Probabilistic Multi-Rater Segmentation via Style-Aware Boundary Conditioning
Multi-rater medical image segmentation reflects the inher ent ambiguity of clinical interpretation, where experts often agree on the semantic extent of a target but differ in fine-grained boundary precision and textural cues. Existing approaches frequently collapse this variabil ity into consensus labels or treat inter-rater differences as noise, leading to overconfident predictions and limited control over personalized out puts. In this work, we view rater variability as structured and preference driven, modeling it through a style factor that captures individual ex pert delineation of ambiguous regions while preserving a shared content representation for anatomical consistency. We introduce a unified prob abilistic framework that disentangles shared content from rater-specific style and employs lightweight rater conditioning to generate personalized yet anatomically consistent segmentations without decoder duplication. To better align learning with actual expert disagreement, we propose a novel disagreement-aware boundary loss that focuses supervision within a controllable band around regions of annotation divergence. This loss heightens sensitivity to contour-level mismatches where inter-rater ambi guity is greatest, while remaining stable in regions of agreement, thereby tying uncertainty estimation and personalization to clinically meaningful boundary disagreement. Experiments on LIDC-IDRI, Kvasir-SEG, and NPC-170 show better aggregated and per-rater performance than strong multi-rater baselines, while generating diverse, plausible segmentations. github.com/sanazkarimi/style-aware
