Beyond scalar losses: calibrating segmentation models via gradient vector field surgery

Laurin Lux, Alexander H. Berger, Moritz Knolle, Daniel Rückert, Johannes C. Paetzold
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:3397-3423, 2026.

Abstract

Region-based loss functions, such as the Dice loss, have established themselves as the de facto standard for highly class- and region-imbalanced segmentation tasks. However, models trained using region-based loss functions are notoriously miscalibrated and typically yield over-confident predictions. In medical imaging applications, such as defining tumor resection margins, this miscalibration is hindering clinical adoption. In this work, we outline a novel gradient perspective on this overconfidence and show how it affects region-based loss functions. We propose a "surgery" on the gradient vector field as a simple, yet effective intervention to mitigate calibration issues. This surgery adds a factor to the loss’s partial derivative, scaling the gradient’s magnitude linearly with the prediction error. In empirical evaluations across 2D and 3D medical segmentation tasks, we demonstrate the effectiveness of this intervention while maintaining high prediction accuracy when used in conjunction with any region-based loss function.

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-lux26a, title = {Beyond scalar losses: calibrating segmentation models via gradient vector field surgery}, author = {Lux, Laurin and Berger, Alexander H. and Knolle, Moritz and R{\"u}ckert, Daniel and Paetzold, Johannes C.}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {3397--3423}, year = {2026}, editor = {Huo, Yuankai and Gao, Mingchen and Kuo, Chang-Fu and Jin, Yueming and Deng, Ruining}, volume = {315}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v315/main/assets/lux26a/lux26a.pdf}, url = {https://proceedings.mlr.press/v315/lux26a.html}, abstract = {Region-based loss functions, such as the Dice loss, have established themselves as the de facto standard for highly class- and region-imbalanced segmentation tasks. However, models trained using region-based loss functions are notoriously miscalibrated and typically yield over-confident predictions. In medical imaging applications, such as defining tumor resection margins, this miscalibration is hindering clinical adoption. In this work, we outline a novel gradient perspective on this overconfidence and show how it affects region-based loss functions. We propose a "surgery" on the gradient vector field as a simple, yet effective intervention to mitigate calibration issues. This surgery adds a factor to the loss’s partial derivative, scaling the gradient’s magnitude linearly with the prediction error. In empirical evaluations across 2D and 3D medical segmentation tasks, we demonstrate the effectiveness of this intervention while maintaining high prediction accuracy when used in conjunction with any region-based loss function.} }
Endnote
%0 Conference Paper %T Beyond scalar losses: calibrating segmentation models via gradient vector field surgery %A Laurin Lux %A Alexander H. Berger %A Moritz Knolle %A Daniel Rückert %A Johannes C. Paetzold %B Proceedings of The 9th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2026 %E Yuankai Huo %E Mingchen Gao %E Chang-Fu Kuo %E Yueming Jin %E Ruining Deng %F pmlr-v315-lux26a %I PMLR %P 3397--3423 %U https://proceedings.mlr.press/v315/lux26a.html %V 315 %X Region-based loss functions, such as the Dice loss, have established themselves as the de facto standard for highly class- and region-imbalanced segmentation tasks. However, models trained using region-based loss functions are notoriously miscalibrated and typically yield over-confident predictions. In medical imaging applications, such as defining tumor resection margins, this miscalibration is hindering clinical adoption. In this work, we outline a novel gradient perspective on this overconfidence and show how it affects region-based loss functions. We propose a "surgery" on the gradient vector field as a simple, yet effective intervention to mitigate calibration issues. This surgery adds a factor to the loss’s partial derivative, scaling the gradient’s magnitude linearly with the prediction error. In empirical evaluations across 2D and 3D medical segmentation tasks, we demonstrate the effectiveness of this intervention while maintaining high prediction accuracy when used in conjunction with any region-based loss function.
APA
Lux, L., Berger, A.H., Knolle, M., Rückert, D. & Paetzold, J.C.. (2026). Beyond scalar losses: calibrating segmentation models via gradient vector field surgery. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:3397-3423 Available from https://proceedings.mlr.press/v315/lux26a.html.

Related Material