- Email: [email protected]
The Effect of Collimator Rotation on IMRT Treatment Planning
S524
I. J. Radiation Oncology
2490
● Biology ● Physics
Volume 63, Number 2, Supplement, 2005
Automatic Contouring via Deformable Image Registration
M. Foskey, B. Davis, L. Goyal, S. Chang, J. Rosenman, S. Joshi Radiation Oncology, University of North Carolina, Chapel Hill, NC Purpose/Objective: The availability of in-treatment-room CT imaging has made it possible to assess organ location and shape at the time of delivery and develop techniques for adaptive radiation therapy. We have been developing deformable image registration algorithms to automatically track anatomical changes in such images. We evaluate our method on a set of 40 intra-treatment CT images from 3 patients undergoing ART. Materials/Methods: Our approach for automatically contouring intra-treatment images requires that organs be contoured manually on the initial planning image. As each treatment image is acquired, the planning image is deformed so that it matches the treatment image taken at the time of a daily fraction, so that organs as well as features within organs are aligned. The contours from the planning image are deformed in the same way, aligning them with the organs in the treatment image. In pelvic images, however, the presence of bowel gas can cause correspondence errors as no correspondence exists in the gaseous regions. We accommodate bowel gas in the image registration algorithm by simulating a gas “deflation” process. The algorithm models the image as a highly viscous fluid, subject to forces that tend to bring features into alignment. The gas deflation works similarly, with the forces being defined by the gradient of the gas tissue boundary. The flow is simulated using the methods of fluid mechanics. The resulting deformation is applied to the manually generated contours. Each CT scan was collected on a Siemens Primatom CT-on-rails scanner with resolution 0.098⫻0.098⫻0.3 cm. We analyze the accuracy of our method by comparing automatically generated contours to hand-drawn ones. Because of inter-rater variability, however, there is no unique gold standard of comparison. We therefore perform a three-way comparison between our method C and two manual raters A and B. The computer-generated contours were based on planning contours drawn by rater A. We assess similarity by relative volume overlap and radial distance maps between the contours. Results: We construct maps, shown below, of the average radial separation between contours created by the different raters. Statistically analyzing the distance maps, we find no significant difference between the CA and AB comparisons [p⫽0.987]. For volume overlap, we found that the automatic segmentation has significantly better overlap with rater A than rater A has with rater B [p⫽0.014]. Conclusions: For the patients we examined, contours generated by our method agree more closely with contours drawn by the same rater than contours drawn by a different human rater. Thus, our automatic method for contouring organs on images taken during treatment appears to be within the range of human variation in contouring. The accuracy of organ deformation also provides a measure of the accuracy of the overall image deformation procedure.
2491
The Effect of Collimator Rotation on IMRT Treatment Planning
1
S. Luan, P.H. Heintz,1 S.A. Sorensen,2 A.A. Jimenez,2 K.D. Roedersheimer,3 D.Z. Chen,3 G. Wong1 University of New Mexico, Albuquerque, NM, 2Radiation Oncology Associates, Albuquerque, NM, 3University of Notre Dame, Notre Dame, IN 1
Purpose/Objective: Intensity-modulated radiation therapy (IMRT) shows significant potential for improving the cancer survival rate by offering a powerful tool for tumor dose escalation and normal tissue toxicity reduction. Currently, IMRT is mainly implemented on a clinical linear accelerator (LINAC) equipped with a multileaf collimator (MLC). The goal of IMRT treatment planning is to determine a control sequence of the LINAC and the MLC to create a conformal radiation dose distribution inside the patient by optimizing (1) gantry rotation angles, (2) couch rotation angles, (3) MLC leaf positions, and (4) MLC rotation angles. Among these four degrees of freedom, MLC rotation is the only factor that so far has not been systematically exploited. This abstract reports the results of our recent study on the effect of collimator rotation on the qualities of IMRT treatment plans. Materials/Methods: We have redone the IMRT planning for 4 previously treated cases, one head and neck, one prostate, one lung, and one breast. The archived CT scans, tumor and sensitive structure outlines, dose prescriptions, and gantry angle information were used to generate the new IMRT plans. The only change we made was to manually rotate the MLC for all gantry angles synchronously by an increment of 20 degrees, from 0 to 180. With the 4 cases, the total number of new IMRT plans we redid is 36, i.e., 9 plans (180/20⫽9) per case. After the planning, we compared the new IMRT plans based on: (1) the dose volume coverage of the 50% and 90% isodose lines of CTV, PTV, and all critical structures, and (2) the total machine beam-on time in monitor units (MU).
Proceedings of the 47th Annual ASTRO Meeting
Results: In all the 4 treatment sites, we observed large fluctuations in the machine beam-on time. Specifically, in the head and neck case, the beam-on time ranges from 873 MU (at collimator angle 20) to 1708 MU (at collimator angle 160), with an average of 1421 MU. In the prostate case, the beam-on time ranges from 908 MU (at collimator angle 40) to 1383 MU (at collimator angle 100), with an average of 1075 MU. In the lung case, the beam-on time ranges from 793 MU (at collimator angle 20) to 1076 MU (at collimator angle 80), with an average of 957. In the breast case, the beam-on time ranges from 286 MU (at collimator angle 160) to 553 MU (at collimator angle 80), with an average of 352 MU. We also observed a big fluctuation in dose on the 50% and 90% isodose lines for the head and neck case that has the most complex tumor and sensitive structure anatomy. For example, the dose to the cord ranges from 250 cGy (at collimator angle 120) to 1791 cGy (at collimator angle 180) for 90% isodose lines, and from 2333 cGy (at collimator angle 20) to 3125 cGy (at collimator angle 100) for 50% isodose lines; the dose to the right parotid ranges from 421 cGy (at collimator angle 180) to 3910 cGy (at collimator angle 140) for 90% isodose lines, and from 2157 cGy (at collimator angle 180) to 5000 cGy (at collimator angle 100) for 50% isodose lines. Conclusions: Our study indicated that (1) using collimator rotations in IMRT planning may significantly reduce the machine beam-on time, and (2) for cases with a complex tumor anatomy, using collimator rotations may considerably reduce the critical structure toxicity. We believe that these observed effects of collimator rotations on IMRT planning are because of the geometry of the tumor and critical structures, and warrant the development of an automated collimator angle selection tool to assist IMRT planning.
2492
Breast Image-Guided Radiation Therapy Using an Optical Laser Scanner
J.W. Sohn, S. Kim, J.I. Monroe, B.W. Wessels Radiation Oncology, Case Western Reserve University, Cleveland, OH Purpose/Objective: We have investigated the magnitude of the intra- and inter-fractional setup error during breast radiation treatments for implementing IMRT to breast and associate nodes using by electronic portal images. Furthermore, the application of the 3D optical surface scanning device as a daily breast patient positioning system was studied. Daily breast surface image was acquired prior to treatment, and compared to the reference image. Image registration was then performed using three fiducial points. From the image transformation, the positional error was obtained. This error can be corrected for daily treatment. This correction may reduce the margin specially for modified breast conservation therapy. Materials/Methods: The liquid-filled ionization detector (PortalVision LC250) was set to run in a fast frame-averaging mode with an image acquisition rate of 1.4 frames per second. The image resolution from the detector was 256 ⫻ 256 pixels. Electronic portal images were taken for 12 patients at least for 10 treatment days. During each treatment day, an average of 8 to 9 images per field were acquired in the dose rate of 400MU/minute. We developed an image analysis tool box to analyze 2931 images quantitatively. We calculated the standard deviation () from intra- and inter-fractional setup uncertainty and breathing motion to estimate the appropriate planning target volume (PTV). The PTV margin added to the clinical target volume (CTV) in 95% of treatments was calculated as 2 ⫻ (1.96 ⫻ ). For the daily patient positioning, FastSCAN™ Cobra™ was used, which projected laser light on the breast while the CCD camera viewed the laser to record cross-sectional profiles. A reference image was taken when the CT was taken, then images were taken in following days. Images were registered with respect to the tattoo marks on the mediastinum. Rigid body transformation method was used for image registration. A computer program was developed to subtract daily images from the reference image, which was capable of displaying the threedimensional geometric differences in color map. This difference in image reflects the topological change and the error occurred from scanning through uncontrolled breathing cycle. Results: The PTV margins required, which encompassed CTV in 95% of treatments, for intra-fractional error due to breathing were ranged from 2mm to 4mm. However, the PTV margins to compensate the inter-fractional error mainly from daily setup error, partially breathing, ranged from 7mm to 31mm. This suggested that daily patient positioning for breast IMRT was very important. With the setup correction using the three-dimensional surface imaging tool, the geometric difference (topological change and error from breathing) was within 6mm induced by the mobile breast tissue characteristics. The three-dimensional surface image can be used for the adaptive radiation therapy to correct the topological changes. Conclusions: The inter-fractional setup error was much significant than patient breathing motion. Daily setup error may vary depending on how patients are setup with an institution-specific immobilization device. However, prior to IMRT delivery to breast, the magnitude of setup error must be measured and it should be properly formulated into the planning target volume (PTV). To reduce the large PTV, the daily setup device is required for breast IMRT.
2493
Distance Between Thoracic Tumor Position and Diaphragm Position During the Course of Radiotherapy: Does It Remain Constant?
J. Wang,1,2 J. Liang,1 G. Hugo,1 L. Kestin,1 D. Yan1 Radiation Oncology, William Beaumont Hospital, Royal Oak, MI, 2Radiation Oncology, Zhongshan Hospital, Fudan University, Shanghai, China
1
Purpose/Objective: Lung cancer treatment techniques for respiratory tumor motion compensation have relied on the assumption that the relative distance between tumor position and diaphragm position remains constant during the entire course of radiotherapy. This assumption allows daily setup verification and correction based on the diaphragm rather than the tumor position. However, this assumption is questionable due to the inter-treatment variation of thoracic structures. In this study, online patient fluoroscopic images acquired during the course of lung cancer treatment were used to evaluate the distance variation between tumor and diaphragm. Materials/Methods: Online weekly fluoroscopic images (56) were obtained during the course of treatment of ten NSCLC patients. Fluoroscopy encompassing the entire lung was acquired weekly for patients in the treatment position. Tumor position and diaphragm position were detected both on patient anterior-to-posterior (AP) and right-to-left projections of fluoroscopic
S525
I. J. Radiation Oncology
2490
● Biology ● Physics
Volume 63, Number 2, Supplement, 2005
Automatic Contouring via Deformable Image Registration
M. Foskey, B. Davis, L. Goyal, S. Chang, J. Rosenman, S. Joshi Radiation Oncology, University of North Carolina, Chapel Hill, NC Purpose/Objective: The availability of in-treatment-room CT imaging has made it possible to assess organ location and shape at the time of delivery and develop techniques for adaptive radiation therapy. We have been developing deformable image registration algorithms to automatically track anatomical changes in such images. We evaluate our method on a set of 40 intra-treatment CT images from 3 patients undergoing ART. Materials/Methods: Our approach for automatically contouring intra-treatment images requires that organs be contoured manually on the initial planning image. As each treatment image is acquired, the planning image is deformed so that it matches the treatment image taken at the time of a daily fraction, so that organs as well as features within organs are aligned. The contours from the planning image are deformed in the same way, aligning them with the organs in the treatment image. In pelvic images, however, the presence of bowel gas can cause correspondence errors as no correspondence exists in the gaseous regions. We accommodate bowel gas in the image registration algorithm by simulating a gas “deflation” process. The algorithm models the image as a highly viscous fluid, subject to forces that tend to bring features into alignment. The gas deflation works similarly, with the forces being defined by the gradient of the gas tissue boundary. The flow is simulated using the methods of fluid mechanics. The resulting deformation is applied to the manually generated contours. Each CT scan was collected on a Siemens Primatom CT-on-rails scanner with resolution 0.098⫻0.098⫻0.3 cm. We analyze the accuracy of our method by comparing automatically generated contours to hand-drawn ones. Because of inter-rater variability, however, there is no unique gold standard of comparison. We therefore perform a three-way comparison between our method C and two manual raters A and B. The computer-generated contours were based on planning contours drawn by rater A. We assess similarity by relative volume overlap and radial distance maps between the contours. Results: We construct maps, shown below, of the average radial separation between contours created by the different raters. Statistically analyzing the distance maps, we find no significant difference between the CA and AB comparisons [p⫽0.987]. For volume overlap, we found that the automatic segmentation has significantly better overlap with rater A than rater A has with rater B [p⫽0.014]. Conclusions: For the patients we examined, contours generated by our method agree more closely with contours drawn by the same rater than contours drawn by a different human rater. Thus, our automatic method for contouring organs on images taken during treatment appears to be within the range of human variation in contouring. The accuracy of organ deformation also provides a measure of the accuracy of the overall image deformation procedure.
2491
The Effect of Collimator Rotation on IMRT Treatment Planning
1
S. Luan, P.H. Heintz,1 S.A. Sorensen,2 A.A. Jimenez,2 K.D. Roedersheimer,3 D.Z. Chen,3 G. Wong1 University of New Mexico, Albuquerque, NM, 2Radiation Oncology Associates, Albuquerque, NM, 3University of Notre Dame, Notre Dame, IN 1
Purpose/Objective: Intensity-modulated radiation therapy (IMRT) shows significant potential for improving the cancer survival rate by offering a powerful tool for tumor dose escalation and normal tissue toxicity reduction. Currently, IMRT is mainly implemented on a clinical linear accelerator (LINAC) equipped with a multileaf collimator (MLC). The goal of IMRT treatment planning is to determine a control sequence of the LINAC and the MLC to create a conformal radiation dose distribution inside the patient by optimizing (1) gantry rotation angles, (2) couch rotation angles, (3) MLC leaf positions, and (4) MLC rotation angles. Among these four degrees of freedom, MLC rotation is the only factor that so far has not been systematically exploited. This abstract reports the results of our recent study on the effect of collimator rotation on the qualities of IMRT treatment plans. Materials/Methods: We have redone the IMRT planning for 4 previously treated cases, one head and neck, one prostate, one lung, and one breast. The archived CT scans, tumor and sensitive structure outlines, dose prescriptions, and gantry angle information were used to generate the new IMRT plans. The only change we made was to manually rotate the MLC for all gantry angles synchronously by an increment of 20 degrees, from 0 to 180. With the 4 cases, the total number of new IMRT plans we redid is 36, i.e., 9 plans (180/20⫽9) per case. After the planning, we compared the new IMRT plans based on: (1) the dose volume coverage of the 50% and 90% isodose lines of CTV, PTV, and all critical structures, and (2) the total machine beam-on time in monitor units (MU).
Proceedings of the 47th Annual ASTRO Meeting
Results: In all the 4 treatment sites, we observed large fluctuations in the machine beam-on time. Specifically, in the head and neck case, the beam-on time ranges from 873 MU (at collimator angle 20) to 1708 MU (at collimator angle 160), with an average of 1421 MU. In the prostate case, the beam-on time ranges from 908 MU (at collimator angle 40) to 1383 MU (at collimator angle 100), with an average of 1075 MU. In the lung case, the beam-on time ranges from 793 MU (at collimator angle 20) to 1076 MU (at collimator angle 80), with an average of 957. In the breast case, the beam-on time ranges from 286 MU (at collimator angle 160) to 553 MU (at collimator angle 80), with an average of 352 MU. We also observed a big fluctuation in dose on the 50% and 90% isodose lines for the head and neck case that has the most complex tumor and sensitive structure anatomy. For example, the dose to the cord ranges from 250 cGy (at collimator angle 120) to 1791 cGy (at collimator angle 180) for 90% isodose lines, and from 2333 cGy (at collimator angle 20) to 3125 cGy (at collimator angle 100) for 50% isodose lines; the dose to the right parotid ranges from 421 cGy (at collimator angle 180) to 3910 cGy (at collimator angle 140) for 90% isodose lines, and from 2157 cGy (at collimator angle 180) to 5000 cGy (at collimator angle 100) for 50% isodose lines. Conclusions: Our study indicated that (1) using collimator rotations in IMRT planning may significantly reduce the machine beam-on time, and (2) for cases with a complex tumor anatomy, using collimator rotations may considerably reduce the critical structure toxicity. We believe that these observed effects of collimator rotations on IMRT planning are because of the geometry of the tumor and critical structures, and warrant the development of an automated collimator angle selection tool to assist IMRT planning.
2492
Breast Image-Guided Radiation Therapy Using an Optical Laser Scanner
J.W. Sohn, S. Kim, J.I. Monroe, B.W. Wessels Radiation Oncology, Case Western Reserve University, Cleveland, OH Purpose/Objective: We have investigated the magnitude of the intra- and inter-fractional setup error during breast radiation treatments for implementing IMRT to breast and associate nodes using by electronic portal images. Furthermore, the application of the 3D optical surface scanning device as a daily breast patient positioning system was studied. Daily breast surface image was acquired prior to treatment, and compared to the reference image. Image registration was then performed using three fiducial points. From the image transformation, the positional error was obtained. This error can be corrected for daily treatment. This correction may reduce the margin specially for modified breast conservation therapy. Materials/Methods: The liquid-filled ionization detector (PortalVision LC250) was set to run in a fast frame-averaging mode with an image acquisition rate of 1.4 frames per second. The image resolution from the detector was 256 ⫻ 256 pixels. Electronic portal images were taken for 12 patients at least for 10 treatment days. During each treatment day, an average of 8 to 9 images per field were acquired in the dose rate of 400MU/minute. We developed an image analysis tool box to analyze 2931 images quantitatively. We calculated the standard deviation () from intra- and inter-fractional setup uncertainty and breathing motion to estimate the appropriate planning target volume (PTV). The PTV margin added to the clinical target volume (CTV) in 95% of treatments was calculated as 2 ⫻ (1.96 ⫻ ). For the daily patient positioning, FastSCAN™ Cobra™ was used, which projected laser light on the breast while the CCD camera viewed the laser to record cross-sectional profiles. A reference image was taken when the CT was taken, then images were taken in following days. Images were registered with respect to the tattoo marks on the mediastinum. Rigid body transformation method was used for image registration. A computer program was developed to subtract daily images from the reference image, which was capable of displaying the threedimensional geometric differences in color map. This difference in image reflects the topological change and the error occurred from scanning through uncontrolled breathing cycle. Results: The PTV margins required, which encompassed CTV in 95% of treatments, for intra-fractional error due to breathing were ranged from 2mm to 4mm. However, the PTV margins to compensate the inter-fractional error mainly from daily setup error, partially breathing, ranged from 7mm to 31mm. This suggested that daily patient positioning for breast IMRT was very important. With the setup correction using the three-dimensional surface imaging tool, the geometric difference (topological change and error from breathing) was within 6mm induced by the mobile breast tissue characteristics. The three-dimensional surface image can be used for the adaptive radiation therapy to correct the topological changes. Conclusions: The inter-fractional setup error was much significant than patient breathing motion. Daily setup error may vary depending on how patients are setup with an institution-specific immobilization device. However, prior to IMRT delivery to breast, the magnitude of setup error must be measured and it should be properly formulated into the planning target volume (PTV). To reduce the large PTV, the daily setup device is required for breast IMRT.
2493
Distance Between Thoracic Tumor Position and Diaphragm Position During the Course of Radiotherapy: Does It Remain Constant?
J. Wang,1,2 J. Liang,1 G. Hugo,1 L. Kestin,1 D. Yan1 Radiation Oncology, William Beaumont Hospital, Royal Oak, MI, 2Radiation Oncology, Zhongshan Hospital, Fudan University, Shanghai, China
1
Purpose/Objective: Lung cancer treatment techniques for respiratory tumor motion compensation have relied on the assumption that the relative distance between tumor position and diaphragm position remains constant during the entire course of radiotherapy. This assumption allows daily setup verification and correction based on the diaphragm rather than the tumor position. However, this assumption is questionable due to the inter-treatment variation of thoracic structures. In this study, online patient fluoroscopic images acquired during the course of lung cancer treatment were used to evaluate the distance variation between tumor and diaphragm. Materials/Methods: Online weekly fluoroscopic images (56) were obtained during the course of treatment of ten NSCLC patients. Fluoroscopy encompassing the entire lung was acquired weekly for patients in the treatment position. Tumor position and diaphragm position were detected both on patient anterior-to-posterior (AP) and right-to-left projections of fluoroscopic
S525