Abstract Book

S1116

ESTRO 37

EP-2042 Use of on-treatment EPID images to detect inter-fractional anatomical variations N. Matsushita 1 , M. Nakamura 2,3 , M. Sasaki 1,2 , S. Yano 1 , M. Yoshimura 3 , T. Mizowaki 3 1 Kyoto University Hospital, Division of Clinical Radiology Service, Kyoto, Japan 2 Kyoto University, Department of Information Technology and Medical Engineering- Faculty of Human Health Science- Graduate School of Medicine, Kyoto, Japan 3 Kyoto University, Department of Radiation Oncology and Image-applied Therapy- Graduate School of Medicine, Kyoto, Japan Purpose or Objective Inter-fractional anatomical variations such as weight loss, tumor shrinkage and organ motion occur through the course of treatment, which may lead to different dose distributions from those of treatment plan. The aim of our study was to detect the inter-fractional anatomical variations by analyzing on-treatment EPID images in VMAT. Material and Methods Eleven patients with head and neck cancer (HNC) and seven patients with prostate cancer (PC) who underwent VMAT with TrueBeam STx (Varian Medical Systems, Palo Alto, CA) were included in this study. Prescribed dose was 60-70 Gy and 70-78 Gy (2 Gy/fraction) for the HNC cases and the PC cases, respectively. The HNC patients were immobilized by thermoplastic shell, and patient setup positions were corrected based on bone anatomy by a kV-X ray image-guidance system. In the PC cases, six and one patients were treated in the supine and prone position, respectively. Their urine was collected for 60 minutes. Patient positions were corrected based on prostate by CBCT images. No patients were re-planned during the course of treatment. In each arc and fraction, on-treatment integrated EPID images were obtained and analyzed with PerFRACTION (Sun Nuclear Corporation, Melbourne, FL). Differences between fluence maps in the first fraction and those in subsequent fractions were calculated for each arc, and the passing rate based on the tolerance level of 3% of maximum dose with a 10% threshold was calculated. A total of 34 arcs including 21 arcs in the HNC cases and 13 arcs in the PC cases was analyzed. Correlation between passing rate and weight loss was investigated for the HNC cases.

Electronic Poster: Physics track: Inter-fraction motion management (excl. adaptive radiotherapy)

EP-2041 Registration accuracy of high-speed single breath-hold kV-CBCT lung cancer imaging A. Arns 1 , J. Fleckenstein 1 , F. Schneider 1 , J. Boda- Heggemann 1 , Y. Abo-Madyan 1 , V. Steil 1 , F. Wenz 1 , H. Wertz 1 1 Universitätsmedizin Mannheim- Medical Faculty Mannheim- Heidelberg University, Department of Radiation Oncology, Mannheim, Germany Purpose or Objective With single breath-hold imaging, hypo-fractionated deep- inspiration breath-hold SABR of lung tumors can be accelerated and precision can be improved because motion artifacts can be reduced to a minimum. To enable single breath-hold kV-CBCT imaging within 10-15s, linac gantry speed was accelerated to 18°/s for patient positioning. To evaluate clinical applicability, registration accuracy was determined and compared to conventional, clinical kV-CBCT with slow gantry speed of 3°/s. Material and Methods A lung tumor SABR case was simulated with four different tumor-mimicking inlays in a thorax phantom and scanned at 10 different pre-defined positions with conventional, clinical CBCT (3°/s) and faster CBCT (18°/s). To assure precise positions, right-left (RL), cranio-caudal (CC) and anterior-posterior (AP) shifts were applied with optical tracking. For both conventional and faster CBCT, the imaging preset setup was 200° rotation, 120kV and 0.4mAs/frame. Registration to planning CT was applied (1) manually by two clinical experts, (2) automatically with the clinical software provided by the vendor, and (3) automatically by an in-house developed independent framework programmed with MATLAB. The offsets between the registration results and the known isocenter shifts were determined and compared. Results With optical tracking, the systematic error of the 10 random pre-selected isocenter shifts of up to 19mm was reduced to 0.05mm. The stochastic mean displacement error for all tumor-mimicking inlays, shifts and translational directions (RL, CC, AP) with (1) manual registration was 0.0±0.2mm for both faster CBCT (18°/s) and conventional CBCT (3°/s); maximum offset was ±0.6mm. With the (2) clinical software, the mean registration accuracy was 0.0±0.3mm for both imaging speeds, with maximum detection errors of ±0.6mm (faster CBCT) and -1.2/0.7mm (conventional CBCT). Objective evaluation was achieved with the (3) self- developed fully-automatic registration method by applying an identical region of interest around the tumor- shapes for all scanned volumes. Here, the mean displacement error was 0.0±0.2mm for both faster CBCT (18°/s) and conventional CBCT (3°/s); with maximum offsets -0.4/0.5mm. Conclusion The detailed comparison study of registration accuracy for pre-treatment patient positioning between high-speed kV-CBCT (18°/s) and conventional, clinical kV-CBCT (3°/s) with different registration methods showed no degradation with higher gantry speeds for lesions in high contrast areas such as the lung. These promising results provide proof of principle for high-speed single breath- hold kV-CBCT lung tumor imaging within 10-15s. Faster gantry speed facilitates avoidance of defective fluctuations between multiple breath-hold phases and increase of patient compliance as well as treatment precision. Furthermore, faster gantry rotation speed paves the way for future intra-fractional combined imaging and treatment within one breath-hold phase.

Results Passing rates tended to decrease with treatment times in the HNC cases. Whereas mean passing rate during first- twenty fractions was 92.7%, that after 21 fractions decreased to 78.3%. The lowest passing rate was 46.0%. Correlation (R=0.6) between passing rate and weight loss was observed in the HNC cases. In the PC cases, passing rate did not decrease through the course of treatment. Mean passing rate was 99.8% and the lowest was 95.2%.