159. Evaluation of a user-guided deformable registration workflow for multi-modal (CT-MRI) prostate imaging

Abstract

Purpose Reg-refine is a tool that adds human knowledge to deformable image registration (DIR) process. In this study a CT-MRI deformable registration workflow for prostate cancer was evaluated. Methods Deformable and rigid alignments of axial CT and axial T2 weighted MRI scans of 8 patients were evaluated using MIM Maestro 6.7.10 (Mim Software Inc.). DIR workflow consisted of three rigid alignments, in the upper, middle, ad lower prostate part, defined by two users in consensus using Reg-refine, and converted in local alignments by the DIR algorithm. Dice coefficient and Hausdorff distance were computed for evaluation of deformable and rigid alignments. Results Rice resulted (0.78 ± 0.12; max 0.9, min 0.58) for rigid registration vs (0.89 ± 0.05; max 0.94, min 0.82). Hausdorff distance resulted resulted (11.7 mm ± 4.5 mm; max 17.1, min 5.4) for rigid registration vs (7.6 mm ± 4.1 mm; max 13.5 mm, min 3.5 mm). Box plots are reported in Fig. 1. Conclusions The DIR workflow evaluated in this study produced good results in all patients analyzed and was able to improve the registration accuracy of rigid alignments. Reg-refine based DIR of CT and MRI scans can be used for prostate contouring.