mlsanityData to 14 Oct 2026

Evaluating the Safety of Deep Learning-Based Brain MRI Reconstruction

arXiv 2608.28714 ↗First posted 28 Aug 2026eess.IV cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Objective: Deep learning accelerates brain MRI four- to tenfold, but models can erase lesions or synthesize false tissue - failures pixel-averaged metrics like PSNR and SSIM miss. We review whether current evaluation practices detect this blind spot. Methods: Following PRISMA 2020, we searched seven databases without date limits, including 263 studies (1995-2026), appraised them using QUADAS-2 and matched instruments, and synthesized narratively. Categories were derived from titles, abstracts, and controlled vocabulary; reported prevalence figures represent floors. Duplicate screening achieved high agreement (Fleiss kappa = 0.877), as did appraisal (0.788; 0.390 where observable). Extraction is unaudited. Results: Only 18 of 263 studies (6.8%) recorded both a fidelity metric and reader assessment on identical data, leaving the central surrogate unmeasured. Reader studies mostly measured inter-reader agreement, which was weak: fastMRI 2020 concordance reached 0.457 and 0.386 (Kendall W), improving only where SSIM diverged. Erasing a 100 mm3 lacunar infarct shifts global PSNR by 0.03 dB under the stated error model. As the corpus grew fivefold, reader assessments dropped from 32% to 18%, recovering to 21%. Generative models - most associated with hallucination (39%) - were among the least reader-evaluated (11.3%), while self-supervised models reached 47% with zero reader evaluation. Only 5% released code and ran reader studies; none evaluated a model observer; no named dataset covered acute stroke or hemorrhage. Conclusions: On these floors, current evaluation practices cannot certify diagnostic safety. We derive five requirements safety-oriented evaluations must meet.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 51 of its 299 references that are in the map. First posted on arXiv 28 Aug 2026.

Builds on

51 of 299 references.

References · 51

  1. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 9,122
  2. 2017MoDL: Model Based Deep Learning Architecture for Inverse Problemscited by 387
  3. 2017Learning a Variational Network for Reconstruction of Accelerated MRI Datacited by 342
  4. 2022Diffusion Models for Medical Image Analysis: A Comprehensive Surveycited by 273
  5. 2017A Deep Cascade of Convolutional Neural Networks for Dynamic MR Image Reconstructioncited by 232
  6. 2019On instabilities of deep learning in image reconstruction - Does AI come at a cost?cited by 212
  7. 2019Causality matters in medical imagingcited by 193
  8. 2022Adaptive Diffusion Priors for Accelerated MRI Reconstructioncited by 118
  9. 2022ISLES 2022: A multi-center magnetic resonance imaging stroke lesion segmentation datasetcited by 106
  10. 2019Deep Learning Methods for Parallel Magnetic Resonance Image Reconstructioncited by 81
  11. 2020On hallucinations in tomographic image reconstructioncited by 68
  12. 2022What is Healthy? Generative Counterfactual Diffusion for Lesion Localizationcited by 52
38 more
  1. 2020The troublesome kernel -- On hallucinations, no free lunches and the accuracy-stability trade-off in inverse problemscited by 40
  2. 2023Self-Supervised MRI Reconstruction with Unrolled Diffusion Modelscited by 38
  3. 2023Guided Reconstruction with Conditioned Diffusion Models for Unsupervised Anomaly Detection in Brain MRIscited by 18
  4. 2021Stacked U-Nets with Self-Assisted Priors Towards Robust Correction of Rigid Motion Artifact in Brain MRIcited by 14
  5. 2023Learning Fourier-Constrained Diffusion Bridges for MRI Reconstructioncited by 14
  6. 2023SADIR: Shape-Aware Diffusion Models for 3D Image Reconstructioncited by 12
  7. 2020Machine Learning in Magnetic Resonance Imaging: Image Reconstructioncited by 8
  8. 2024Diff-Def: Diffusion-Generated Deformation Fields for Conditional Atlasescited by 8
  9. 2025Regression is all you need for medical image translationcited by 8
  10. 2024Latent Causal Modeling for 3D Brain MRI Counterfactualscited by 7
  11. 2021Alternating Learning Approach for Variational Networks and Undersampling Pattern in Parallel MRI Applicationscited by 6
  12. 2021Sinogram upsampling using Primal-Dual UNet for undersampled CT and radial MRI reconstructioncited by 6
  13. 2022Distortion-Corrected Image Reconstruction with Deep Learning on an MRI-Linaccited by 6
  14. 2023Zero-DeepSub: Zero-Shot Deep Subspace Reconstruction for Rapid Multiparametric Quantitative MRI Using 3D-QALAScited by 5
  15. 2021Model-based Synthetic Data-driven Learning (MOST-DL): Application in Single-shot T2 Mapping with Severe Head Motion Using Overlapping-echo Acquisitioncited by 4
  16. 2023SISMIK for brain MRI: Deep-learning-based motion estimation and model-based motion correction in k-spacecited by 4
  17. 2024NPB-REC: A Non-parametric Bayesian Deep-learning Approach for Undersampled MRI Reconstruction with Uncertainty Estimationcited by 4
  18. 2024Non-Reference Quality Assessment for Medical Imaging: Application to Synthetic Brain MRIscited by 4
  19. 2024Robust multi-coil MRI reconstruction via self-supervised denoisingcited by 4
  20. 2024Enhancing Reconstruction-Based Out-of-Distribution Detection in Brain MRI with Model and Metric Ensemblescited by 4
  21. 2025A Physics-Informed Deep Learning Model for MRI Brain Motion Correctioncited by 4
  22. 2025Sparsity-Driven Parallel Imaging Consistency for Improved Self-Supervised MRI Reconstructioncited by 4
  23. 2025DIMA: DIffusing Motion Artifacts for unsupervised correction in brain MRI imagescited by 3
  24. 2020Spatiotemporal Flexible Sparse Reconstruction for Rapid Dynamic Contrast-enhanced MRIcited by 2
  25. 2023Towards retrospective motion correction and reconstruction for clinical 3D brain MRI protocols with a reference contrastcited by 2
  26. 2023DD-CISENet: Dual-Domain Cross-Iteration Squeeze and Excitation Network for Accelerated MRI Reconstructioncited by 2
  27. 2024Diffusion based multi-domain neuroimaging harmonization method with preservation of anatomical detailscited by 2
  28. 2024MCDDPM: Multichannel Conditional Denoising Diffusion Model for Unsupervised Anomaly Detection in Brain MRIcited by 2
  29. 2025Assessing the use of Diffusion models for motion artifact correction in brain MRIcited by 2
  30. 2025Highly Undersampled MRI Reconstruction via a Single Posterior Sampling of Diffusion Modelscited by 2
  31. 2025MRIQT: Physics-Aware Diffusion Model for Image Quality Transfer in Neonatal Ultra-Low-Field MRIcited by 2
  32. 2023Quantitative MR Image Reconstruction using Parameter-Specific Dictionary Learning with Adaptive Dictionary-Size and Sparsity-Level Choicecited by 1
  33. 2024Iterative Learning for Joint Image Denoising and Motion Artifact Correction of 3D Brain MRIcited by 1
  34. 2024Contrast-Optimized Basis Functions for Self-Navigated Motion Correction in Quantitative MRIcited by 1
  35. 2025Automatic quality control in multi-centric fetal brain MRI super-resolution reconstructioncited by 1
  36. 2025Res-MoCoDiff: Residual-guided diffusion models for motion artifact correction in brain MRIcited by 1
  37. 2025RealDeal: Enhancing Realism and Details in Brain Image Generation via Image-to-Image Diffusion Modelscited by 1
  38. 2025Lesion-Aware Post-Training of Latent Diffusion Models for Synthesizing Diffusion MRI from CT Perfusioncited by 1

50 most cited of 51 All 51 (JSON)

Built on by

Not cited yet.