Spatially-Aware Evaluation of Segmentation Uncertainty

TL;DR

This paper introduces three spatially-aware metrics for evaluating segmentation uncertainty that incorporate structural and boundary information, improving clinical relevance and pattern discrimination in medical imaging.

Contribution

The paper proposes novel spatially-aware uncertainty evaluation metrics that account for anatomical structure and boundary information, addressing limitations of existing voxel-wise methods.

Findings

Metrics better align with clinical importance

Improved discrimination of meaningful vs. spurious uncertainty

Validated on prostate segmentation challenge data

Abstract

Uncertainty maps highlight unreliable regions in segmentation predictions. However, most uncertainty evaluation metrics treat voxels independently, ignoring spatial context and anatomical structure. As a result, they may assign identical scores to qualitatively distinct patterns (e.g., scattered vs. boundary-aligned uncertainty). We propose three spatially aware metrics that incorporate structural and boundary information and conduct a thorough validation on medical imaging data from the prostate zonal segmentation challenge within the Medical Segmentation Decathlon. Our results demonstrate improved alignment with clinically important factors and better discrimination between meaningful and spurious uncertainty patterns.