ESTRO 2025 - Abstract Book

S3165

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2025

3193

Digital Poster Performance evaluation and comparison of algorithms for deformable dose accumulation Yun Ming Wong 1 , James Cheow Lei Lee 2,1 , Wen Siang Lew 1 , Hong Qi Tan 2,1,3 1 Division of Physics and Applied Physics, Nanyang Technological University, Singapore, Singapore. 2 Division of Radiation Oncology, National Cancer Centre Singapore, Singapore, Singapore. 3 Oncology Academic Clinical Programme, Duke-NUS Medical School, Singapore, Singapore Purpose/Objective: Deformable image registration (DIR) is an important tool for performing dose accumulation during adaptive radiotherapy. This application, known as deformable dose accumulation (DDA), requires a high voxel mapping accuracy, which may not be representable by contour-based metrics commonly used for contour mapping accuracy evaluation. On that account, our current work aimed to 1) investigate and compare the voxel mapping accuracy of four DIR software by quantifying the target registration error (TRE), and 2) study the organ-wise correlation between Dice similarity coefficient (DSC) and TRE. Material/Methods: In our previous work 1 , a deformable anthropomorphic pelvic phantom was fabricated via 3D printing, and four deformation scenarios were simulated by varying the bladder and rectum volumes. CT scans were taken for each scenario, after which the CT numbers were overridden based on real patient images. Four DIR software were tested, i.e., RayStation v10B, Velocity v4.1, Slicer, and Plastimatch. A number of DIR runs were done for each software. For RayStation, different organs were selected in turn as the controlling or focus region of interest; For Velocity, five available algorithms were used: Rigid Registration + Deformable Multi Pass, Deformable, Extended Deformable Multi Pass, Structure Guided Deformable, and Extended Deformable Multi Pass + Structure Guided Deformable; For Slicer, the Elastix extension was studied and the registration parameters were varied in each run; Lastly for Plastimatch, the Demons method was employed and the registration parameters were also varied. The TRE was quantified by calculating the difference between the true and mapped marker positions. Subsequently, Pearson correlation analyses were done to examine the correlation between DSC and mean TRE, separately for bladder, prostate, rectum and all organs combined. A P -value of 0.05 marked the significance of the two-tailed tests.