aiqbench analysis
This paper proposes a lightweight variational DNN plugin that refines degraded medical segmentation maps when scanner or protocol shifts push inputs out of distribution. It learns anatomical geometry priors from small in-distribution datasets, aiming to avoid target-domain data and retraining, and it provides uncertainty estimates without sampling. For ML readers, the claims rest on a softmax reinterpretation and benchmarks across 14 public datasets against 15 baselines, so the main things to check are the generalization gains and whether the uncertainty estimates are well calibrated.
Learning a Sampling-Free Variational DNN Plugin from Tiny Training Sets to Refine OOD Segmentation With Uncertainty Estimation
arXiv:2606.15837v3 Announce Type: replace-cross Abstract: Deep neural networks (DNNs) frequently fail to generalize to out-of-distribution (OOD) medical images because of variations in scanners and acquisition protocols. Retraining DNN models to address these distribution shifts is often impractical due to the high cost of acquiring and annotating new medical datasets. To address this, we introduce VarDeepPCA, a novel lightweight variational DNN framework designed to restore/refine degraded segmentation maps by leveraging intrinsic geometric priors. Unlike existing approaches that require target-domain data or extensive pre-training, our VarDeepPCA explicitly learns a distribution of valid anatomical geometries using only small in-distribution (ID) datasets. Theoretically, our novel variational learning framework leverages a reinterpretation of the softmax mapping to implicitly perform exact distribution modeling, thereby enabling computationally efficient, sampling-free learning and inference. This also enables VarDeepPCA to provide uncertainty estimates associated with its restored segmentation maps. We empirically validate our framework across 4 distinct clinical applications, using 14 publicly available datasets, involving segmentation of the myocardium, neuroretinal rim, prostate, and fetal head. Comparisons against 15 existing methods demonstrate that VarDeepPCA consistently restores segmentation maps produced by the existing methods on OOD data to (i) significantly improve anatomical plausibility of geometries and clinical utility of the segmentations, and (ii) significantly reduce errors, without needing any more training data than that used by existing methods.
Abstract:Deep neural networks (DNNs) frequently fail to generalize to out-of-distribution (OOD) medical images because of variations in scanners and acquisition protocols. Retraining DNN models to address these distribution shifts is often impractical due to the high cost of acquiring and annotating new medical datasets. To address this, we introduce VarDeepPCA, a novel lightweight variational DNN framework designed to restore/refine degraded segmentation maps by leveraging intrinsic geometric priors. Unlike existing approaches that require target-domain data or extensive pre-training, our VarDeepPCA explicitly learns a distribution of valid anatomical geometries using only small in-distribution (ID) datasets. Theoretically, our novel variational learning framework leverages a reinterpretation of the softmax mapping to implicitly perform exact distribution modeling, thereby enabling computationally efficient, sampling-free learning and inference. This also enables VarDeepPCA to provide uncertainty estimates associated with its restored segmentation maps. We empirically validate our framework across 4 distinct clinical applications, using 14 publicly available datasets, involving segmentation of the myocardium, neuroretinal rim, prostate, and fetal head. Comparisons against 15 existing methods demonstrate that VarDeepPCA consistently restores segmentation maps produced by the existing methods on OOD data to (i) significantly improve anatomical plausibility of geometries and clinical utility of the segmentations, and (ii) significantly reduce errors, without needing any more training data than that used by existing methods.
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Methodology (stat.ME); Machine Learning (stat.ML)
arXiv:2606.15837 [cs.CV]
(or arXiv:2606.15837v3 [cs.CV] for this version)
Machine.Learning.for.Biomedical.Imaging. 2026 (2026)
https://doi.org/10.59275/j.melba.2026-6d54 DOI(s) linking to related resources
From: Jimut Bahan Pal Mr. [view email] [v1] Sun, 14 Jun 2026 14:36:59 UTC (29,487 KB) [v2] Tue, 30 Jun 2026 09:24:22 UTC (29,487 KB) [v3] Mon, 12 Oct 2026 15:47:21 UTC (29,470 KB)
Full text under the publisher's open licence · 276 words
Related coverage
Learning a Sampling-Free Variational DNN Plugin from Tiny Training Sets to Refine OOD Segmentation With Uncertainty Estimation
arXiv · Jun 16, 2026
[AINews] not much happened today
Latent Space · Oct 3, 2026
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · Sep 4, 2026
AUP Learning Cloud: Streamlining AI Education on AMD
AMD · Aug 6, 2026