deepMRI: Deep Learning Methods for MRI
April 29, 2026 · View on GitHub
Authors: Yang Gao, Zhuang Xiong, Xuanyu Zhu, Samiha Prima, Hongfu Sun
This repository is a collection of deep learning tools for MRI reconstruction and quantitative mapping. Each method lives in its own standalone repository (linked below). Each subfolder here contains a short README that links directly to the corresponding standalone repo.
Clinicians / non-coders: iQSM+ is available as a browser-based web app — no MATLAB or command line needed. See the iQSM+ repo to get started.
Projects
| Method | Task | Paper |
|---|---|---|
| AFTER-QSM | QSM for oblique/anisotropic scans | NeuroImage 2022 |
| BFRnet | QSM background field removal | Z Med Phys 2022 |
| DCRNet | QSM + R2* acceleration | NeuroImage 2021 |
| DeepRelaxo | Fast brain R2* mapping | MRM 2026 |
| DIP-UP | Deep image prior phase unwrapping | Information 2025 |
| iQSM | Single-step instant QSM | NeuroImage 2022 |
| iQSM+ | Orientation-adaptive single-step QSM | Med Image Anal 2024 |
| MoDIP | Model-based deep image prior QSM | NeuroImage 2024 |
| QSMDiff | Diffusion model-based QSM | MRM 2026 |
| xQSM | QSM dipole inversion | NMR Biomed 2021 |
Each method has its own standalone GitHub repository (linked in the table above) with full setup instructions, model weights, and demo data.