Medical Physics Oral Presentations: 0053 Validation of an atlas consistent with RTOG for automated contouring in radiotherapy planning for high-risk prostate cancers Adly Naga 1 ,3 , Yasir Bahadur 2 , Camelia Constantinescu 1 , Maha Eltaher 2 ,3 , Mohammad Attar 2 1 King Faisal Specialist Hospital & Research Center, Jeddah, Saudi Arabia, 2 King Abdulaziz University Hospital, Jeddah, Saudi Arabia, 3 National Cancer Institute, Cairo University, Cairo, Egypt Purpose: To create a CT-based atlas of high-risk prostate-cancers, and evaluate the performance of a commercially available automated-atlas-based-segmentation (AABS) software in delineation of target- volumes and organs-at-risk (OARs). Methods and materials: Treatment planning CT-images of 20 high-risk prostate-cancer patients were retrospectively reviewed. All patients had non-contrast CT-simulation with full bladder and empty rectum. Volumes-of-interest included prostate and pelvic-lymph-nodes (LN) CTVs; OARs (rectum, bladder, left and right femoral heads); and patient's body. Patients were stratified according to AP-diameter as small (S; AP≤26cm) and large (L; AP>26cm). Multi- patient CT-atlases were generated for each group using a deformable-registration ABAS software. The atlas-predicted-contours were compared against manual contours, using volume-percentage-difference, Dice Similarity Coefficient (DSC) and percentage-overlap (PO). Correlation of ABAS algorithm performance with patient's body-size was evaluated by a logistic regression test. Results: We identified 9 (45%) patients in S-group and 11 (55%) in L-group. ABAS tended to underestimate the segmented volumes for all structures. High levels-of-agreement (DSC≥0.9) were found for femoral heads and body, good levels-of-agreement (0.8≤DSC≤0.9) for rectum and bladder, and limited levels-of-agreement (DSC≤0.8) for prostate and pelvic-LN CTVs. Significant correlation was found between DSC and AP diameter for LN-CTV (p=0.0472), bladder (p=0.0192), femoral heads (p=0.0445), and lateral diameter for rectum (p=0.0195), suggesting that patient's body-size has to be considered in atlas generation. Conclusion: If generated according to patient's body-size, CT-based multi-patient atlases for automated-contouring can be reasonably used in treatment planning of high-risk prostate radiotherapy. However, deformable organs (bladder and rectum), and structures with not clearly defined borders (prostate and pelvic-LN CTVs) need review by a radiation oncologist. 0062