CARD: Calibration via Agreement in Reverse Diffusion for Out-of-Domain MRI Segmentation

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, $\mathcal{L}_{\mathrm{CalSAM}}$ 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$\to$GE) CalSAM shows a $+7.4\%$ relative improvement in $\mathrm{DSC}$ (80.1% vs.\ 74.6%), a $-26.9\%$ reduction in $\mathrm{HD95}$ (4.6 mm vs.\ 6.3 mm), and a $-39.5\%$ reduction in $\mathrm{ECE}$ (5.2% vs.\ 8.6%). On ATLAS-C (motion corruptions), CalSAM achieves a $+5.3\%$ relative improvement in $\mathrm{DSC}$ (75.9%) and a $-32.6\%$ reduction in $\mathrm{ECE}$ (5.8%). Ablations show FIP and CMP contribute complementary gains ($p<0.01$), and the Fisher penalty incurs a modest $\sim$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.

Reliable confidence estimates are essential in semantic segmentation, yet modern models often remain miscalibrated. We investigate two overlooked issues in post-hoc calibration. First, adding a constant to all logits leaves softmax probabilities unchanged, but several standard calibrators depend on this arbitrary offset. In segmentation, this offset can vary across pixels or voxels, introducing spatially varying representation dependence. We characterize translation-invariant (TI) calibrators and construct TI counterparts of shift-sensitive methods. Second, calibrating with cross-entropy can degrade segmentation quality due to mismatched training and calibration objectives and limited calibration data. We investigate decision-preserving calibration under argmax- and order-preservation constraints. Since these constraints restrict affine softmax calibrators to temperature scaling, we introduce more expressive class-conditional affine calibrators that preserve decisions. Across natural-image and medical segmentation benchmarks, including corruption-based covariate shift, TI variants generally improve calibration, while decision-preserving variants prevent segmentation degradation by construction and retain strong calibration performance. Our findings provide practical design principles for post-hoc calibration in semantic segmentation.

Clinical cardiac imaging pipelines currently deploy separate models for each dataset and modality, incurring redundant training costs and precluding knowledge sharing across anatomically related tasks. Consolidating semi-supervised learning, unsupervised domain adaptation, and domain generalisation into one model is therefore a practical necessity, yet naive joint training exposes a fundamental barrier: conflicting label semantics between datasets collapse LA Dice from 90.31% to 83.38%, while gradient imbalance across tasks of unequal complexity suppresses the weaker tasks throughout training. We present UniT-Diff, a unified diffusion segmentation framework that resolves these conflicts through three targeted mechanisms. An 11-channel task-specific output space physically partitions label categories, eliminating cross-task gradient sign reversal by construction. SNR-Adaptive Task Conditioning (SATC) scales the task token by the log signal-to-noise ratio of the current diffusion timestep, suppressing domain-specific bias during coarse denoising and restoring full task guidance as the signal clears. Task-Type-Aware Conditional Dropout (TTACD) permanently removes the task token for domain-generalisation inputs, routing them through a shared neutral pathway that draws on cross-dataset cardiac anatomy rather than source-vendor statistics. Under a single parameter set, UniT-Diff surpasses independently trained task-specific baselines on all three benchmarks simultaneously: +0.87% on LA, +1.77% on MMWHS, and +0.88% on MNMS.