Towards reliable WMH segmentation under domain shift: An application study using maximum entropy regularization to improve uncertainty estimation

Franco Matzkin, Agostina J. Larrazabal, Diego H. Milone, Jose Dolz, Enzo Ferrante

Abstract: Highlights•Entropy-based uncertainty estimates can be used as a proxy for segmentation errors.•Maximum-entropy regularization improves model calibration and uncertainty quantification under domain shift.•Models trained with maximum-entropy regularization achieve stronger alignment between uncertainty and segmentation errors.•Validation performed on multicenter WMH datasets highlights robustness to different imaging conditions.

External IDs:dblp:journals/cbm/MatzkinLMDF25