Benchmarking Self-Supervised Learning Methods for Accelerated MRI Reconstruction
eess.IV, cs.LG
Submitted: 2025-02-19
Updated: 2026-09-11
Comments: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) https://melba-journal.org/2026:028
Journal ref: Machine.Learning.for.Biomedical.Imaging. 2026 (2026)
DOI: 10.59275/j.melba.2026-b6a3
Code: https://github.com/Andrewwango/ssibench
License: http://creativecommons.org/licenses/by/4.0/
The gist: Reconstructing MRI from highly undersampled measurements is crucial for accelerating medical imaging, but is challenging due to the ill-posedness of the inverse problem.
Terminology
Abstract
Reconstructing MRI from highly undersampled measurements is crucial for accelerating medical imaging, but is challenging due to the ill-posedness of the inverse problem. While supervised deep learning (DL) approaches have shown remarkable success, they traditionally rely on fully-sampled ground truth (GT) images, which are expensive or impossible to obtain in real scenarios. This problem has created a recent surge in interest in self-supervised learning methods that do not require GT. Although recent methods are now fast approaching "oracle" supervised performance, the lack of systematic comparison and standard experimental setups are hindering targeted methodological research and precluding widespread trustworthy industry adoption. We present SSIBench, a modular and flexible comparison framework to unify and thoroughly benchmark Self-Supervised Imaging methods (SSI) without GT. We focus on end-to-end trained DL methods, which do not require long inference time, large datasets, or per-image training. We evaluate 21 such recent methods across seven realistic MRI scenarios on real data, showing a wide performance landscape whose method ranking differs across scenarios and metrics, exposing the need for further SSI research. To accelerate reproducible research and lower the barrier to entry, we provide the extensible benchmark and open-source reimplementations of all methods at https://github.com/Andrewwango/ssibench, allowing researchers to rapidly and fairly contribute and evaluate new methods on the standardised setup for potential leaderboard ranking, or benchmark existing methods on custom datasets, forward operators, or models, unlocking the application of SSI to other valuable nascent GT-free scientific imaging modalities.
Sources
- Ambient Diffusion: Learning Clean Distributions from Corrupted Data
- A Survey on Diffusion Models for Inverse Problems
- Accelerated MRI with Un-trained Neural Networks
- VORTEX: Physics-Driven Data Augmentations Using Consistency Training for Robust Accelerated MRI Reconstruction
- Noise2Recon: Enabling Joint MRI Reconstruction and Denoising with Semi-Supervised and Self-Supervised Learning
- SKM-TEA: A Dataset for Accelerated MRI Reconstruction with Dense Image Labels for Quantitative Clinical Evaluation
- SPICER: Self-Supervised Learning for MRI with Automatic Coil Sensitivity Estimation and Reconstruction
- Self-Supervised Noise Adaptive MRI Denoising via Repetition to Repetition (Rep2Rep) Learning
- GSURE-Based Diffusion Model Training with Corrupted Data
- Clean self-supervised MRI reconstruction from noisy, sub-sampled training data with Robust SSDU
- Noisier2Noise: Learning to Denoise from Unpaired Noisy Data
- Unsupervised Accelerated MRI Reconstruction via Ground-Truth-Free Flow Matching
- Unsupervised Learning with Stein's Unbiased Risk Estimator
- Self-Supervised Learning from Noisy and Incomplete Data
- UNSURE: self-supervised learning with Unknown Noise level and Stein's Unbiased Risk Estimate
- Unsupervised Deep Basis Pursuit: Learning inverse problems without ground-truth data
- K-band: Self-supervised MRI Reconstruction via Stochastic Gradient Descent over K-space Subsets
- Benchmarking 3D multi-coil NC-PDNet MRI reconstruction
- Reconstruct Anything Model: a lightweight general model for computational imaging
- Unsupervised Data Augmentation for Consistency Training
Related papers
- Revisiting Integration of Image and Metadata for DICOM Series Classification: Cross-Attention and Dictionary Learning
- VesselSDF: Distance Field Priors for Vascular Network Reconstruction
- cSVR: Convolutional Slice-to-Volume Reconstruction
- NAIMA: Semantics Aware RGB Guided Depth Super-Resolution
- AneumoBench: A Source-Linked Benchmark for Synthetic-Geometry Transfer in Aneurysm CFD
- RETO: A Rotary-Enhanced Transformer Operator for High-Fidelity Prediction of Automotive Aerodynamics