CS Paper Radar

VariaFlow: Uncertainty-Aware Flow Matching for Unsupervised Anomaly Detection

Arnela Hadzic, Martin Urschler, Cosmin I. Bercea

cs.CV ·submitted 2026-10-30 09:48 UTC · v1
published

Abstract

Reconstruction-based anomaly detection no longer fails at reconstruction. Generative models produce high-quality pseudo-healthy reconstructions, and the remaining difficulty is the step of scoring residuals. The residual between a scan and its reconstruction has two indistinguishable sources, genuine pathology and the anatomical variation that separates one healthy brain from another. Scoring against a single reconstruction treats both as anomalous, inflating false positives where healthy anatomy varies most while small or subtle lesions are lost in the surrounding reconstruction error. We identify the missing quantity as the predictive variance, how much a healthy reconstruction can vary at each location. We introduce VariaFlow, which estimates this variance from an ensemble of mask-guided restorations under an unconditional flow matching prior, each started from a different noise initialization, and which normalizes the residual by the resulting standard deviation. Few-step flow matching keeps the ensemble computationally affordable, and restoring at two spatial scales captures lesions of different size. Across three brain MRI datasets spanning stroke and tumor, VariaFlow is trained once on healthy anatomy and generalizes across all three, attaining the highest AUROC (0.94, 0.80, 0.78) on each and the highest AUPRC on both cross-domain datasets, exceeding the strongest baseline by 37% and 50%. Under domain shift, the same baseline drops by up to 52% in AUPRC, whereas VariaFlow stays more stable and achieves the best small-lesion Dice on all three datasets.

Comment
Accepted for WACV 2027

Closest papers

Nearest neighbours by meaning across the whole index.

Unsupervised Brain Anomaly Detection as a Bayesian Inverse Problem with Diffusion Prior

cs.AI cs.CV ·2026-09-21 ·Hugues Roy, Reuben Dorent, Ninon Burgos
top venuecode/data similarity 0.68
Abstract

Unsupervised anomaly detection (UAD) aims to localize abnormal regions in medical scans without pixel-level annotations. A typical strategy seeks to reconstruct a pseudo-healthy image that preserves subject-specific anatomy. Recently, diffusion models have been proposed to perform UAD. However, these methods rely on heuristic noise schedules or synthetic corruptions to balance subject-specificity and anomaly removal. In this work, we propose an alternative formulation of UAD as a Bayesian inverse problem under a diffusion prior. First, we introduce a latent spatial anomaly mask that models pixel-wise consistency between a test image and its latent corresponding pseudo-healthy image. Then, we propose an approximation of the unknown generation process that links healthy anatomy, anomalies, and the observed image, enabling a well-defined likelihood within the Bayesian framework. Building on recent advances in diffusion-based inverse problem methods, we jointly infer the pseudo-healthy image and the anomaly mask via annealed posterior sampling. We evaluate our approach on FDG PET (ADNI) and FLAIR MRI (BraTS 2021), demonstrating improved anomaly localization performance compared to other diffusion-based approaches and validating the contribution of our introduced model. Our code is available at https://github.com/HuguesRoy/UAD_DAPS.

0.45

RARF: Region-Aware Rectified Flows for 3D Brain MRI Inpainting

cs.CV cs.AIcs.LG ·2026-09-03 ·Tomas Guija-Valiente, Blanca Rodriguez-Gonzalez, Norberto Malpica, Angel Torrado-Carvajal
code/data similarity 0.66
Abstract · 11 pages, 2 figures. Preprint version corresponding to the initial submission prior to peer review, submitted as part of our participation i…

Medical image inpainting has the potential to improve automated brain MRI analysis by reconstructing healthy tissue within pathological regions. We introduce RARF, a task-agnostic region-aware rectified flow framework for masked data generation. We instantiate the framework for 3D brain MRI inpainting as our submission to the BraTS Inpainting Challenge 2026. RARF restricts the stochastic interpolation process to the inpainting region, while the observed voxels remain fixed and provide patient-specific anatomical context. A three-dimensional neural network receives the partially voided image, with Gaussian noise filling the missing region, together with the inpainting mask and the corresponding timestep. The model is trained using masked flow-matching and reconstruction-consistency objectives, combined with mask-aware preprocessing and data augmentation. During inference, the learned velocity field transports the initial noise toward a plausible reconstruction of the missing tissue, which is then combined with the unchanged observed anatomy. Experiments under the BraTS evaluation protocol show that the proposed approach produces competitive reconstructions while maintaining anatomical consistency. Source code is available at: https://github.com/TomasGuija/rarf.

0.41

MOSAIC-Flow: Orthogonal-Axis and Status-Conditioned Fusion for Missing-Sequence Completion in Brain Tumor MRI

cs.CV ·2026-10-29 ·Pengli Zhu, Zhihao Peng, Yixuan Yuan
code/data similarity 0.64
Abstract · 5 pages, 4 figures

Missing-sequence completion is important for comprehensive brain tumor characterization, but cross-modal fusion can obscure sequence-specific contrast and lose fine-grained three-dimensional tumor detail. To address this challenge, we introduce Modality-aware Orthogonal-axis Status-conditioned Attentive Integration for Completion (MOSAIC-Flow), a unified framework for missing-sequence completion that learns a target-conditioned velocity field while keeping the seven acquired volumes fixed throughout the transport. Specifically, modality-aware stems independently encode each sequence, and orthogonal-axis refinement captures axial, coronal, and sagittal dependencies while retaining the three-dimensional representation. Status-conditioned prototype fusion uses compact modality and availability tokens to route and modulate pooled queries from the fused visual representation, enabling target-adaptive conditioning without dense voxel-to-voxel attention. A conditional flow matching objective transports target-channel noise toward the single missing sequence, while four Heun trajectories are averaged to produce the completion. Evaluation across structural, diffusion, and contrast-enhanced MRI in UCSF-PDGM demonstrates consistent whole-brain and tumor-region fidelity across all eight single-target completion tasks, with the strongest performance on structural sequences. Code is available at https://github.com/CUHK-AIM-Group/MOSAIC-FLOW.

0.21

Uncertainty-Aware 3D Residual Wavelet Diffusion for Ultra Low-Field MRI Super-Resolution

cs.CV ·2026-09-21 ·Rui W. Yeow, Millie Beament, Fred Dick, Raha Razin, Martina Bocchetta +4 more
published similarity 0.64
Abstract · 11 pages, 3 figures, 1 table. Accepted at SASHIMI 2026 (MICCAI 2026 workshop). This is the version submitted for peer review

Ultra low-field MRI expands global access to neuroimaging but produces scans with low signal-to-noise ratio, reduced contrast, and thick slices. While regression-based super-resolution can recover anatomical detail for segmentation, it returns a single deterministic estimate that gives no indication of regions where the low-field input leaves anatomy underdetermined. Generative diffusion models offer an alternative by sampling the posterior distribution of plausible high-field images, quantifying this anatomical ambiguity. However, applying them to 3D whole-brain MRI is restricted by memory bottlenecks, slow sampling, and scanner domain shifts. We propose a 3D residual wavelet diffusion model that combines three ideas to overcome these hurdles. A lossless wavelet reparameterisation shrinks the spatial grid to fit a whole brain on a single GPU, residual shifting accelerates sampling by starting from the low-field input, and domain randomisation promotes scanner generalisation without paired training data. As the high-field reference is not a voxel-aligned ground truth, we evaluate downstream volumetric agreement. On a healthy cohort (n=19) imaged at 0.064T and 3T, our method matches a leading general-purpose regression approach in volumetric accuracy while additionally generating per-voxel uncertainty maps highlighting underdetermined regions. Furthermore, on a pilot dataset (n=11) of participants with cognitive impairment, disease-relevant atrophy is preserved rather than normalised towards a healthy prior. Our framework brings whole-brain posterior sampling to low-field super-resolution without sacrificing volumetric accuracy.

0.35

CAPS: Clinically-Verifiable Anatomy-Prioritized Synthesis for Missing MRI Imputation in Brain Tumor Segmentation

eess.IV cs.CV ·2026-10-20 ·Cecilia Diana-Albelda, Arthur Longuefosse, Mario Rubén Paz Campos, Álvaro García-Martín, Jesus Bescos
similarity 0.63
Abstract · 20 pages, 6 figures, 8 tables. Cecilia Diana-Albelda and Arthur Longuefosse contributed equally

Multi-parametric MRI is the gold standard for neuro-oncology, yet acquiring complete 4-modality protocols may be unfeasible in real-world clinical practice. When handling missing modalities, existing automated systems force a critical compromise: robust segmenters operate as uninterpretable "black boxes" that bypass image generation, while standard synthesis models often produce blurry reconstructions lacking diagnostic reliability. To address this issue, we propose CAPS, a Clinically-Verifiable Anatomy-Prioritized Synthesis framework that reconciles computational precision with clinical transparency. Our multi-modal generative architecture is driven by a hybrid objective function (L_1 + L_AFP + L_DualDice) that explicitly enforces the preservation of deep anatomical features and strict topological boundaries, rather than relying solely on pixel-wise optimization. Evaluated comprehensively on the BraTS Adult Glioma dataset, the framework achieves superior synthesis fidelity across intensity-based metrics, and demonstrates robust downstream segmentation performance (82.38% mean Dice), successfully restoring verifiable visual context. Furthermore, the framework exhibits exceptional zero-shot cross-domain generalization, maintaining high structural fidelity across completely unseen Meningioma, Pediatric, and Sub-Saharan African cohorts without domain-specific adaptation. Finally, a blinded radiological assessment confirms that our anatomy-aware approach significantly enhances both perceived image quality and clinical utility, providing clinicians with the reliable visual evidence required to trust downstream automated workflows.

0.20

SliceBridge: Context-Consistent Repair of Corrupted Slice Intervals in T1-Weighted MRI

cs.CV ·2026-09-01 ·Jiheng Li, Michael E. Kim, Trent M. Schwartz, Gaurav Rudravaram, Derek B. Archer +4 more
similarity 0.63
Abstract

Structural magnetic resonance imaging (MRI) images are sometimes corrupted over a contiguous set of slices, where acquisition, motion, hardware, or reconstruction effects leave a single slice or short interval inconsistent with its neighbors while the rest of the image remains usable. Such localized corruption can bias downstream morphometric analysis, yet discarding or reacquiring an otherwise usable image is costly. We formulate this as an image restoration problem: given the location of the affected interval, reconstruct those slices from the surrounding anatomical and imaging context. We propose SliceBridge, a framework for restoring corrupted slice intervals in T1-weighted MRI using rectified flow matching conditioned on the surrounding intact slices and their relative slice positions. Through-plane consistency is encouraged by coupling the slices within the interval through interval-correlated initial noise, a shared flow time, and synchronized sampling. The restored interval is then inserted back, leaving all other slices unchanged. We trained and validated the model on 9,877 T1-weighted brain MRI volumes from four datasets and evaluated it on 581 external subjects using clean interval withholding and controlled corruptions. Compared with a matched model that reconstructed target slices independently, SliceBridge reduced error in slice-to-slice changes within repaired intervals by 32.9%-41.3% across interval lengths and achieved higher SSIM at every interval length. In controlled-corruption cases, SliceBridge reduced the median error in regional brain volume estimates produced by a downstream segmentation model from 1.95% in corrupted volumes to 1.05%.

0.31

Beyond Training from Scratch: Foundation Models for Data-Efficient and Generalizable Cardiac MRI Reconstruction

cs.CV ·2026-10-06 ·Anam Hashmi, Mayug Maniparambil, Julia Dietlmeier, Kathleen M. Curran, Noel E. O'Connor
publishedcode/data similarity 0.63
Abstract · Accepted at ECCVW 2026

Cardiac magnetic resonance imaging reconstruction aims to recover high-quality images from undersampled acquisitions, enabling faster scans while preserving diagnostic fidelity. Recent reconstruction methods are typically trained from scratch and often require large amounts of task-specific data, limiting their robustness under data scarcity and distribution shifts. In this work, we investigate whether pretrained vision foundation models can serve as effective priors for accelerated cardiac MRI reconstruction. We propose a reconstruction framework that integrates frozen and parameter-efficiently adapted visual encoders, including CLIP, BiomedCLIP, and DINOv2, within a transformer-based reconstruction architecture. Extensive experiments on the CMRxRecon2023 and CMRxRecon2024 benchmarks demonstrate that pretrained representations consistently outperform a transformer trained from scratch across multiple acceleration factors. We further evaluate performance under limited supervision and cross-dataset transfer, showing that foundation models provide superior data efficiency and generalization. While frozen representations are particularly effective in extreme low-data regimes, Low-Rank Adaptation (LoRA) yields additional gains when moderate amounts of training data are available. Among the evaluated backbones, DINOv2 achieves the strongest overall performance. These findings highlight the potential of vision foundation models as robust and transferable priors for cardiac MRI reconstruction.

0.31

Unsupervised Anomaly Detection for Image Dataset Quality Assurance in Multi-Center Breast MRI

cs.CV cs.AI ·2026-08-17 ·Chiara Tappermann, Steffen Renisch, Lars Ole Schwen, Hans Meine, Horst K. Hahn +1 more
benchmark similarity 0.63
Abstract

Corrupted, inconsistent, or anomalous data silently threatens the safety and reliability of medical AI. Despite growing regulatory recognition of dataset quality assurance (QA) for high-risk medical AI, scalable automated detection remains underdeveloped. We employ unsupervised anomaly detection (AD) and out-of-distribution (OOD) detection as an automated dataset QA mechanism for multi-center dynamic contrast-enhanced breast MRI. We build a controlled AD benchmark of 17 realistic QA-relevant anomaly types from six public datasets (protocol violations, processing errors, incorrect anatomical regions) and propose a taxonomy of radiological image anomalies based on human visual perception, enabling fine-grained analysis of AD failure modes. The benchmark includes near-, medium-far-, far-OOD samples, as well as in-distribution and external normal data. Four methods are evaluated: a projection-based method extended with a domain-specific feature extractor and a novel positional encoding, a reconstruction-based approach extended to full 3D volumes with an augmented training objective, and two unmodified hybrid OOD detection methods. Medium-far- and far-OOD samples are detected reliably, whereas near-OOD samples and external normal data from unseen institutions expose method-specific differences. The 3D reconstruction-based approach best balances detection performance (AUROC: 0.936) and generalization to unseen institutions. The projection-based method with positional encoding achieves the highest overall detection performance (AUROC: 0.954). Both hybrid methods exhibit critical failure modes, confirming that methods validated for one modality or anatomy may not generalize without domain-specific adaptation. Implants and mastectomies remain an open challenge for all methods. Our results establish a foundation and practical guidance on scalable unsupervised QA in medical AI pipelines.

0.20