment of patients for radiotherapy of the thorax.," Phys Med Biol 52, 39493963 (2007). 46. D. Jaffray, D. Drake, M. Moreau, A. Martinez and J. Wong, "A radiographic and tomographic imaging system integrated i nto a medical linear accelerator for localization of bone and soft tissue targets.," Int J Radiat Oncol Biol Phys 45, 773789 (1999). 47. Z. Han, J. Zhu, R. Drzymala, D. Rangaraj, S. Mutic, D. Low and H. Li, "The use of a commercial multi -detector device for daily QA of linear acclerators," Med.Phys 35, 2781 (2008).
126 48. C. Liu, L. C. Peng, D. Kahler, S. Samant, J. Li, J. Palta and R. Amdur, "Quality Assessment of FSRT using CBCT," International Journal of Radiation Oncology*Biology*Physics 75, S588 (2009). 49. R. Sweeney, W. Arnold, E. Steixner, M. Nevinny Stickel and P. Lukas, "Compensating for tumor motion by a 6 degree of freedom treatment couch: is patient tolerance an issue?," Int J Radiat Oncol Biol Phys 74 168 171 (2009). 50. P. Ahn, A. Ah n, C. Lee, J. Shen, E. Miller, A. Lukaj, E. Milan, R. Yaparpalvi, S. Kalnicki and M. Garg, "Random positional variation among the skull, mandible, and cervical spine with treatment progression during head and neck radiotherapy.," Int J Radiat Oncol Biol Phys 73 626 633 (2009). 51 R. Wurm, S. Erbel, I. Schwenkert, F. Gum, D. Agaoglu, R. Schild, L. Schlenger, D. Scheffler, M. Brock and V. Budach, "Novalis frameless image-guided noninvasive radiosurgery: initial experience.," Neurosurgery 62, A11 17; discus sion A17 18 (2008). 52 N. Smith, I. Meir, G. Hale, R. Howe, L. Johnson, P. Edwards, D. Dawkes, M. Bidmead and D. Landau, "Real time 3D surface imaging for patient positioning in radiotherapy," Int J Radiat Oncol Biol Phys 57, S187 (2003). 53 D. Gierga M. Riboldi, J. Turcotte, G. Sharp, S. Jiang, A. Taghian and G. Chen, "Comparison of target registration errors for multiple image guided techniques in accelerated partial breast irradiation.," Int J Radiat Oncol Biol Phys 70, 12391246 (2008). 54 M. Krengli, S. Gaiano, E. Mones, A. Ballar, D. Beld, C. Bolchini and G. Loi, "Reproducibility of patient setup by surface image registration system in conformal radiotherapy of prostate cancer.," Radiat Oncol 4 9 (2009). 55 L. C. Peng, D. Kahler, J Li and C. Liu, "Commissioning a new frameless intracranial stereotactic positioning system: The real time 3D stereo -vision surface imaging system," Med Phys 36, 2507 (2009). 56 N. Ramakrishna, F. Rosca, S. Friesen, E. Tezcanli, P. Zygmanszki and F. Hac ker, "A clinical comparison of patient setup and intra -fraction motion using frame -based radiosurgery versus a frameless image-guided radiosurgery system for intracranial lesions.," Radiother Oncol (2010).
127 BIOGRAPHICAL SKETCH Lee Cheng Peng grew up in the big city, Taipei, Taiwan where she attended medical school and worked her first job in the cancer center until 2004. She studied radiation sciences and worked as a clinical medical physicist in the D epartment of Radiation O ncology in Ta iwan for two and half years. Her working experiences focused on radiotherapy of cancer treatment technology. Since Sept 2004, she studied abroad at the University of Michigan and received her M.S. in spring 2005. After graduation, She worked as clinical me dical physicist in the department of radiation oncology of Monmouth Medical Center in New Jersey, for 2.5 years and pa ssed the certification exam of American Board of Radiology (ABR) in June 2007 to become the board -certified medical physicist. In June 2007, she had decided to pursue a Ph.D in the medical physics program under the department of nuclear engineering in the University of Florida. Her reasons to continue Ph.D studying are to help cancer patients and to improve current treatment techniques by ap plying her research after working for many cancer patients in clinical environments. Therefore, her research interests are focused on developing the new radiation treatment technique and then applying to the current radiotherapy procedures. Her goals are e arning more research experiences in her studying life so she would like to contribute her professional knowledge to all cancer patients and family in the future.