Public

MRS — Public Tier

(3 scenes)

Full-access development tier with all data visible.

What you get

Measurements (y), ideal forward operator (H), spec ranges, ground truth (x_true), and true mismatch spec.

How to use

Load HDF5 → compare reconstruction vs x_true → check consistency → iterate.

What to submit

Reconstructed signals (x_hat) and corrected spec as HDF5.

Parameter Specifications

✓

True spec visible — use these exact values for Scenario III oracle reconstruction.

Parameter Spec Range True Value Unit
linewidth -2.0 – 4.0 1.0 Hz
freq_drift -1.5 – 3.0 0.75 Hz
phase_error -5.0 – 10.0 2.5 deg
baseline -0.05 – 0.1 0.025

InverseNet Baseline Scores

Method: CPU_baseline — Mismatch parameter: nominal

Scenario I (Ideal)

17.61 dB

SSIM 0.3262

Scenario II (Mismatch)

13.04 dB

SSIM 0.0468

Scenario III (Oracle)

16.34 dB

SSIM 0.1224

Per-scene breakdown (4 scenes)
Scene PSNR I SSIM I PSNR II SSIM II PSNR III SSIM III
scene_00 16.58 0.3251 12.51 0.0447 15.93 0.1214
scene_01 18.88 0.3252 13.74 0.0480 17.02 0.1162
scene_02 18.30 0.3298 13.37 0.0496 16.55 0.1259
scene_03 16.68 0.3248 12.53 0.0450 15.85 0.1262

Public Tier Leaderboard

# Method Score PSNR SSIM Consistency Trust Source
1 SwinMR++ + gradient 0.908 42.34 0.992 0.94 ✓ 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 0.900 41.55 0.991 0.94 ✓ Certified Fabian et al., dHUMUS-Net 2023 — k-space DC per module + dynamic multi-scale weighting + INR head + perceptual-structural loss + axial attention
3 HybridCascade++ + gradient 0.894 41.03 0.99 0.94 ✓ 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
4 MR-IPT + gradient 0.894 41.16 0.99 0.93 ✓ Certified Sci. Reports 2025
5 U-Net++ + gradient 0.885 40.49 0.989 0.92 ✓ 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
6 PnP-DnCNN-Pro + gradient 0.877 39.62 0.987 0.93 ✓ 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
7 MRI-FM + gradient 0.871 40.31 0.989 0.86 ✓ Certified Wang et al., Nature MI 2026
8 MoDL-Net++ + gradient 0.868 39.48 0.987 0.89 ✓ 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
9 PromptMR-SFM + gradient 0.863 39.41 0.986 0.87 ✓ Certified PWM 2026
10 ReconFormer++ + gradient 0.862 39.19 0.986 0.88 ✓ Certified Pan et al., IEEE TMI 2025
11 PromptMR + gradient 0.862 38.31 0.983 0.93 ✓ Certified Bai et al., ECCV 2024
12 BrainID-MRI + gradient 0.858 39.2 0.986 0.86 ✓ Certified Liu et al., CVPR 2025
13 MMR-Mamba + gradient 0.858 39.15 0.986 0.86 ✓ Certified Zhao et al., Med. Image Anal. 2025
14 ReconFormer + gradient 0.855 37.81 0.981 0.92 ✓ Certified Guo et al., IEEE TMI 2024
15 MRDynamo + gradient 0.853 38.57 0.984 0.87 ✓ Certified Chen et al., NeurIPS 2024
16 SwinMR + gradient 0.849 36.95 0.978 0.94 ✓ Certified Huang et al., MICCAI 2022
17 MRI-DiffusionNet + gradient 0.848 37.76 0.981 0.89 ✓ Certified Song et al., ICCV 2024
18 E2E-VarNet + gradient 0.837 36.65 0.977 0.9 ✓ Certified Sriram et al., MICCAI 2020
19 HUMUS-Net + gradient 0.834 36.72 0.977 0.88 ✓ Certified Fabian et al., NeurIPS 2022
20 MoDL + gradient 0.827 35.47 0.971 0.92 ✓ Certified Aggarwal et al., IEEE TMI 2019
21 HybridCascade + gradient 0.821 35.81 0.972 0.87 ✓ Certified Fastmri, arXiv 2020
22 U-Net + gradient 0.818 34.8 0.966 0.92 ✓ Certified Zbontar et al., arXiv 2018
23 DCCNN + gradient 0.811 33.78 0.959 0.95 ✓ Certified Schlemper et al., IEEE TMI 2018
24 Deep-ADMM-Net + gradient 0.808 33.66 0.958 0.94 ✓ Certified Yang et al., NeurIPS 2016
25 PnP-DnCNN + gradient 0.805 33.46 0.957 0.94 ✓ Certified Ahmad et al., IEEE TCI 2019 (DOI:10.1109/TCI.2019.2944521)
26 LORAKS + gradient 0.789 32.27 0.946 0.94 ✓ Certified Haldar, IEEE TMI 2014
27 ESPIRiT + gradient 0.787 32.37 0.947 0.92 ✓ Certified Uecker et al., MRM 2014
28 ALOHA + gradient 0.778 32.52 0.948 0.87 ✓ Certified Jin et al., IEEE TMI 2016
29 BM3D-MRI + gradient 0.771 31.62 0.938 0.9 ✓ Certified Eksioglu, Comput. Math. Meth. Med. 2016
30 k-t SPARSE-SENSE + gradient 0.746 30.2 0.92 0.88 ✓ Certified Lustig et al., MRM 2006
31 L1-Wavelet + gradient 0.740 29.67 0.912 0.89 ✓ Certified Lustig et al., MRM 2007
32 GRAPPA + gradient 0.729 29.26 0.905 0.87 ✓ Certified Griswold et al., MRM 2002
33 Score-MRI + gradient 0.726 28.64 0.894 0.91 ✓ Certified Chung & Ye, Med. Image Anal. 2022
34 SENSE + gradient 0.718 27.79 0.876 0.95 ✓ Certified Pruessmann et al., MRM 1999
35 Zero-Filled IFFT + gradient 0.621 23.97 0.767 0.87 ✓ Certified Pruessmann et al., MRM 1999

Visible Data Fields

y H_ideal spec_ranges x_true true_spec

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%
Back to MRS