- Email: [email protected]
Dose Recalculation and the Dose-Guided Radiation Therapy (DGRT) Process Using Megavoltage Cone-Beam CT
Int. J. Radiation Oncology Biol. Phys., Vol. 74, No. 2, pp. 583–592, 2009 Copyright Ó 2009 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/09/$–see front matter
doi:10.1016/j.ijrobp.2008.12.034
PHYSICS CONTRIBUTION
DOSE RECALCULATION AND THE DOSE-GUIDED RADIATION THERAPY (DGRT) PROCESS USING MEGAVOLTAGE CONE-BEAM CT JOEY CHEUNG, B.A.,* JEAN-FRANC¸OIS AUBRY, M.S.,*y SUE S. YOM, M.D., PH.D.,* ALEXANDER R. GOTTSCHALK, M.D., PH.D.,* JUAN CARLOS CELI, PH.D.,xyy AND JEAN POULIOT, PH.D.* y
* Department of Radiation Oncology, University of California San Francisco, Comprehensive Cancer Center, San Francisco, CA; De´partement de physique, genie physique et d’optique, Universite´ Laval, Que´bec, Que´bec, Canada; x Oncology Care Systems Group, Siemens Medical Solutions USA, Inc., Erlangen, Germany; yy currently at IBA Dosimetry, Schwarzenbruck, Germany Purpose: At the University of California San Francisco, daily or weekly three-dimensional images of patients in treatment position are acquired for image-guided radiation therapy. These images can be used for calculating the actual dose delivered to the patient during treatment. In this article, we present the process of performing dose recalculation on megavoltage cone-beam computed tomography images and discuss possible strategies for dose-guided radiation therapy (DGRT). Materials and Methods: A dedicated workstation has been developed to incorporate the necessary elements of DGRT. Patient image correction (cupping, missing data artifacts), calibration, completion, recontouring, and dose recalculation are all implemented in the workstation. Tools for dose comparison are also included. Examples of image correction and dose analysis using 6 head-and-neck and 2 prostate patient datasets are presented to show possible tracking of interfraction dosimetric endpoint variation over the course of treatment. Results: Analysis of the head-and-neck datasets shows that interfraction treatment doses vary compared with the planning dose for the organs at risk, with the mean parotid dose and spinal cord D1 increasing by as much as 52% and 10%, respectively. Variation of the coverage to the target volumes was small, with an average D5 dose difference of 1%. The prostate patient datasets revealed accurate dose coverage to the targeted prostate and varying interfraction dose distributions to the organs at risk. Conclusions: An effective workflow for the clinical implementation of DGRT has been established. With these techniques in place, future clinical developments in adaptive radiation therapy through daily or weekly dosimetric measurements of treatment day images are possible. Ó 2009 Elsevier Inc.
Dose-guided radiation therapy, Megavoltage cone-beam CT, Dose calculation, Treatment replanning, Imageguided radiation therapy.
known as image-guided radiation therapy (IGRT) and has quickly become a popular solution for patient setup (3, 4). At the University of California San Francisco Comprehensive Cancer Center, IGRT has been implemented using a Siemens MVision megavoltage cone-beam computed tomography (MVCBCT) system (5–8) to correct for patient setup errors. For head-and-neck patients, a thermoplastic patient-specific mask is used to position the patient daily, and weekly cone-beam images are acquired to correct for setup errors and to assess the long-term patient positioning accuracy. For prostate patients, daily low-dose cone-beam images
INTRODUCTION With recent advancements in radiation treatment planning and delivery systems such as intensity-modulated radiation therapy (IMRT), very sharp dose gradients can be achieved to attain optimal dose distributions. These advancements, however, require accurate patient positioning and setup to ensure the proper treatment delivery (1, 2). This problem has been addressed through the development of in-room patient imaging systems that can image the patient on the treatment table, which can then be used to verify and correct for patient misalignments or target organ movement. This technique is Reprint requests to: Joey Cheung, University of California San Francisco Dept. of Radiation Oncology, Helen Diller Family Comprehensive Cancer Center, 1600 Divisadero St., Suite HM006, San Francisco, CA 94115. Tel: (415) 353-7179; E-mail: CheungJ@ radonc.ucsf.edu Partly supported by Siemens Oncology Care Systems. Acknowledgments—The authors would like to thank Chandrasekhar Nunna, Thomas Boettger, and Venkataramana Abbaraju for their
aid in the development of the workstation and Sherry Leeper for her contribution in this work. One of the authors (J.F.A.) acknowledges financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC). Received Sept 17, 2008, and in revised form Nov 26, 2008. Accepted for publication Dec 19, 2008.
583
584
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 1. Overview of the dose-guided radiation therapy (DGRT) workflow and its integration into the image-guided radiation therapy (IGRT) workflow. Dashed lines show possible feedback loops into the clinical process.
are acquired and the patients are realigned based on the position of the prostate determined by implanted gold seeds (9, 10). For some prostate patients, weekly cone-beams of slightly higher dose are also acquired to validate alignment with soft tissue information or for research purposes. Because these images are taken directly before treatment, it has been proposed that they can be used for daily assessment of the actual delivered dose to the patient, providing feedback to the radiation physician and therapists on the accuracy of the treatment delivery. As a result, deviations from the plan that are not taken into account by IGRT, such as tumor shrinkage, weight loss, and other internal anatomical changes, can be assessed. Furthermore, the verification of the actual treatment dose can help the physician track the evolution of the treatment and determine if and when a replan or plan adaptation is needed (11, 12). Recently, our group has worked to develop a workstation that integrates all of the necessary elements of dose-guided radiation therapy (DGRT) in an attempt to streamline and introduce the process into the clinical workflow (13, 14). The workstation is currently built on top of a Syngo-based software platform and is codeveloped by the Siemens Oncology Care Systems innovation group and members of the University of California San Francisco research staff. This article will go through the steps required to perform dose recalculation on MVCBCT treatment images and the possible strategies that can be developed using DGRT. In addition, several examples are chosen to illustrate how this process can be performed on patient images. Case studies for tracking daily delivered patient dose are explored.
MATERIALS AND METHODS Acquisition protocols Six head-and-neck and 2 prostate patient datasets were used to illustrate the possibilities of the dedicated DGRT workstation. These patients received MVCBCT imaging with low-dose image
Fig. 2. The conversion process for converting cone-bean (CB) numbers in CB images to computed tomography (CT) numbers. The calibration curves are obtained from measurements of phantoms with inserts of known densities. The CB numbers are converted to CT numbers by matching the calibrated physical densities between both systems.
DGRT process using MVCBCT d J. CHEUNG et al.
585
Table 1. Tumor location, diagnosis, and treatment for head-and-neck and prostate cases # Age/Gender 1 55/m 2 69/m 3 48/m 4 48/m 5 59/m 6 64/m
Location (Primary Site)
8 57/M
Prescription (cGy) No. Fraction Total Dose (Gy) Replan? Chemo?
Nasopharynx T1N1M0 IIB Undifferentiated carcinoma Oral cavity T1N2bM0 IVA Squamous cell carcinoma Left tonsillar pillar T3N2b Squamous cell carcinoma Unknown primary with metastasis to left TXN2bM0 IV neck Squamous cell carcinoma Base of tongue T2N3M0 Squamous cell carcinoma Melanoma of right ear pT1N0 IA Malignant melanoma
# Age/gender 7 70/M
Stage
Location (primary site)
Stage
Prostate adenocarcinoma cT1c Gleason 4+4 PSA 4.27 Prostate adenocarcinoma cT2a/uT3a Gleason 4+5 PSA 4.3
212
33
69.96
Concurrent
200
33
66
Concurrent
212
33
69.96
Concurrent
212
33
69.96
No
212
33
69.96
600
5
No. Total Dose Prescription (cGy) Fraction (Gy)
30
Yes
Concurrent No
Postradiotherapy CyberKnife Boost
180
25
45
2 fractions
9.5 Gy per fraction
180
25
45
2 fractions
9.5 Gy per fraction
acquisition protocols under institutional review board approval. A total of 31 head-and-neck cone-beams and 8 weekly prostate cone-beams were used. In the following discussion, the necessary procedures to perform dose recalculation are outlined in detail. A proposed clinical workflow diagram of the DGRT process is shown in Fig. 1.
Image processing and reconstruction Before reconstruction, a diffusion filter is applied to each conebeam projection image using a built-in function in MeVisLab (MeVis Medical Solutions) to improve the contrast-to-noise ratio (15, 16). The cone-beam images are then reconstructed offline through a modified Feldkamp-Davis-Kress back-projection reconstruction algorithm and imported to the DGRT workstation (17). Because of the geometry of the MVCBCT system setup, the center of the image is the isocenter (7).
Image correction and calibration methods As with most cone-beam systems, a typical cupping artifact, resulting from scatter and beam hardening, is present in the images obtained from our MVCBCT system (18). In addition, a missing data artifact also appears if the section of the patient being imaged is larger than the field-of-view of the system. For the dose to be accurately calculated using these images, these artifacts need to be corrected and the images need to be calibrated to the correct electron densities. Several methods have been developed to correct for these artifacts in head-and-neck cases (19, 20). The workstation uses the method developed by Aubry et al. (20) in which a reference conventional kVCT image is used to create a set of patient-specific correction factors for the MVCBCT images. The slice-by-slice correction is applied to the cone-beam image based on the CT number values of the blurred and filtered kVCT image. After applying the correction factors, the corrected image is already calibrated to the correct CT numbers because the planning kVCT image was used as a reference. These CT numbers can then be calibrated in the dose calculation
engine to the correct physical and electron densities through measurements of phantoms with inserts of known densities. For the prostate patients, a different method developed by Aubry et al. (21) was used that utilizes a set of correction factors calibrated using custom-built pelvic-shaped water phantoms of various sizes. The correction factors are determined by interpolating between the values obtained from the water phantoms based on measured average radiological thicknesses of the patients. Unlike the head-andneck correction, however, the resulting image is not calibrated to the correct CT numbers because of the nonlinear response of the kVCT and MVCBCT systems to the density of imaged materials and further calibration is needed for image correction. This calibration is done in the process of MVCBCT completion.
MVCBCT completion The maximum field of view of our MVCBCT system is 27 cm in diameter in the axial plane and 27 cm in length along the craniocaudal axis. Because the field of view of our MVCBCT system is smaller than most patients, the images exclude part of the patient’s anatomy. The resulting treatment beam attenuation for dose calculation will be incorrect if this missing information is not accounted for. To correct for this, the cone-beam image is completed using the planning kVCT image as a reference. The cone-beam images are registered to the kVCT image based on bony alignment in the area of interest. A new image is created called a cone-beam+ (cone-beam plus) image that takes the kVCT image and replaces the overlapping region of the two images with the cone-beam image. In creating the cone-beam+, the machine isocenter from the original cone-beam is preserved. For the prostate patients, the calibration curves of both imaging systems are used to convert the CT numbers from the MVCBCT image (CB numbers) to CT numbers of the planning kVCT image. This process is shown in Fig. 2.
Recontouring For this study, two radiation oncologists specialized in head-andneck and prostate radiation therapy recontoured several of the
586
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 3. Images of a head-and-neck cone-beam (a) uncorrected, (b) corrected, and (c) completed using a kVCT image. organs at risk on the cone-beam+ images. For head and neck, the contours used were: brain stem, spinal cord, parotid glands, temporomandibular joints (TMJ), mandible, and larynx. For the prostate patients, they were: prostate, rectum, bladder, and seminal vesicles. However, because these images were initially used only for patient positioning and not for dose recalculation, some of the images were taken with a reduced field of view to protect critical organs from additional exposure. Thus, in some cases, a few of the organs of interest were outside the field of view of the cone-beam and were therefore excluded from this study. In addition to the organs at risk, the target volumes for the headand-neck datasets were also recontoured. At the University of California San Francisco, the technique for replanning patients in the middle of treatment involves maintaining the original gross tumor volume (GTV) in the second plan while removing any volumes that extend beyond the skin. Similar attempts are made for
the clinical tumor volumes (CTV). This ensures that adequate dose is given to the high-risk volumes to eliminate the cancer and that original planned coverage to the targets are not compromised (13). The same process was performed on the cone-beam+ images in this study for the GTV and the CTV59.4 (CTV receiving a prescription dose of 59.4 Gy).
Image registration and dose calculation The DGRT workstation uses a pencil beam dose calculation engine (22–25) based on the KonRad Inverse Planning Software (Siemens Medical) commissioned to a 6MV Siemens ONCOR linear accelerator. The original treatment plan for each patient was created using the Pinnacle3 Radiation Therapy Planning System (Philips Medical Systems), which uses a collapse cone algorithm for dose calculation. To accommodate for this difference in dose calculation
DGRT process using MVCBCT d J. CHEUNG et al.
587
Fig. 4. Comparisons of the dose different maps for Patient V on the first week of treatment and the third week overlaid on the planning kVCT image. The colors are windowed to show PDDiffs of greater than 5% (red), less than –5% (blue), and within 5% (green). Although visual verification of local changes to parotid dose may not be accurate, overall dose changes to the spinal cord, larynx, brain stem, and mandible can be useful. engines, a new plan for the kVCT image is created on the workstation that preserves the beam shapes, angles, and energies of the original plan but renormalizes all the beam weights by the same multiplicative factor in order to give the same dose to the isocenter as the original plan. The dose distribution on the planning kVCT is then recalculated using this renormalized plan so that differences in treatment dose will be due to anatomical and positioning changes rather than differences in dose calculation algorithms. This new plan is copied from the kVCT to the cone-beam+ treatment image based on the actual treatment registration to ensure that the treatment isocenter for dose recalculation purposes is the same as it was on the day of treatment. Note that the renormalization process is only performed once per patient on the planning kVCT image and not on the cone-beam images.
Dose comparison Several methods of comparing the planned dose to the treatment dose have been integrated into the DGRT workstation. A simple side-by-side comparison of the two dose distributions using the same isodose markers allows for a visual overview of the difference in doses between the two images. The percentage dose difference (PDDiff) of the local treatment dose to the reference planning dose can also be calculated and viewed as an overlay on top of the cone-beam+ image with the contoured anatomy. Although voxel-by-voxel dose differences may not accurately represent dose changes on the cellular level, the PDDiff does allow the user to see the local difference in overall dose to different parts of the patient anatomy if the anatomy has had a minimal shift in location. The dose difference is normalized to the reference plan dose so that underdoses and overdoses can be easily visualized and understood. The dose and PDDiff information with user-defined limiting parameters for the contoured anatomy can be extracted as a text file and analyzed. Furthermore, dose–volume histograms (DVHs) can be plotted and viewed on the workstation.
RESULTS Head-and-neck cases The six head-and-neck patient datasets used for this study were all newly diagnosed with cancers in different primary
tumor sites. Table 1 lists the treatment information along with the location and stages of the tumors. Figure 3 shows an example of the application of the correction method and image completion on a head-and-neck patient image. For this particular dataset, the cone-beam was taken on the first week of treatment. The cupping and missing data artifacts are clearly seen on the images on the left (a). After correction (b), the images are uniform and the density values are calibrated to the correct CT numbers needed for dose calculation. The image can then be completed (c) using the original planning kVCT image to complement the missing data. Figure 4 shows the evolution of the dose differences map of one head-and-neck patient (Patient V) throughout treatment. The color overlay is windowed to show PDDiffs greater than 5% (red), less than –5% (blue), and within 5% (green) of the planning dose. A number of factors can result in larger overdosed (red) regions including patient misalignments, tumor or organ regression, more global anatomical changes, and weight loss. Prostate cases Both prostate cases were treated with concurrent hormone therapy. The treatment information can be found in Table 1. Figure 5 shows the application of the correction method and image completion on a pelvic patient. As with head-and-neck cases, the cone-beam correction and calibration (B) and completion (C) on pelvic images result in a final image that is density accurate. Figure 6 shows a side-by-side dose grid comparison of a prostate planning kVCT image with a treatment conebeam+ image for Patient VII. Analysis shows that the volumes of the prostate, rectum, seminal vesicles, and bladder in this image changed by –4%, –4%, –44%, and 36%, respectively. Dosimetric end point tracking Figure 7 shows the time evolution of the percent difference in mean dose compared to planning dose to the left and right
588
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 5. Images of a pelvic cone-beam (a) uncorrected (b) corrected and (c) completed using a kVCT image.
parotids for three of the 6 patients (Patients I, II, and V) and the difference in D1 dose (Dx dose is the minimum dose received by the highest dosed x-percent of the volume) to the spinal cord for all 6 patients. The D1 dose was used instead of the maximum dose because of the high sensitivity and variation of the maximum dose. For the parotids, 3 head-andneck patients were omitted because their cone-beam images were taken with a limited field of view that resulted in the parotids being incompletely imaged or not imaged at all. An average net increase in dose to both parotids is observed for all 3 patients. In addition, all 6 patients also received an average net increase in dose to the spinal cord over the course of treatment. A summary of the percent dose differences to the target and various recontoured organs at risk averaged across all 6 patients is shown in Table 2. For the prostate patients, DVHs were plotted to observe the differences in dose to the various recontoured organs. Figure 8
shows the DVHs of the bladder, prostate, seminal vesicles, and rectum for Patient VII (a) and Patient VIII (b). The dotted lines show the original planned DVHs to the regions of interest. Figure 9 shows the percentage volume changes from the planning kVCT volume for the bladder, prostate, rectum, and seminal vesicles for the 2 prostate patients in our study. DISCUSSION Of the different head-and-neck structures that were contoured, the parotids tended to be the only ones that exhibited a long-term unidirectional change in volume and geometry. As has been shown in other studies (26–28), the irradiated parotids tended to shrink in size over the course of treatment and migrate inward toward the center of the skull. This moves the parotids into higher dose regions over time and a general increase in dose to the right parotid was seen across all
DGRT process using MVCBCT d J. CHEUNG et al.
589
Fig. 6. A side-by-side comparison of the dose maps from the planning kVCT and the treatment cone-beam+ images. Both images use the same dose range colors and the values are the lower bounds of the ranges. The yellow contour is the bladder, purple is the prostate, green is the seminal vesicles, and brown is the rectum.
3 patients after the third week of treatment. Patient I showed some weight loss during treatment and received a higher dose to the right side. Patient II received a right unilateral treatment, which could explain the large increase in dose to the right parotid over the course of the treatment. Patient V exhibited a large amount of weight loss (about 4 cm of soft tissue on both sides of the neck) and tumor regression, which can be seen in the increase in dose to both parotids. Weekly variation of the spinal cord dose seemed to be fairly random but tended to be shifted toward overdoses (Fig. 7). These errors could be attributed to daily setup errors, weight loss, or changes in anatomy between the date of the planning kVCT and the start of the treatment. One notable exception in the study is Patient V, who had lost a large amount of weight between the cone-beam images, resulting in a large (10.7%) increase in D1 spinal cord dose. In the study, the TMJ was too small and too close to the edge of the cone-beam images for 2 of the patients to give accurate results. Of the 4 patients left, there was an average net decrease in dose to both TMJ (Table 2). The D1 dose to the mandibles tended to vary randomly based on the data. This is likely based on patient positioning rather than weight
loss. Slight movement in the patient’s jaw position was apparent between the weekly cone beams and the planning kVCT as the anatomy was recontoured. Brain stem variability appeared to be the smallest of all the anatomy studied. This is likely due to the relative position and size of the brain stem. Because the base of the skull is fairly easy to align for IGRT, the brain stem is likely to be positioned accurately during treatment and receive close to the planned dose. The larynx, on the other hand, showed high variability for all 6 patients, most likely from the non-rigidity of the organ and its close location to the skin surface where high dose variability can occur. Dose coverage to the target volumes remained fairly consistent for all of the cone-beam images. The D95 dose to the GTV changed by an average (standard deviation) of 0.1% (0.8%) with a maximum decrease of 0%, showing a consistent overall coverage to the GTV. High-dose coverage to the GTV also remained fairly consistent, with a mean dose increase of 1.0% (1.7%) and a maximum decrease of –4.0% for the D5 dose. Similar results were found for the CTV59.4 with values for the D95 and D5 with a maximum absolute change of 3.0% and 5.1% and an average change of 0.1% and 0.7%, respectively.
590
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 7. Time evolution of the percentage mean dose difference to the parotids (a) and the D1 percentage dose difference to the spinal cord (b).
For the prostate patients, the variation of dose to the bladder and rectum are likely the result of volume changes. The current procedure used in our institution aligns the patients based on daily prostate images. Therefore, although the prostate may be well aligned for each fraction, the bladder, rectum, and other organs may not be well aligned, leading to varying fractional dose distributions to these regions. It is also important to consider the accuracy of the recontoured anatomy on the lower contrast treatment day images when evaluating the DVH curves. The same oncologist who prepared the treatment plan recontoured the images used in this study. Figure 9 examines the changes in volume for the bladder, prostate, rectum, and seminal vesicles. These Table 2. Statistics for head-and-neck targets and organs at risk percentage dose differences Mean Standard Minimum Maximum Spinal cord D1* 2.2 Left parotid mean dose 5.8 Right parotid mean dose 18.0 Mandible D1* 1.8 1.1 Brain stem D1* –1.6 Left TMJ D1* –0.5 Right TMJ D1* Larynx mean dose 6.7 GTV D95* 0.1 1.0 GTV D5* 0.1 CTV59.4 D95* CTV59.4 D5* 0.7
2.9 9.5 18.0 4.1 2.7 8.0 5.1 9.3 0.8 1.7 1.1 1.3
–3.3 –4.2 –2.1 –6.1 –4.8 –16.0 –12.3 –11.5 0.0 –4.0 –3.0 –1.2
10.7 37.9 51.8 12.2 10.1 10.0 5.2 25.4 2.1 5.8 2.3 5.1
Abbreviations: TMJ = temporomandibular joint; GTV = gross tumor volume; CTV = clinical tumor volume. * D1 and D5 are the minimum doses received by at least 1% and 5% of the volume that received the largest dose, respectively. This was used instead of the maximum dose due to the high sensitivity and variation of the maximum dose. D95 is the maximum dose received by at least 95% of the volume. These values were chosen based on the planning dosimetric considerations for the organs at risk. All values are given as percent increases from the planning dose.
variations in volume will likely affect the dose received by these organs as reflected by the variation in DVHs in Fig. 8. However, because these volume changes occur in three dimensions, the actual dosimetric consequences of these volume changes are much more complicated and must be evaluated on a case-by-case basis. The dose to the prostate and seminal vesicles tended to be fairly accurate when compared with the original planned dose. On one of the cone-beam days, Patient VIII received 3.3% less dose to 95% of the prostate volume (D95). However, on closer observation of the images themselves, the prostate was not aligned properly on that day of treatment, which explains the diminished prostate dose coverage. Because the patients are aligned daily based on prostate location using gold seed implants, the DGRT process provides a good check that the primary tumor volume is getting the prescribed dose throughout treatment. The DGRT process presented can be used for verification of the accuracy of the treatment delivery and for an assessment of the daily trend of dose differences over the course of treatment. Dose accumulation is not currently implemented in the process. Although non-rigid deformation algorithms could be used as an approximation for organ movement, it is still insufficient in providing dose accumulation on the cellular level. The DGRT process as presented does not implement non-rigid deformation in order to reduce calculation times for clinical use, but the process can be easily integrated if desired and is an avenue that we are currently exploring for future clinical use. However, even without the ability to perform dose accumulation, DGRT can provide valuable feedback about the dose delivered during each treatment fraction. Trends in dose differences can reveal important consequences of anatomy change or setup errors. CONCLUSION All of the functionalities required to provide an effective workflow for the clinical implementation of DGRT have
DGRT process using MVCBCT d J. CHEUNG et al.
591
Fig. 8. Dose–volume histograms of the contoured anatomy for the prostate patients: (a) Patient VII, (b) Patient VIII. The dotted line shows the original planning dose–volume histogram.
been integrated in our workstation, enabling the regular comparison of the dose of the day with the plan. In the examples presented, it was found that although dose coverage to the target volumes have remained fairly consistent for head-and-neck cases, doses to various organs at risk, including the spinal cord, parotids, mandible, brain stem, TMJ, and larynx, have been observed to vary between fractions. For the prostate cases, we have shown that accurate prostate positioning using IGRT has allowed for accurate dose coverage to the targeted prostate. However, doses to the organs at risk including the bladder, rectum, and seminal vesicles are observed to vary. With DGRT, this variation can be monitored to ensure that the dose to the organs at risk do not surpass critical limits. Regular monitoring of the daily patient dose based on images acquired of the patient on the treatment table can be
a powerful tool for tracking the progress and accuracy of the treatment. In addition, with IGRT implemented in many facilities, DGRT can be easily integrated into the clinical workflow. Access to information about the doseof-the-day will open the doors to many areas of research and clinical improvements in treatment verification and dose accumulation. A systematic approach can also be developed to determine when a patient requires replanning based on dosimetric considerations for either the organs at risk or the target tumor volume. Furthermore, dose– response relations can be developed based on long-term patient follow-up studies. The studies presented in this article are only for a limited set of patients and serve as an example of the many possible case studies that can be performed using the proposed DGRT process.
Fig. 9. Plots of the percentage volume change from planning of the bladder, prostate, rectum, and seminal vesicles for the 2 pelvic patients.
592
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
REFERENCES 1. O’Daniel JC, Garden AS, Schwartz DL, et al. Parotid gland dose in intensity-modulated radiotherapy for head and neck cancer: is what you plan what you get? Int J Radiat Oncol Biol Phys 2007;69:1290–1296. 2. Webb S. Intensity-modulated radiation therapy. Bristol, UK: Institute of Physics Publishing; 2001. 3. Ma CM, Paskalev K. In-room CT techniques for image-guided radiation therapy. Med Dosim 2006;31:30–39. 4. Verellen D, Ridder MD, Storme G. A (short) history of imageguided radiotherapy. Radiother Oncol 2008;86:4–13. 5. Pouliot J, Bani-Hashemi A, Chen J, et al. Low-dose megavoltage cone-beam CT for radiation therapy. Int J Radiat Oncol Biol Phys 2005;61:552–560. 6. Sillanpaa J, Chang J, Mageras G, et al. Developments in megavoltage cone beam CT with an amorphous silicon EPID: reduction of exposure and synchronization with respiratory gating. Med Phys 2005;32:819–829. 7. Morin O, Gillis A, Chen J, et al. Megavoltage cone-beam CT: System description and clinical applications. Med Dosim 2006;31:51–61. 8. Morin O, Gillis A, Descovich M, et al. Patient dose considerations for routine megavoltage cone-beam CT imaging. Med Phys 2007;34:1819–1827. 9. Pouliot J, Aubin M, Langden K, et al. (Non)-migration of radiopaque markers used for on-line localization of the prostate with an electronic portal imaging device. Int J Radiat Oncol Biol Phys 2003;56:862–866. 10. Balter JM, Sandler HM, Lam K, et al. Measurement of prostate movement over the course of routine radiotherapy using implanted markers. Int J Radiat Oncol Biol Phys 1995;31: 113–118. 11. Hansen E, Bucci MK, Quivey JM, et al. Repeat CT imaging and replanning during the course of IMRT for head-and-neck cancer. Int J Radiat Oncol Biol Phys 2006;64:355–362. 12. Lu W, Olivera GH, Chen Q, et al. Deformable registration of the planning image (kVCT) and the daily images (MVCT) for adaptive radiation therapy. Phys Med Biol 2006;51:4357–4374. 13. Pouliot J. From dose to image to dose: IGRT to DGRT. In: Mould RF, editor. Choices in advanced radiotherapy. Nucletron, BV: The Netherlands; 2007. p. 243–250. 14. Chen J, Morin O, Aubin M, et al. Dose-guided radiation therapy with megavoltage cone-beam CT. Br J Radiol 2006;79: S87–S98. 15. Monteil J, Beghdadi A. New adaptive nonlinear anisotropic diffusion for noise smoothing. Volume 3 of the IEEE International
16. 17. 18. 19. 20. 21.
22. 23. 24. 25. 26.
27.
28.
Conference on Image Processing, page 254. Chicago, IL, USA, 1998. IEEE Comp Soc, Los Alamitos, CA. Perona P, Malik J. Scale-space and edge detection using anisotropic diffusion. IEEE Trans Pattern Anal Machine Intell 1990; 12:629–639. Feldkamp LA, Davis LC, Kress JW. Practical cone-beam algorithm. J Opt Soc Am A 1984;1:612–619. Spies L, Ebert M, Groh BA, et al. Correction of scatter in megavoltage cone-beam CT. Phys Med Biol 2001;46:821–833. Morin O, Chen J, Gillis A, et al. Dose calculations using megavoltage cone-beam CT. Int J Radiat Oncol Biol Phys 2007;67: 1202–1210. Aubry JF, Pouliot J, Beaulieu L. Correction of megavoltage cone-beam CT images for dose calculation in the head and neck region. Med Phys 2008;35:900–908. Aubry JF, Cheung J, Gottschalk A, et al. Correction of megavoltage cone-beam CT images of the pelvic region based on phantom measurements for dose calculation purposes. J Appl Clin Med Phys 2009;10:2852. Mohan R, Chui C, Lidofsky L. Differential pencil beam dose computation model for photons. Med Phys 1986;13:64–73. Bortfeld T, Schlegel W, Rhein B. Decomposition of pencil beam kernels for fast dose calculations in three-dimensional treatment planning. Med Phys 1993;20:311–318. Boudreau C, Heath E, Seuntjens J, et al. IMRT head and neck treatment planning with a commercially available Monte Carlo based planning system. Phys Med Biol 2005;50:879–890. Yang J, Li J, Chen L, et al. Dosimetric verification of IMRT treatment planning using Monte Carlo simulations for prostate cancer. Phys Med Biol 2005;50:869–878. Barker JL, Garden AS, Ang KK, et al. Quantification of volumetric and geometric changes occurring during fractionated radiotherapy for head-and-neck cancer using an integrated CT/ linear accelerator system. Int J Radiat Oncol Biol Phys 2004; 59:960–970. Vasquez Osorio EM, Hoogeman MS, Al-Mamgani A, et al. Local anatomic changes in parotid and submandibular glands during radiotherapy for oropharynx cancer and correlation with dose, studied in detail with nonrigid registration. Int J Radiat Oncol Biol Phys 2008;70:875–882. Lee C, Langen KM, Lu W, et al. Assessment of parotid gland dose changes during head and neck cancer radiotherapy using daily megavoltage computed tomography and deformable image registration. Int J Radiat Oncol Biol Phys 2008;71: 1563–1571.
doi:10.1016/j.ijrobp.2008.12.034
PHYSICS CONTRIBUTION
DOSE RECALCULATION AND THE DOSE-GUIDED RADIATION THERAPY (DGRT) PROCESS USING MEGAVOLTAGE CONE-BEAM CT JOEY CHEUNG, B.A.,* JEAN-FRANC¸OIS AUBRY, M.S.,*y SUE S. YOM, M.D., PH.D.,* ALEXANDER R. GOTTSCHALK, M.D., PH.D.,* JUAN CARLOS CELI, PH.D.,xyy AND JEAN POULIOT, PH.D.* y
* Department of Radiation Oncology, University of California San Francisco, Comprehensive Cancer Center, San Francisco, CA; De´partement de physique, genie physique et d’optique, Universite´ Laval, Que´bec, Que´bec, Canada; x Oncology Care Systems Group, Siemens Medical Solutions USA, Inc., Erlangen, Germany; yy currently at IBA Dosimetry, Schwarzenbruck, Germany Purpose: At the University of California San Francisco, daily or weekly three-dimensional images of patients in treatment position are acquired for image-guided radiation therapy. These images can be used for calculating the actual dose delivered to the patient during treatment. In this article, we present the process of performing dose recalculation on megavoltage cone-beam computed tomography images and discuss possible strategies for dose-guided radiation therapy (DGRT). Materials and Methods: A dedicated workstation has been developed to incorporate the necessary elements of DGRT. Patient image correction (cupping, missing data artifacts), calibration, completion, recontouring, and dose recalculation are all implemented in the workstation. Tools for dose comparison are also included. Examples of image correction and dose analysis using 6 head-and-neck and 2 prostate patient datasets are presented to show possible tracking of interfraction dosimetric endpoint variation over the course of treatment. Results: Analysis of the head-and-neck datasets shows that interfraction treatment doses vary compared with the planning dose for the organs at risk, with the mean parotid dose and spinal cord D1 increasing by as much as 52% and 10%, respectively. Variation of the coverage to the target volumes was small, with an average D5 dose difference of 1%. The prostate patient datasets revealed accurate dose coverage to the targeted prostate and varying interfraction dose distributions to the organs at risk. Conclusions: An effective workflow for the clinical implementation of DGRT has been established. With these techniques in place, future clinical developments in adaptive radiation therapy through daily or weekly dosimetric measurements of treatment day images are possible. Ó 2009 Elsevier Inc.
Dose-guided radiation therapy, Megavoltage cone-beam CT, Dose calculation, Treatment replanning, Imageguided radiation therapy.
known as image-guided radiation therapy (IGRT) and has quickly become a popular solution for patient setup (3, 4). At the University of California San Francisco Comprehensive Cancer Center, IGRT has been implemented using a Siemens MVision megavoltage cone-beam computed tomography (MVCBCT) system (5–8) to correct for patient setup errors. For head-and-neck patients, a thermoplastic patient-specific mask is used to position the patient daily, and weekly cone-beam images are acquired to correct for setup errors and to assess the long-term patient positioning accuracy. For prostate patients, daily low-dose cone-beam images
INTRODUCTION With recent advancements in radiation treatment planning and delivery systems such as intensity-modulated radiation therapy (IMRT), very sharp dose gradients can be achieved to attain optimal dose distributions. These advancements, however, require accurate patient positioning and setup to ensure the proper treatment delivery (1, 2). This problem has been addressed through the development of in-room patient imaging systems that can image the patient on the treatment table, which can then be used to verify and correct for patient misalignments or target organ movement. This technique is Reprint requests to: Joey Cheung, University of California San Francisco Dept. of Radiation Oncology, Helen Diller Family Comprehensive Cancer Center, 1600 Divisadero St., Suite HM006, San Francisco, CA 94115. Tel: (415) 353-7179; E-mail: CheungJ@ radonc.ucsf.edu Partly supported by Siemens Oncology Care Systems. Acknowledgments—The authors would like to thank Chandrasekhar Nunna, Thomas Boettger, and Venkataramana Abbaraju for their
aid in the development of the workstation and Sherry Leeper for her contribution in this work. One of the authors (J.F.A.) acknowledges financial support from the Natural Sciences and Engineering Research Council of Canada (NSERC). Received Sept 17, 2008, and in revised form Nov 26, 2008. Accepted for publication Dec 19, 2008.
583
584
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 1. Overview of the dose-guided radiation therapy (DGRT) workflow and its integration into the image-guided radiation therapy (IGRT) workflow. Dashed lines show possible feedback loops into the clinical process.
are acquired and the patients are realigned based on the position of the prostate determined by implanted gold seeds (9, 10). For some prostate patients, weekly cone-beams of slightly higher dose are also acquired to validate alignment with soft tissue information or for research purposes. Because these images are taken directly before treatment, it has been proposed that they can be used for daily assessment of the actual delivered dose to the patient, providing feedback to the radiation physician and therapists on the accuracy of the treatment delivery. As a result, deviations from the plan that are not taken into account by IGRT, such as tumor shrinkage, weight loss, and other internal anatomical changes, can be assessed. Furthermore, the verification of the actual treatment dose can help the physician track the evolution of the treatment and determine if and when a replan or plan adaptation is needed (11, 12). Recently, our group has worked to develop a workstation that integrates all of the necessary elements of dose-guided radiation therapy (DGRT) in an attempt to streamline and introduce the process into the clinical workflow (13, 14). The workstation is currently built on top of a Syngo-based software platform and is codeveloped by the Siemens Oncology Care Systems innovation group and members of the University of California San Francisco research staff. This article will go through the steps required to perform dose recalculation on MVCBCT treatment images and the possible strategies that can be developed using DGRT. In addition, several examples are chosen to illustrate how this process can be performed on patient images. Case studies for tracking daily delivered patient dose are explored.
MATERIALS AND METHODS Acquisition protocols Six head-and-neck and 2 prostate patient datasets were used to illustrate the possibilities of the dedicated DGRT workstation. These patients received MVCBCT imaging with low-dose image
Fig. 2. The conversion process for converting cone-bean (CB) numbers in CB images to computed tomography (CT) numbers. The calibration curves are obtained from measurements of phantoms with inserts of known densities. The CB numbers are converted to CT numbers by matching the calibrated physical densities between both systems.
DGRT process using MVCBCT d J. CHEUNG et al.
585
Table 1. Tumor location, diagnosis, and treatment for head-and-neck and prostate cases # Age/Gender 1 55/m 2 69/m 3 48/m 4 48/m 5 59/m 6 64/m
Location (Primary Site)
8 57/M
Prescription (cGy) No. Fraction Total Dose (Gy) Replan? Chemo?
Nasopharynx T1N1M0 IIB Undifferentiated carcinoma Oral cavity T1N2bM0 IVA Squamous cell carcinoma Left tonsillar pillar T3N2b Squamous cell carcinoma Unknown primary with metastasis to left TXN2bM0 IV neck Squamous cell carcinoma Base of tongue T2N3M0 Squamous cell carcinoma Melanoma of right ear pT1N0 IA Malignant melanoma
# Age/gender 7 70/M
Stage
Location (primary site)
Stage
Prostate adenocarcinoma cT1c Gleason 4+4 PSA 4.27 Prostate adenocarcinoma cT2a/uT3a Gleason 4+5 PSA 4.3
212
33
69.96
Concurrent
200
33
66
Concurrent
212
33
69.96
Concurrent
212
33
69.96
No
212
33
69.96
600
5
No. Total Dose Prescription (cGy) Fraction (Gy)
30
Yes
Concurrent No
Postradiotherapy CyberKnife Boost
180
25
45
2 fractions
9.5 Gy per fraction
180
25
45
2 fractions
9.5 Gy per fraction
acquisition protocols under institutional review board approval. A total of 31 head-and-neck cone-beams and 8 weekly prostate cone-beams were used. In the following discussion, the necessary procedures to perform dose recalculation are outlined in detail. A proposed clinical workflow diagram of the DGRT process is shown in Fig. 1.
Image processing and reconstruction Before reconstruction, a diffusion filter is applied to each conebeam projection image using a built-in function in MeVisLab (MeVis Medical Solutions) to improve the contrast-to-noise ratio (15, 16). The cone-beam images are then reconstructed offline through a modified Feldkamp-Davis-Kress back-projection reconstruction algorithm and imported to the DGRT workstation (17). Because of the geometry of the MVCBCT system setup, the center of the image is the isocenter (7).
Image correction and calibration methods As with most cone-beam systems, a typical cupping artifact, resulting from scatter and beam hardening, is present in the images obtained from our MVCBCT system (18). In addition, a missing data artifact also appears if the section of the patient being imaged is larger than the field-of-view of the system. For the dose to be accurately calculated using these images, these artifacts need to be corrected and the images need to be calibrated to the correct electron densities. Several methods have been developed to correct for these artifacts in head-and-neck cases (19, 20). The workstation uses the method developed by Aubry et al. (20) in which a reference conventional kVCT image is used to create a set of patient-specific correction factors for the MVCBCT images. The slice-by-slice correction is applied to the cone-beam image based on the CT number values of the blurred and filtered kVCT image. After applying the correction factors, the corrected image is already calibrated to the correct CT numbers because the planning kVCT image was used as a reference. These CT numbers can then be calibrated in the dose calculation
engine to the correct physical and electron densities through measurements of phantoms with inserts of known densities. For the prostate patients, a different method developed by Aubry et al. (21) was used that utilizes a set of correction factors calibrated using custom-built pelvic-shaped water phantoms of various sizes. The correction factors are determined by interpolating between the values obtained from the water phantoms based on measured average radiological thicknesses of the patients. Unlike the head-andneck correction, however, the resulting image is not calibrated to the correct CT numbers because of the nonlinear response of the kVCT and MVCBCT systems to the density of imaged materials and further calibration is needed for image correction. This calibration is done in the process of MVCBCT completion.
MVCBCT completion The maximum field of view of our MVCBCT system is 27 cm in diameter in the axial plane and 27 cm in length along the craniocaudal axis. Because the field of view of our MVCBCT system is smaller than most patients, the images exclude part of the patient’s anatomy. The resulting treatment beam attenuation for dose calculation will be incorrect if this missing information is not accounted for. To correct for this, the cone-beam image is completed using the planning kVCT image as a reference. The cone-beam images are registered to the kVCT image based on bony alignment in the area of interest. A new image is created called a cone-beam+ (cone-beam plus) image that takes the kVCT image and replaces the overlapping region of the two images with the cone-beam image. In creating the cone-beam+, the machine isocenter from the original cone-beam is preserved. For the prostate patients, the calibration curves of both imaging systems are used to convert the CT numbers from the MVCBCT image (CB numbers) to CT numbers of the planning kVCT image. This process is shown in Fig. 2.
Recontouring For this study, two radiation oncologists specialized in head-andneck and prostate radiation therapy recontoured several of the
586
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 3. Images of a head-and-neck cone-beam (a) uncorrected, (b) corrected, and (c) completed using a kVCT image. organs at risk on the cone-beam+ images. For head and neck, the contours used were: brain stem, spinal cord, parotid glands, temporomandibular joints (TMJ), mandible, and larynx. For the prostate patients, they were: prostate, rectum, bladder, and seminal vesicles. However, because these images were initially used only for patient positioning and not for dose recalculation, some of the images were taken with a reduced field of view to protect critical organs from additional exposure. Thus, in some cases, a few of the organs of interest were outside the field of view of the cone-beam and were therefore excluded from this study. In addition to the organs at risk, the target volumes for the headand-neck datasets were also recontoured. At the University of California San Francisco, the technique for replanning patients in the middle of treatment involves maintaining the original gross tumor volume (GTV) in the second plan while removing any volumes that extend beyond the skin. Similar attempts are made for
the clinical tumor volumes (CTV). This ensures that adequate dose is given to the high-risk volumes to eliminate the cancer and that original planned coverage to the targets are not compromised (13). The same process was performed on the cone-beam+ images in this study for the GTV and the CTV59.4 (CTV receiving a prescription dose of 59.4 Gy).
Image registration and dose calculation The DGRT workstation uses a pencil beam dose calculation engine (22–25) based on the KonRad Inverse Planning Software (Siemens Medical) commissioned to a 6MV Siemens ONCOR linear accelerator. The original treatment plan for each patient was created using the Pinnacle3 Radiation Therapy Planning System (Philips Medical Systems), which uses a collapse cone algorithm for dose calculation. To accommodate for this difference in dose calculation
DGRT process using MVCBCT d J. CHEUNG et al.
587
Fig. 4. Comparisons of the dose different maps for Patient V on the first week of treatment and the third week overlaid on the planning kVCT image. The colors are windowed to show PDDiffs of greater than 5% (red), less than –5% (blue), and within 5% (green). Although visual verification of local changes to parotid dose may not be accurate, overall dose changes to the spinal cord, larynx, brain stem, and mandible can be useful. engines, a new plan for the kVCT image is created on the workstation that preserves the beam shapes, angles, and energies of the original plan but renormalizes all the beam weights by the same multiplicative factor in order to give the same dose to the isocenter as the original plan. The dose distribution on the planning kVCT is then recalculated using this renormalized plan so that differences in treatment dose will be due to anatomical and positioning changes rather than differences in dose calculation algorithms. This new plan is copied from the kVCT to the cone-beam+ treatment image based on the actual treatment registration to ensure that the treatment isocenter for dose recalculation purposes is the same as it was on the day of treatment. Note that the renormalization process is only performed once per patient on the planning kVCT image and not on the cone-beam images.
Dose comparison Several methods of comparing the planned dose to the treatment dose have been integrated into the DGRT workstation. A simple side-by-side comparison of the two dose distributions using the same isodose markers allows for a visual overview of the difference in doses between the two images. The percentage dose difference (PDDiff) of the local treatment dose to the reference planning dose can also be calculated and viewed as an overlay on top of the cone-beam+ image with the contoured anatomy. Although voxel-by-voxel dose differences may not accurately represent dose changes on the cellular level, the PDDiff does allow the user to see the local difference in overall dose to different parts of the patient anatomy if the anatomy has had a minimal shift in location. The dose difference is normalized to the reference plan dose so that underdoses and overdoses can be easily visualized and understood. The dose and PDDiff information with user-defined limiting parameters for the contoured anatomy can be extracted as a text file and analyzed. Furthermore, dose–volume histograms (DVHs) can be plotted and viewed on the workstation.
RESULTS Head-and-neck cases The six head-and-neck patient datasets used for this study were all newly diagnosed with cancers in different primary
tumor sites. Table 1 lists the treatment information along with the location and stages of the tumors. Figure 3 shows an example of the application of the correction method and image completion on a head-and-neck patient image. For this particular dataset, the cone-beam was taken on the first week of treatment. The cupping and missing data artifacts are clearly seen on the images on the left (a). After correction (b), the images are uniform and the density values are calibrated to the correct CT numbers needed for dose calculation. The image can then be completed (c) using the original planning kVCT image to complement the missing data. Figure 4 shows the evolution of the dose differences map of one head-and-neck patient (Patient V) throughout treatment. The color overlay is windowed to show PDDiffs greater than 5% (red), less than –5% (blue), and within 5% (green) of the planning dose. A number of factors can result in larger overdosed (red) regions including patient misalignments, tumor or organ regression, more global anatomical changes, and weight loss. Prostate cases Both prostate cases were treated with concurrent hormone therapy. The treatment information can be found in Table 1. Figure 5 shows the application of the correction method and image completion on a pelvic patient. As with head-and-neck cases, the cone-beam correction and calibration (B) and completion (C) on pelvic images result in a final image that is density accurate. Figure 6 shows a side-by-side dose grid comparison of a prostate planning kVCT image with a treatment conebeam+ image for Patient VII. Analysis shows that the volumes of the prostate, rectum, seminal vesicles, and bladder in this image changed by –4%, –4%, –44%, and 36%, respectively. Dosimetric end point tracking Figure 7 shows the time evolution of the percent difference in mean dose compared to planning dose to the left and right
588
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 5. Images of a pelvic cone-beam (a) uncorrected (b) corrected and (c) completed using a kVCT image.
parotids for three of the 6 patients (Patients I, II, and V) and the difference in D1 dose (Dx dose is the minimum dose received by the highest dosed x-percent of the volume) to the spinal cord for all 6 patients. The D1 dose was used instead of the maximum dose because of the high sensitivity and variation of the maximum dose. For the parotids, 3 head-andneck patients were omitted because their cone-beam images were taken with a limited field of view that resulted in the parotids being incompletely imaged or not imaged at all. An average net increase in dose to both parotids is observed for all 3 patients. In addition, all 6 patients also received an average net increase in dose to the spinal cord over the course of treatment. A summary of the percent dose differences to the target and various recontoured organs at risk averaged across all 6 patients is shown in Table 2. For the prostate patients, DVHs were plotted to observe the differences in dose to the various recontoured organs. Figure 8
shows the DVHs of the bladder, prostate, seminal vesicles, and rectum for Patient VII (a) and Patient VIII (b). The dotted lines show the original planned DVHs to the regions of interest. Figure 9 shows the percentage volume changes from the planning kVCT volume for the bladder, prostate, rectum, and seminal vesicles for the 2 prostate patients in our study. DISCUSSION Of the different head-and-neck structures that were contoured, the parotids tended to be the only ones that exhibited a long-term unidirectional change in volume and geometry. As has been shown in other studies (26–28), the irradiated parotids tended to shrink in size over the course of treatment and migrate inward toward the center of the skull. This moves the parotids into higher dose regions over time and a general increase in dose to the right parotid was seen across all
DGRT process using MVCBCT d J. CHEUNG et al.
589
Fig. 6. A side-by-side comparison of the dose maps from the planning kVCT and the treatment cone-beam+ images. Both images use the same dose range colors and the values are the lower bounds of the ranges. The yellow contour is the bladder, purple is the prostate, green is the seminal vesicles, and brown is the rectum.
3 patients after the third week of treatment. Patient I showed some weight loss during treatment and received a higher dose to the right side. Patient II received a right unilateral treatment, which could explain the large increase in dose to the right parotid over the course of the treatment. Patient V exhibited a large amount of weight loss (about 4 cm of soft tissue on both sides of the neck) and tumor regression, which can be seen in the increase in dose to both parotids. Weekly variation of the spinal cord dose seemed to be fairly random but tended to be shifted toward overdoses (Fig. 7). These errors could be attributed to daily setup errors, weight loss, or changes in anatomy between the date of the planning kVCT and the start of the treatment. One notable exception in the study is Patient V, who had lost a large amount of weight between the cone-beam images, resulting in a large (10.7%) increase in D1 spinal cord dose. In the study, the TMJ was too small and too close to the edge of the cone-beam images for 2 of the patients to give accurate results. Of the 4 patients left, there was an average net decrease in dose to both TMJ (Table 2). The D1 dose to the mandibles tended to vary randomly based on the data. This is likely based on patient positioning rather than weight
loss. Slight movement in the patient’s jaw position was apparent between the weekly cone beams and the planning kVCT as the anatomy was recontoured. Brain stem variability appeared to be the smallest of all the anatomy studied. This is likely due to the relative position and size of the brain stem. Because the base of the skull is fairly easy to align for IGRT, the brain stem is likely to be positioned accurately during treatment and receive close to the planned dose. The larynx, on the other hand, showed high variability for all 6 patients, most likely from the non-rigidity of the organ and its close location to the skin surface where high dose variability can occur. Dose coverage to the target volumes remained fairly consistent for all of the cone-beam images. The D95 dose to the GTV changed by an average (standard deviation) of 0.1% (0.8%) with a maximum decrease of 0%, showing a consistent overall coverage to the GTV. High-dose coverage to the GTV also remained fairly consistent, with a mean dose increase of 1.0% (1.7%) and a maximum decrease of –4.0% for the D5 dose. Similar results were found for the CTV59.4 with values for the D95 and D5 with a maximum absolute change of 3.0% and 5.1% and an average change of 0.1% and 0.7%, respectively.
590
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
Fig. 7. Time evolution of the percentage mean dose difference to the parotids (a) and the D1 percentage dose difference to the spinal cord (b).
For the prostate patients, the variation of dose to the bladder and rectum are likely the result of volume changes. The current procedure used in our institution aligns the patients based on daily prostate images. Therefore, although the prostate may be well aligned for each fraction, the bladder, rectum, and other organs may not be well aligned, leading to varying fractional dose distributions to these regions. It is also important to consider the accuracy of the recontoured anatomy on the lower contrast treatment day images when evaluating the DVH curves. The same oncologist who prepared the treatment plan recontoured the images used in this study. Figure 9 examines the changes in volume for the bladder, prostate, rectum, and seminal vesicles. These Table 2. Statistics for head-and-neck targets and organs at risk percentage dose differences Mean Standard Minimum Maximum Spinal cord D1* 2.2 Left parotid mean dose 5.8 Right parotid mean dose 18.0 Mandible D1* 1.8 1.1 Brain stem D1* –1.6 Left TMJ D1* –0.5 Right TMJ D1* Larynx mean dose 6.7 GTV D95* 0.1 1.0 GTV D5* 0.1 CTV59.4 D95* CTV59.4 D5* 0.7
2.9 9.5 18.0 4.1 2.7 8.0 5.1 9.3 0.8 1.7 1.1 1.3
–3.3 –4.2 –2.1 –6.1 –4.8 –16.0 –12.3 –11.5 0.0 –4.0 –3.0 –1.2
10.7 37.9 51.8 12.2 10.1 10.0 5.2 25.4 2.1 5.8 2.3 5.1
Abbreviations: TMJ = temporomandibular joint; GTV = gross tumor volume; CTV = clinical tumor volume. * D1 and D5 are the minimum doses received by at least 1% and 5% of the volume that received the largest dose, respectively. This was used instead of the maximum dose due to the high sensitivity and variation of the maximum dose. D95 is the maximum dose received by at least 95% of the volume. These values were chosen based on the planning dosimetric considerations for the organs at risk. All values are given as percent increases from the planning dose.
variations in volume will likely affect the dose received by these organs as reflected by the variation in DVHs in Fig. 8. However, because these volume changes occur in three dimensions, the actual dosimetric consequences of these volume changes are much more complicated and must be evaluated on a case-by-case basis. The dose to the prostate and seminal vesicles tended to be fairly accurate when compared with the original planned dose. On one of the cone-beam days, Patient VIII received 3.3% less dose to 95% of the prostate volume (D95). However, on closer observation of the images themselves, the prostate was not aligned properly on that day of treatment, which explains the diminished prostate dose coverage. Because the patients are aligned daily based on prostate location using gold seed implants, the DGRT process provides a good check that the primary tumor volume is getting the prescribed dose throughout treatment. The DGRT process presented can be used for verification of the accuracy of the treatment delivery and for an assessment of the daily trend of dose differences over the course of treatment. Dose accumulation is not currently implemented in the process. Although non-rigid deformation algorithms could be used as an approximation for organ movement, it is still insufficient in providing dose accumulation on the cellular level. The DGRT process as presented does not implement non-rigid deformation in order to reduce calculation times for clinical use, but the process can be easily integrated if desired and is an avenue that we are currently exploring for future clinical use. However, even without the ability to perform dose accumulation, DGRT can provide valuable feedback about the dose delivered during each treatment fraction. Trends in dose differences can reveal important consequences of anatomy change or setup errors. CONCLUSION All of the functionalities required to provide an effective workflow for the clinical implementation of DGRT have
DGRT process using MVCBCT d J. CHEUNG et al.
591
Fig. 8. Dose–volume histograms of the contoured anatomy for the prostate patients: (a) Patient VII, (b) Patient VIII. The dotted line shows the original planning dose–volume histogram.
been integrated in our workstation, enabling the regular comparison of the dose of the day with the plan. In the examples presented, it was found that although dose coverage to the target volumes have remained fairly consistent for head-and-neck cases, doses to various organs at risk, including the spinal cord, parotids, mandible, brain stem, TMJ, and larynx, have been observed to vary between fractions. For the prostate cases, we have shown that accurate prostate positioning using IGRT has allowed for accurate dose coverage to the targeted prostate. However, doses to the organs at risk including the bladder, rectum, and seminal vesicles are observed to vary. With DGRT, this variation can be monitored to ensure that the dose to the organs at risk do not surpass critical limits. Regular monitoring of the daily patient dose based on images acquired of the patient on the treatment table can be
a powerful tool for tracking the progress and accuracy of the treatment. In addition, with IGRT implemented in many facilities, DGRT can be easily integrated into the clinical workflow. Access to information about the doseof-the-day will open the doors to many areas of research and clinical improvements in treatment verification and dose accumulation. A systematic approach can also be developed to determine when a patient requires replanning based on dosimetric considerations for either the organs at risk or the target tumor volume. Furthermore, dose– response relations can be developed based on long-term patient follow-up studies. The studies presented in this article are only for a limited set of patients and serve as an example of the many possible case studies that can be performed using the proposed DGRT process.
Fig. 9. Plots of the percentage volume change from planning of the bladder, prostate, rectum, and seminal vesicles for the 2 pelvic patients.
592
I. J. Radiation Oncology d Biology d Physics
Volume 74, Number 2, 2009
REFERENCES 1. O’Daniel JC, Garden AS, Schwartz DL, et al. Parotid gland dose in intensity-modulated radiotherapy for head and neck cancer: is what you plan what you get? Int J Radiat Oncol Biol Phys 2007;69:1290–1296. 2. Webb S. Intensity-modulated radiation therapy. Bristol, UK: Institute of Physics Publishing; 2001. 3. Ma CM, Paskalev K. In-room CT techniques for image-guided radiation therapy. Med Dosim 2006;31:30–39. 4. Verellen D, Ridder MD, Storme G. A (short) history of imageguided radiotherapy. Radiother Oncol 2008;86:4–13. 5. Pouliot J, Bani-Hashemi A, Chen J, et al. Low-dose megavoltage cone-beam CT for radiation therapy. Int J Radiat Oncol Biol Phys 2005;61:552–560. 6. Sillanpaa J, Chang J, Mageras G, et al. Developments in megavoltage cone beam CT with an amorphous silicon EPID: reduction of exposure and synchronization with respiratory gating. Med Phys 2005;32:819–829. 7. Morin O, Gillis A, Chen J, et al. Megavoltage cone-beam CT: System description and clinical applications. Med Dosim 2006;31:51–61. 8. Morin O, Gillis A, Descovich M, et al. Patient dose considerations for routine megavoltage cone-beam CT imaging. Med Phys 2007;34:1819–1827. 9. Pouliot J, Aubin M, Langden K, et al. (Non)-migration of radiopaque markers used for on-line localization of the prostate with an electronic portal imaging device. Int J Radiat Oncol Biol Phys 2003;56:862–866. 10. Balter JM, Sandler HM, Lam K, et al. Measurement of prostate movement over the course of routine radiotherapy using implanted markers. Int J Radiat Oncol Biol Phys 1995;31: 113–118. 11. Hansen E, Bucci MK, Quivey JM, et al. Repeat CT imaging and replanning during the course of IMRT for head-and-neck cancer. Int J Radiat Oncol Biol Phys 2006;64:355–362. 12. Lu W, Olivera GH, Chen Q, et al. Deformable registration of the planning image (kVCT) and the daily images (MVCT) for adaptive radiation therapy. Phys Med Biol 2006;51:4357–4374. 13. Pouliot J. From dose to image to dose: IGRT to DGRT. In: Mould RF, editor. Choices in advanced radiotherapy. Nucletron, BV: The Netherlands; 2007. p. 243–250. 14. Chen J, Morin O, Aubin M, et al. Dose-guided radiation therapy with megavoltage cone-beam CT. Br J Radiol 2006;79: S87–S98. 15. Monteil J, Beghdadi A. New adaptive nonlinear anisotropic diffusion for noise smoothing. Volume 3 of the IEEE International
16. 17. 18. 19. 20. 21.
22. 23. 24. 25. 26.
27.
28.
Conference on Image Processing, page 254. Chicago, IL, USA, 1998. IEEE Comp Soc, Los Alamitos, CA. Perona P, Malik J. Scale-space and edge detection using anisotropic diffusion. IEEE Trans Pattern Anal Machine Intell 1990; 12:629–639. Feldkamp LA, Davis LC, Kress JW. Practical cone-beam algorithm. J Opt Soc Am A 1984;1:612–619. Spies L, Ebert M, Groh BA, et al. Correction of scatter in megavoltage cone-beam CT. Phys Med Biol 2001;46:821–833. Morin O, Chen J, Gillis A, et al. Dose calculations using megavoltage cone-beam CT. Int J Radiat Oncol Biol Phys 2007;67: 1202–1210. Aubry JF, Pouliot J, Beaulieu L. Correction of megavoltage cone-beam CT images for dose calculation in the head and neck region. Med Phys 2008;35:900–908. Aubry JF, Cheung J, Gottschalk A, et al. Correction of megavoltage cone-beam CT images of the pelvic region based on phantom measurements for dose calculation purposes. J Appl Clin Med Phys 2009;10:2852. Mohan R, Chui C, Lidofsky L. Differential pencil beam dose computation model for photons. Med Phys 1986;13:64–73. Bortfeld T, Schlegel W, Rhein B. Decomposition of pencil beam kernels for fast dose calculations in three-dimensional treatment planning. Med Phys 1993;20:311–318. Boudreau C, Heath E, Seuntjens J, et al. IMRT head and neck treatment planning with a commercially available Monte Carlo based planning system. Phys Med Biol 2005;50:879–890. Yang J, Li J, Chen L, et al. Dosimetric verification of IMRT treatment planning using Monte Carlo simulations for prostate cancer. Phys Med Biol 2005;50:869–878. Barker JL, Garden AS, Ang KK, et al. Quantification of volumetric and geometric changes occurring during fractionated radiotherapy for head-and-neck cancer using an integrated CT/ linear accelerator system. Int J Radiat Oncol Biol Phys 2004; 59:960–970. Vasquez Osorio EM, Hoogeman MS, Al-Mamgani A, et al. Local anatomic changes in parotid and submandibular glands during radiotherapy for oropharynx cancer and correlation with dose, studied in detail with nonrigid registration. Int J Radiat Oncol Biol Phys 2008;70:875–882. Lee C, Langen KM, Lu W, et al. Assessment of parotid gland dose changes during head and neck cancer radiotherapy using daily megavoltage computed tomography and deformable image registration. Int J Radiat Oncol Biol Phys 2008;71: 1563–1571.