- Email: [email protected]
Automated Localization Method for a Lung Tumor on EPID Images without Implanted Markers in Stereotactic Body Radiotherapy
Proceedings of the 51st Annual ASTRO Meeting
205
Automated Localization Method for a Lung Tumor on EPID Images without Implanted Markers in Stereotactic Body Radiotherapy
H. Arimura1, Y. Shioyama1, K. Nakamura1, S. Anai2, S. Nomoto3, H. Hirata1, H. Honda1, F. Toyofuku1, Y. Onizuka1, H. Terashima4 1 4
Kyushu University, Fukuoka, Japan, 2Kyushu University Hospital, Fukuoka, Japan, 3Fukuoka University, Fukuoka, Japan, Harasanshin Hospital, Fukuoka, Japan
Purpose/Objective(s): It is very important to monitor the location of a tumor in cine images by using an electronic portal imaging device (EPID) during stereotactic body radiotherapy (SBRT), because it has been unclear whether the internal margin is appropriate for displacements of a tumor due to respiratory motion. Our motivation for this study is a clinical need of a tool for investigating whether the internal margin is appropriate for a moving tumor due to respiration, and tracking the moving tumor in real time without implanted markers, because a majority of conventional methods have used markers implanted adjacent to a tumor, which are invasive approaches for patients. Our goal of the study was to develop an automated method for localization of a lung tumor in cine images on an EPID without implanted markers during the SBRT. Materials/Methods: The localization for a lung tumor was based on a template matching technique between a ‘‘tumor template’’ image obtained in first EPID cine image and the subsequent image. The ‘‘tumor template’’ image was segmented in the first EPID cine image, i.e., reference portal image, based on a multiple-gray level threshold technique and a region growing technique. The tumor location was determined as the position where the tumor template image took the maximum cross-correlation value within each subsequent cine image. EPID images (matrix size: 512 384, pixel size: 0.56 mm) were acquired in cine mode at a sampling rate of 0.5 frame/second by using energies of 4, 6, or 10 MV on linear accelerators. We applied the proposed method to EPID cine images (226 frames) of 12 clinical cases (ages: 51–83, mean: 72) with a non–small-cell lung cancer. Results: The average location deviation between tumor center points obtained by our method and the manual method was 1.47 ± 0.60 mm. The location deviations for 79% of all frames (214 frames) in 12 cases were smaller than 2.0 mm, and those for 94% of all frames were smaller than 3.0 mm. Conclusions: The results of this study suggest that our method based on the tumor template matching technique might be feasible for localization of a lung tumor without implanted markers in SBRT. Author Disclosure: H. Arimura, None; Y. Shioyama, None; K. Nakamura, None; S. Anai, None; S. Nomoto, None; H. Hirata, None; H. Honda, None; F. Toyofuku, None; Y. Onizuka, None; H. Terashima, None.
206
Use of Deformable Registration Tool for Verifying Tumor Position on Kilo-voltage Cone-beam CT Scans for Stage III Lung Cancer
J. R. van Sornsen de Koste, B. J. Slotman, S. Senan VU University Medical Center, Amsterdam, The Netherlands Purpose/Objective(s): Individualized target volumes for Stage III lung cancer can be generated using 4D-CT scans. However, verification of treatment delivery during respiration-gated and image-guided radiotherapy is suboptimal as: (1) conventional cone beam CT (KV-CBCT) images are time-averaged; (2) mediastinal nodes are poorly visualized on blurred images; and (3) coregistration on CBCT tumor volumes is unreliable. We evaluated the use of deformable image registration to propagate planning target volumes on CBCT images. Materials/Methods: The 4D-CT scans were repeated after a dose of 15 Gy in patients with Stage III lung cancer who participated in a prospective study. Five such patients who had free-breathing CBCT scans for setup verification were identified. Gross tumor volumes (GTV) of tumor and nodes were contoured in an end-inspiration phase of 4D-CT using prespecified window level settings. A free-form (B-spline) deformable registration tool (VelocityAI, Velocity Medical Solutions) was applied to (1) deform targets from end-inspiration phase to other phase sets of initial 4D-CT, and (2) to deform end-inspiration targets of initial 4D-CT to all phases of the repeat 4D-CT. Internal targets volumes (ITV) of individually contoured targets were derived that encompassed 9 deformed volumes plus corresponding contour in the end-inspiration phase. The ITV contours were superimposed on average intensity projection (AV-IP) CT scan created of all images of initial 4D-CT scan. The AV-IP and CBCT were rigidly and deformably coregistered, and corresponding ITVs transferred to CBCT. All ITVs were visually reviewed on CBCT, and 3D position relative to planned ITV scored using deformation maps. Conformality between rigid and deformed ITVs was assessed by Dice’s coefficient (DC). Results: Deformable registration revealed decreases in GTVs from inhale to exhale positions (p\0.0001), and also after 15 Gy (28% in 2 patients). Deformed ITV contours of mediastinal nodes were located similarly on AV-IP and CBCT scan relative to surrounding structures. Analysis of serial CBCT in 2 patients showed no change in GTV ITV positions. A tumor fixed to the chest wall showed progressive 3D displacement of ITV-GTV to a mean 5.4 mm relative to initial position. In 2 other patients, each with 5 CBCT studies, a mean shift of 6.7 mm (maximum 10.5 mm) and 5.6 mm, respectively, was observed. In all patients, corresponding DC values decreased over time, and correlated highly with mean displacement. Deformation map results showed varying changes in ITV shape. Conclusions: This deformable registration tool facilitated rapid review of target volumes on repeat CBCT, and detected changes in ITV shape and position during a treatment. This approach can significantly improve treatment verification during image-guided radiotherapy. Author Disclosure: J.R. van Sornsen de Koste, None; B.J. Slotman, Varian Medical Systems, B. Research Grant; S. Senan, Varian Medical Systems, B. Research Grant; Velocity Medical Solutions, C. Other Research Support.
207
Image Quality of On-board Cone-beam CT Acquired during VMAT Delivery
T. Marchant, C. Rowbottom, C. Moore The Christie NHS Foundation Trust, Manchester, United Kingdom Purpose/Objective(s): Emerging rotational treatment techniques such as Volumetric Modulated Arc Therapy (VMAT) give the possibility to acquire a simultaneous kV cone-beam CT (CBCT) image, showing the patient anatomy during treatment delivery. If
S97
205
Automated Localization Method for a Lung Tumor on EPID Images without Implanted Markers in Stereotactic Body Radiotherapy
H. Arimura1, Y. Shioyama1, K. Nakamura1, S. Anai2, S. Nomoto3, H. Hirata1, H. Honda1, F. Toyofuku1, Y. Onizuka1, H. Terashima4 1 4
Kyushu University, Fukuoka, Japan, 2Kyushu University Hospital, Fukuoka, Japan, 3Fukuoka University, Fukuoka, Japan, Harasanshin Hospital, Fukuoka, Japan
Purpose/Objective(s): It is very important to monitor the location of a tumor in cine images by using an electronic portal imaging device (EPID) during stereotactic body radiotherapy (SBRT), because it has been unclear whether the internal margin is appropriate for displacements of a tumor due to respiratory motion. Our motivation for this study is a clinical need of a tool for investigating whether the internal margin is appropriate for a moving tumor due to respiration, and tracking the moving tumor in real time without implanted markers, because a majority of conventional methods have used markers implanted adjacent to a tumor, which are invasive approaches for patients. Our goal of the study was to develop an automated method for localization of a lung tumor in cine images on an EPID without implanted markers during the SBRT. Materials/Methods: The localization for a lung tumor was based on a template matching technique between a ‘‘tumor template’’ image obtained in first EPID cine image and the subsequent image. The ‘‘tumor template’’ image was segmented in the first EPID cine image, i.e., reference portal image, based on a multiple-gray level threshold technique and a region growing technique. The tumor location was determined as the position where the tumor template image took the maximum cross-correlation value within each subsequent cine image. EPID images (matrix size: 512 384, pixel size: 0.56 mm) were acquired in cine mode at a sampling rate of 0.5 frame/second by using energies of 4, 6, or 10 MV on linear accelerators. We applied the proposed method to EPID cine images (226 frames) of 12 clinical cases (ages: 51–83, mean: 72) with a non–small-cell lung cancer. Results: The average location deviation between tumor center points obtained by our method and the manual method was 1.47 ± 0.60 mm. The location deviations for 79% of all frames (214 frames) in 12 cases were smaller than 2.0 mm, and those for 94% of all frames were smaller than 3.0 mm. Conclusions: The results of this study suggest that our method based on the tumor template matching technique might be feasible for localization of a lung tumor without implanted markers in SBRT. Author Disclosure: H. Arimura, None; Y. Shioyama, None; K. Nakamura, None; S. Anai, None; S. Nomoto, None; H. Hirata, None; H. Honda, None; F. Toyofuku, None; Y. Onizuka, None; H. Terashima, None.
206
Use of Deformable Registration Tool for Verifying Tumor Position on Kilo-voltage Cone-beam CT Scans for Stage III Lung Cancer
J. R. van Sornsen de Koste, B. J. Slotman, S. Senan VU University Medical Center, Amsterdam, The Netherlands Purpose/Objective(s): Individualized target volumes for Stage III lung cancer can be generated using 4D-CT scans. However, verification of treatment delivery during respiration-gated and image-guided radiotherapy is suboptimal as: (1) conventional cone beam CT (KV-CBCT) images are time-averaged; (2) mediastinal nodes are poorly visualized on blurred images; and (3) coregistration on CBCT tumor volumes is unreliable. We evaluated the use of deformable image registration to propagate planning target volumes on CBCT images. Materials/Methods: The 4D-CT scans were repeated after a dose of 15 Gy in patients with Stage III lung cancer who participated in a prospective study. Five such patients who had free-breathing CBCT scans for setup verification were identified. Gross tumor volumes (GTV) of tumor and nodes were contoured in an end-inspiration phase of 4D-CT using prespecified window level settings. A free-form (B-spline) deformable registration tool (VelocityAI, Velocity Medical Solutions) was applied to (1) deform targets from end-inspiration phase to other phase sets of initial 4D-CT, and (2) to deform end-inspiration targets of initial 4D-CT to all phases of the repeat 4D-CT. Internal targets volumes (ITV) of individually contoured targets were derived that encompassed 9 deformed volumes plus corresponding contour in the end-inspiration phase. The ITV contours were superimposed on average intensity projection (AV-IP) CT scan created of all images of initial 4D-CT scan. The AV-IP and CBCT were rigidly and deformably coregistered, and corresponding ITVs transferred to CBCT. All ITVs were visually reviewed on CBCT, and 3D position relative to planned ITV scored using deformation maps. Conformality between rigid and deformed ITVs was assessed by Dice’s coefficient (DC). Results: Deformable registration revealed decreases in GTVs from inhale to exhale positions (p\0.0001), and also after 15 Gy (28% in 2 patients). Deformed ITV contours of mediastinal nodes were located similarly on AV-IP and CBCT scan relative to surrounding structures. Analysis of serial CBCT in 2 patients showed no change in GTV ITV positions. A tumor fixed to the chest wall showed progressive 3D displacement of ITV-GTV to a mean 5.4 mm relative to initial position. In 2 other patients, each with 5 CBCT studies, a mean shift of 6.7 mm (maximum 10.5 mm) and 5.6 mm, respectively, was observed. In all patients, corresponding DC values decreased over time, and correlated highly with mean displacement. Deformation map results showed varying changes in ITV shape. Conclusions: This deformable registration tool facilitated rapid review of target volumes on repeat CBCT, and detected changes in ITV shape and position during a treatment. This approach can significantly improve treatment verification during image-guided radiotherapy. Author Disclosure: J.R. van Sornsen de Koste, None; B.J. Slotman, Varian Medical Systems, B. Research Grant; S. Senan, Varian Medical Systems, B. Research Grant; Velocity Medical Solutions, C. Other Research Support.
207
Image Quality of On-board Cone-beam CT Acquired during VMAT Delivery
T. Marchant, C. Rowbottom, C. Moore The Christie NHS Foundation Trust, Manchester, United Kingdom Purpose/Objective(s): Emerging rotational treatment techniques such as Volumetric Modulated Arc Therapy (VMAT) give the possibility to acquire a simultaneous kV cone-beam CT (CBCT) image, showing the patient anatomy during treatment delivery. If
S97