Dev
MRI — Dev Tier
(3 scenes)Blind evaluation tier — no ground truth available.
What you get
Measurements (y), ideal forward operator (H), and spec ranges only.
How to use
Apply your pipeline from the Public tier. Use consistency as self-check.
What to submit
Reconstructed signals and corrected spec. Scored server-side.
Parameter Specifications
🔒
True spec hidden — estimate parameters from spec ranges below.
| Parameter | Spec Range | Unit |
|---|---|---|
| B0_inhomog | -1.8 – 2.7 | ppm |
| gradient_nonlin | -2.4 – 3.6 | % |
| coil_sensitivity | -6.0 – 9.0 | % |
| k_trajectory | -1.2 – 1.8 | % |
Dev Tier Leaderboard
| # | Method | Score | PSNR | SSIM | Consistency | Trust | Source |
|---|---|---|---|---|---|---|---|
| 1 | SwinMR++ + gradient | 0.851 | 38.9 | 0.985 | 0.84 | ✓ Certified | Huang et al., IEEE TMI 2025 — multi-scale axial attention + INR head + k-space DC per module + LPIPS+SSIM+k-space joint loss + dynamic feature fusion |
| 2 | HUMUS-Net++ + gradient | 0.823 | 37.15 | 0.979 | 0.8 | ✓ Certified | Fabian et al., dHUMUS-Net 2023 — k-space DC per module + dynamic multi-scale weighting + INR head + perceptual-structural loss + axial attention |
| 3 | HUMUS-Net + gradient | 0.815 | 34.82 | 0.967 | 0.9 | ✓ Certified | Fabian et al., NeurIPS 2022 |
| 4 | MRI-FM + gradient | 0.809 | 35.17 | 0.969 | 0.85 | ✓ Certified | Wang et al., Nature MI 2026 |
| 5 | PnP-DnCNN-Pro + gradient | 0.808 | 34.72 | 0.966 | 0.87 | ✓ Certified | PnP-DnCNN-Pro IEEE TMI 2025 (DOI:10.1109/TMI.2025.3441240) — multi-scale DnCNN denoiser + adaptive mu/sigma schedule + SIREN INR output head + joint LPIPS+SSIM denoiser training + dynamic PnP regularization scheduling |
| 6 | MR-IPT + gradient | 0.803 | 35.34 | 0.97 | 0.81 | ✓ Certified | Sci. Reports 2025 |
| 7 | ReconFormer++ + gradient | 0.802 | 34.15 | 0.962 | 0.88 | ✓ Certified | Pan et al., IEEE TMI 2025 |
| 8 | HybridCascade++ + gradient | 0.801 | 35.32 | 0.97 | 0.8 | ✓ Certified | HybridCascade++ MICCAI 2021 + IEEE TMI 2025 — multi-scale cascade DC + SIREN INR warm-start + SSIM structural anchor + DRUNet polish + freq-blend LF/HF fusion |
| 9 | U-Net++ + gradient | 0.800 | 34.79 | 0.966 | 0.83 | ✓ Certified | Chen & Boning, IEEE TMI 2024 (DOI: 10.1109/TMI.2024.3367890) — Residual U-Net + data consistency layers + plug-and-play prior + residual connections + dense skip paths |
| 10 | SwinMR + gradient | 0.794 | 33.64 | 0.958 | 0.87 | ✓ Certified | Huang et al., MICCAI 2022 |
| 11 | ReconFormer + gradient | 0.792 | 33.95 | 0.96 | 0.84 | ✓ Certified | Guo et al., IEEE TMI 2024 |
| 12 | MMR-Mamba + gradient | 0.784 | 33.24 | 0.955 | 0.85 | ✓ Certified | Zhao et al., Med. Image Anal. 2025 |
| 13 | E2E-VarNet + gradient | 0.774 | 32.92 | 0.952 | 0.82 | ✓ Certified | Sriram et al., MICCAI 2020 |
| 14 | MRDynamo + gradient | 0.771 | 32.46 | 0.947 | 0.84 | ✓ Certified | Chen et al., NeurIPS 2024 |
| 15 | PromptMR-SFM + gradient | 0.768 | 32.49 | 0.948 | 0.82 | ✓ Certified | PWM 2026 |
| 16 | MoDL-Net++ + gradient | 0.757 | 32.25 | 0.945 | 0.78 | ✓ Certified | MoDL-Net++ IEEE TMI 2025 — multi-scale pyramid fusion + RDN/Swin deep prior + differentiable DC layers + LPIPS+SSIM+L1 joint loss + two-stage training strategy |
| 17 | U-Net + gradient | 0.754 | 31.08 | 0.932 | 0.85 | ✓ Certified | Zbontar et al., arXiv 2018 |
| 18 | PnP-DnCNN + gradient | 0.749 | 31.43 | 0.936 | 0.8 | ✓ Certified | Ahmad et al., IEEE TCI 2019 (DOI:10.1109/TCI.2019.2944521) |
| 19 | PromptMR + gradient | 0.746 | 30.88 | 0.929 | 0.83 | ✓ Certified | Bai et al., ECCV 2024 |
| 20 | MRI-DiffusionNet + gradient | 0.744 | 29.95 | 0.916 | 0.89 | ✓ Certified | Song et al., ICCV 2024 |
| 21 | BrainID-MRI + gradient | 0.737 | 30.81 | 0.928 | 0.79 | ✓ Certified | Liu et al., CVPR 2025 |
| 22 | MoDL + gradient | 0.723 | 28.71 | 0.895 | 0.89 | ✓ Certified | Aggarwal et al., IEEE TMI 2019 |
| 23 | DCCNN + gradient | 0.719 | 28.94 | 0.899 | 0.85 | ✓ Certified | Schlemper et al., IEEE TMI 2018 |
| 24 | GRAPPA + gradient | 0.713 | 28.8 | 0.897 | 0.83 | ✓ Certified | Griswold et al., MRM 2002 |
| 25 | Deep-ADMM-Net + gradient | 0.703 | 28.16 | 0.884 | 0.84 | ✓ Certified | Yang et al., NeurIPS 2016 |
| 26 | HybridCascade + gradient | 0.698 | 27.68 | 0.874 | 0.86 | ✓ Certified | Fastmri, arXiv 2020 |
| 27 | SENSE + gradient | 0.681 | 26.58 | 0.848 | 0.88 | ✓ Certified | Pruessmann et al., MRM 1999 |
| 28 | BM3D-MRI + gradient | 0.672 | 26.34 | 0.841 | 0.86 | ✓ Certified | Eksioglu, Comput. Math. Meth. Med. 2016 |
| 29 | ESPIRiT + gradient | 0.657 | 25.44 | 0.816 | 0.88 | ✓ Certified | Uecker et al., MRM 2014 |
| 30 | ALOHA + gradient | 0.656 | 25.85 | 0.828 | 0.83 | ✓ Certified | Jin et al., IEEE TMI 2016 |
| 31 | Score-MRI + gradient | 0.615 | 24.41 | 0.783 | 0.79 | ✓ Certified | Chung & Ye, Med. Image Anal. 2022 |
| 32 | LORAKS + gradient | 0.612 | 23.69 | 0.757 | 0.86 | ✓ Certified | Haldar, IEEE TMI 2014 |
| 33 | L1-Wavelet + gradient | 0.586 | 23.14 | 0.737 | 0.8 | ✓ Certified | Lustig et al., MRM 2007 |
| 34 | k-t SPARSE-SENSE + gradient | 0.582 | 22.31 | 0.703 | 0.89 | ✓ Certified | Lustig et al., MRM 2006 |
| 35 | Zero-Filled IFFT + gradient | 0.574 | 22.29 | 0.702 | 0.85 | ✓ Certified | Pruessmann et al., MRM 1999 |
Visible Data Fields
y
H_ideal
spec_ranges
Dataset
Format: HDF5
Scenes: 3
Scoring Formula
0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)
PSNR: 40%
SSIM: 40%
Consistency: 20%