mlsanityData to 14 Oct 2026

Confidence-Calibrating Regularization for Robust Brain MRI Segmentation Under Domain Shift

arXiv 2509.23176 ↗First posted 27 Sep 2025cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

The Segment Anything Model (SAM) exhibits strong zero-shot performance on natural images but suffers from domain shift and overconfidence when applied to medical volumes. We propose CalSAM, a lightweight adaptation framework that (i) reduces encoder sensitivity to domain shift via a Feature Fisher Information Penalty (FIP) computed on 3D feature maps and (ii) penalizes overconfident voxel-wise errors through a Confidence Misalignment Penalty (CMP). The combined loss, (LCalSAM) fine-tunes only the mask decoder while keeping SAM's encoders frozen. On cross-center and scanner-shift evaluations, CalSAM substantially improves accuracy and calibration: e.g., on the BraTS scanner split (Siemens→GE) CalSAM shows a +7.4% relative improvement in DSC (80.1% vs. 74.6%), a -26.9% reduction in HD95 (4.6 mm vs. 6.3 mm), and a -39.5% reduction in ECE (5.2% vs. 8.6%). On ATLAS-C (motion corruptions), CalSAM achieves a +5.3% relative improvement in DSC (75.9%) and a -32.6% reduction in ECE (5.8%). Ablations show FIP and CMP contribute complementary gains (p<0.01), and the Fisher penalty incurs a modest ~15% training-time overhead. CalSAM therefore delivers improved domain generalization and better-calibrated uncertainty estimates for brain MRI segmentation, while retaining the computational benefits of freezing SAM's encoder.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 22 of its 42 references that are in the map. First posted on arXiv 27 Sep 2025.

Builds on

22 of 42 references.

References · 22

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  2. 2023Segment Anythingcited by 8,234
  3. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,974
  4. 2017On Calibration of Modern Neural Networkscited by 4,960
  5. 2023BiomedCLIP: a multimodal biomedical foundation model pretrained from fifteen million scientific image-text pairscited by 520
  6. 2022MONAI: An open-source framework for deep learning in healthcarecited by 448
  7. 2019Beyond temperature scaling: Obtaining well-calibrated multiclass probabilities with Dirichlet calibrationcited by 364
  8. 2019Verified Uncertainty Calibrationcited by 309
  9. 2018A Probabilistic U-Net for Segmentation of Ambiguous Imagescited by 283
  10. 2019Limitations of the Empirical Fisher Approximation for Natural Gradient Descentcited by 217
  11. 2016Unsupervised domain adaptation in brain lesion segmentation with adversarial networkscited by 125
  12. 2020DoDNet: Learning to segment multi-organ and tumors from multiple partially labeled datasetscited by 94
10 more
  1. 2022Fisher SAM: Information Geometry and Sharpness Aware Minimisationcited by 84
  2. 2023CXR-CLIP: Toward Large Scale Chest X-ray Language-Image Pre-trainingcited by 82
  3. 2017Entropy-SGD optimizes the prior of a PAC-Bayes bound: Generalization properties of Entropy-SGD and data-dependent priorscited by 81
  4. 2019Assessing Reliability and Challenges of Uncertainty Estimations for Medical Image Segmentationcited by 75
  5. 2023Segment Anything Model for Medical Images?cited by 39
  6. 2023Dual-Reference Source-Free Active Domain Adaptation for Nasopharyngeal Carcinoma Tumor Segmentation across Multiple Hospitalscited by 25
  7. 2021On the Variance of the Fisher Information for Deep Learningcited by 22
  8. 2021Calibrating the Dice loss to handle neural network overconfidence for biomedical image segmentationcited by 17
  9. 2023Threshold-Consistent Margin Loss for Open-World Deep Metric Learningcited by 7
  10. 2025Confidence-calibrated covariate shift correction for few-shot classification in Vision-Language Modelscited by 6

Built on by

Not cited yet.