Contour-informed inter-patient deformable registration for more reliable voxel-based analysis of Head-and-Neck cancer patients.

Background and purpose: The registration of individual dose distributions to a reference anatomy represents a key step in voxel-based analysis (VBA), a tool for spatially informed dose-response assessment. Accurate deformable image registration (DIR) is essential for addressing anatomical variability across patients. To improve both global and region-specific alignment, we enhanced our in-house DIR algorithm (CPT-DIR) by incorporating contour-informed regularization. Materials and methods: We evaluated contour-informed CPT-DIR using CT images from 37 Head-and-Neck patients, with seven cases providing ground-truth dose distribution for dose warping validation. Organs at risk (OARs) were delineated manually, with bone contours auto-generated using TotalSegmentator. Contour-informed constraints (Dice Similarity) were integrated to enhance registration in clinically relevant regions. The global registration results were evaluated using MAE, SSIM and PSNR. Geometric accuracy and warped dose accuracy were assessed using Dice Similarity Coefficient (DSC) and Dose-Organ Overlap (DOO). The performance of CPT-DIR, with and without constraints, was benchmarked against conventional B-spline. Results: CPT-DIR achieved superior accuracy with a MAE of 98.9 ± 6.3 HU, lower than 179.1 ± 17.8 HU for B-spline. Incorporating brainstem contours as regularization improved the DSC from 0.604 ± 0.116 to 0.878 ± 0.017 and DOO from 0.430 ± 0.117 to 0.753 ± 0.043 for brainstem. For the remaining OARs, the enhanced CPT-DIR consistently achieved higher DSC and DOO metrics. Conclusions: The integration of contour-informed regularization in CPT-DIR improved DIR accuracy, particularly in anatomically and dosimetrically relevant regions. This enhanced spatial alignment demonstrated strong potential for advancing reliable inter-patient dosimetric studies in HN radiotherapy.