{"slug": "card-calibration-via-agreement-in-reverse-diffusion-for-out-of-domain-mri", "title": "CARD: Calibration via Agreement in Reverse Diffusion for Out-of-Domain MRI Segmentation", "summary": "A new method called CARD (Calibration via Agreement in Reverse Diffusion) improves out-of-domain MRI segmentation calibration, lowering calibration error in 45 of 49 comparisons across cardiac, prostate, and brain MRI shifts. The approach, detailed in arXiv:2608.28681v1, leverages categorical diffusion's generative shape prior and reverse-step class distributions to correlate with Dice at 0.788 versus 0.521 for a matched discriminative control, enabling per-pixel temperature adjustment without altering segmentations.", "body_md": "arXiv:2608.28681v1 Announce Type: new\nAbstract: Probability calibration aligns model confidence with predictive accuracy, enabling clinicians to identify unreliable segmentation regions. This alignment breaks down under domain shift, where artifacts and unseen protocols produce confident errors. Existing post-hoc methods adapt the correction at test time, conditioning on predictive entropy, the logit pattern, or augmentation response, but each proxy is read from the terminal prediction, the very quantity that shift corrupts. This motivates reliability evidence beyond the terminal prediction, which categorical diffusion provides in two ways. First, a generative shape prior keeps a capacity-limited reference intact when appearance is corrupted, so its disagreement with the primary segmentor highlights primary-model errors. Second, every reverse step yields a class distribution, separating persistent disagreement from transient discrepancy. Aggregated over the trajectory, this disagreement correlates with Dice at 0.788, against 0.521 for a matched discriminative control. We therefore propose CARD (Calibration via Agreement in Reverse Diffusion), which maps the temporal aggregate of this disagreement to a temperature field applied per pixel across all classes, so that confidence changes while the segmentation does not. Across cardiac, prostate and brain MRI shifts, CARD lowers calibration error in 45 of 49 comparisons against the strongest baseline in each setting.", "url": "https://wpnews.pro/news/card-calibration-via-agreement-in-reverse-diffusion-for-out-of-domain-mri", "canonical_source": "https://arxiv.org/abs/2608.28681", "published_at": "2026-09-01 04:00:00+00:00", "updated_at": "2026-09-01 04:22:37.346818+00:00", "lang": "en", "topics": ["machine-learning", "artificial-intelligence", "computer-vision"], "entities": ["CARD", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/card-calibration-via-agreement-in-reverse-diffusion-for-out-of-domain-mri", "markdown": "https://wpnews.pro/news/card-calibration-via-agreement-in-reverse-diffusion-for-out-of-domain-mri.md", "text": "https://wpnews.pro/news/card-calibration-via-agreement-in-reverse-diffusion-for-out-of-domain-mri.txt", "jsonld": "https://wpnews.pro/news/card-calibration-via-agreement-in-reverse-diffusion-for-out-of-domain-mri.jsonld"}}