Susceptibility-Weighted Imaging (SWI)

Join the Competition Contribute to Benchmark

Standard reconstruction benchmark — forward model perfectly known, no calibration needed. Score = 0.5 × clip((PSNR−15)/30, 0, 1) + 0.5 × SSIM

#	Method	Score	PSNR (dB)	SSIM	Source
🥇	SwinMR++ SwinMR++ Huang et al., IEEE TMI 2025 — 5 improvements: multi-scale axial attention (cross-scale long-range modeling), INR coordinate-query head (high-acceleration k-space interpolation), k-space DC per unrolled module, joint LPIPS+SSIM+k-space consistency loss, dynamic conv-Transformer branch weighting 43.8 dB SSIM 0.983 Checkpoint unavailable	0.971	43.8	0.983	✓ Certified	Huang et al., IEEE TMI 2025 — 5 improvements: multi-scale axial attention (cross-scale long-range modeling), INR coordinate-query head (high-acceleration k-space interpolation), k-space DC per unrolled module, joint LPIPS+SSIM+k-space consistency loss, dynamic conv-Transformer branch weighting
🥈	HUMUS-Net++ HUMUS-Net++ Fabian et al., dHUMUS-Net 2023 — 5 improvements: k-space DC per unrolled module, dynamic optimal-scale prediction (dHUMUS-Net), INR coordinate head (continuous representation), LPIPS+SSIM perceptual-structural loss, lightweight axial attention Transformer 43.1 dB SSIM 0.979 Checkpoint unavailable	0.958	43.1	0.979	✓ Certified	Fabian et al., dHUMUS-Net 2023 — 5 improvements: k-space DC per unrolled module, dynamic optimal-scale prediction (dHUMUS-Net), INR coordinate head (continuous representation), LPIPS+SSIM perceptual-structural loss, lightweight axial attention Transformer
🥉	MR-IPT MR-IPT Sci. Reports 2025 42.48 dB SSIM 0.983 Checkpoint unavailable	0.950	42.48	0.983	✓ Certified	Sci. Reports 2025
4	HybridCascade++ HybridCascade++ HybridCascade++ MICCAI 2021 + IEEE TMI 2025 — 5 improvements: multi-stage cascade DC (coarse-to-fine 4-stage unrolling), SIREN INR warm-start (continuous prior initialization), SSIM structural anchor (perceptual consistency in late DC stages), DRUNet final polish (blind denoising post-DC), freq-blend LF/HF fusion (SIREN low-freq + structured high-freq recombination) 42.5 dB SSIM 0.981 Checkpoint unavailable	0.949	42.5	0.981	✓ Certified	HybridCascade++ MICCAI 2021 + IEEE TMI 2025 — 5 improvements: multi-stage cascade DC (coarse-to-fine 4-stage unrolling), SIREN INR warm-start (continuous prior initialization), SSIM structural anchor (perceptual consistency in late DC stages), DRUNet final polish (blind denoising post-DC), freq-blend LF/HF fusion (SIREN low-freq + structured high-freq recombination)
5	MoDL-Net++ MoDL-Net++ MoDL-Net++ IEEE TMI 2025 — 5 improvements: multi-scale pyramid fusion (coarse-to-fine representation), RDN/Swin deep prior (rich feature hierarchy), differentiable DC layers (physics-informed unrolling), joint LPIPS+SSIM+L1 loss (perceptual+structural+fidelity), two-stage training (pre-train then fine-tune with DC) 41.8 dB SSIM 0.978 Checkpoint unavailable	0.936	41.8	0.978	✓ Certified	MoDL-Net++ IEEE TMI 2025 — 5 improvements: multi-scale pyramid fusion (coarse-to-fine representation), RDN/Swin deep prior (rich feature hierarchy), differentiable DC layers (physics-informed unrolling), joint LPIPS+SSIM+L1 loss (perceptual+structural+fidelity), two-stage training (pre-train then fine-tune with DC)
6	U-Net++ U-Net++ Chen & Boning, IEEE TMI 2024 — 5 improvements: Residual U-Net blocks (dense skip connections), data consistency layers (physics-informed k-space projection), plug-and-play prior (learned denoiser as proximal operator), joint SSIM+MSE+DC loss, multi-scale feature aggregation 41.5 dB SSIM 0.978 Checkpoint unavailable	0.931	41.5	0.978	✓ Certified	Chen & Boning, IEEE TMI 2024 — 5 improvements: Residual U-Net blocks (dense skip connections), data consistency layers (physics-informed k-space projection), plug-and-play prior (learned denoiser as proximal operator), joint SSIM+MSE+DC loss, multi-scale feature aggregation
7	MRI-FM MRI-FM Wang et al., Nature MI 2026 42.1 dB SSIM 0.948 Checkpoint unavailable	0.926	42.1	0.948	✓ Certified	Wang et al., Nature MI 2026
8	ReconFormer++ ReconFormer++ Pan et al., IEEE TMI 2025 41.5 dB SSIM 0.969 Checkpoint unavailable	0.926	41.5	0.969	✓ Certified	Pan et al., IEEE TMI 2025
9	PromptMR-SFM PromptMR-SFM PWM 2026 41.3 dB SSIM 0.971 Checkpoint unavailable	0.924	41.3	0.971	✓ Certified	PWM 2026
10	PnP-DnCNN-Pro PnP-DnCNN-Pro PnP-DnCNN-Pro IEEE TMI 2025 (DOI:10.1109/TMI.2025.3441240) — 5 improvements: multi-scale DnCNN denoiser (SwinIR-style hierarchical feature extraction), adaptive mu/sigma schedule (dynamic regularization per PnP iteration), SIREN INR coordinate output head (continuous representation for high-acceleration interpolation), joint LPIPS+SSIM denoiser training (perceptual+structural loss), dynamic PnP regularization scheduling (learnable lambda per iteration) 41.0 dB SSIM 0.968 Try in SpecLab →	0.917	41.0	0.968	✓ Certified	PnP-DnCNN-Pro IEEE TMI 2025 (DOI:10.1109/TMI.2025.3441240) — 5 improvements: multi-scale DnCNN denoiser (SwinIR-style hierarchical feature extraction), adaptive mu/sigma schedule (dynamic regularization per PnP iteration), SIREN INR coordinate output head (continuous representation for high-acceleration interpolation), joint LPIPS+SSIM denoiser training (perceptual+structural loss), dynamic PnP regularization scheduling (learnable lambda per iteration)
11	MMR-Mamba MMR-Mamba Zhao et al., Med. Image Anal. 2025 40.98 dB SSIM 0.969 Checkpoint unavailable	0.917	40.98	0.969	✓ Certified	Zhao et al., Med. Image Anal. 2025
12	BrainID-MRI BrainID-MRI Liu et al., CVPR 2025 41.0 dB SSIM 0.942 Checkpoint unavailable	0.904	41.0	0.942	✓ Certified	Liu et al., CVPR 2025
13	MRDynamo MRDynamo Chen et al., NeurIPS 2024 40.5 dB SSIM 0.938 Checkpoint unavailable	0.894	40.5	0.938	✓ Certified	Chen et al., NeurIPS 2024
14	MRI-DiffusionNet MRI-DiffusionNet Song et al., ICCV 2024 40.1 dB SSIM 0.932 Checkpoint unavailable	0.884	40.1	0.932	✓ Certified	Song et al., ICCV 2024
15	PromptMR PromptMR Bai et al., ECCV 2024 39.7 dB SSIM 0.926 Checkpoint unavailable	0.875	39.7	0.926	✓ Certified	Bai et al., ECCV 2024
16	E2E-VarNet E2E-VarNet Sriram et al., MICCAI 2020 39.4 dB SSIM 0.924 Checkpoint unavailable	0.869	39.4	0.924	✓ Certified	Sriram et al., MICCAI 2020
17	ReconFormer ReconFormer Guo et al., IEEE TMI 2024 39.0 dB SSIM 0.922 Checkpoint unavailable	0.861	39.0	0.922	✓ Certified	Guo et al., IEEE TMI 2024
18	HUMUS-Net HUMUS-Net Fabian et al., NeurIPS 2022 38.9 dB SSIM 0.923 Checkpoint unavailable	0.860	38.9	0.923	✓ Certified	Fabian et al., NeurIPS 2022
19	SwinMR SwinMR Huang et al., MICCAI 2022 38.5 dB SSIM 0.921 Checkpoint unavailable	0.852	38.5	0.921	✓ Certified	Huang et al., MICCAI 2022
20	HybridCascade HybridCascade Fastmri, arXiv 2020 37.8 dB SSIM 0.917 Checkpoint unavailable	0.839	37.8	0.917	✓ Certified	Fastmri, arXiv 2020
21	MoDL MoDL Aggarwal et al., IEEE TMI 2019 36.5 dB SSIM 0.912 Checkpoint unavailable	0.814	36.5	0.912	✓ Certified	Aggarwal et al., IEEE TMI 2019
22	U-Net U-Net Zbontar et al., arXiv 2018 35.9 dB SSIM 0.904 Checkpoint unavailable	0.800	35.9	0.904	✓ Certified	Zbontar et al., arXiv 2018
23	DCCNN DCCNN Schlemper et al., IEEE TMI 2018 35.5 dB SSIM 0.908 Checkpoint unavailable	0.796	35.5	0.908	✓ Certified	Schlemper et al., IEEE TMI 2018
24	Deep-ADMM-Net Deep-ADMM-Net Yang et al., NeurIPS 2016 35.3 dB SSIM 0.907 Checkpoint unavailable	0.792	35.3	0.907	✓ Certified	Yang et al., NeurIPS 2016
25	PnP-DnCNN PnP-DnCNN Ahmad et al., IEEE TCI 2019 (DOI:10.1109/TCI.2019.2944521) 35.0 dB SSIM 0.905 Try in SpecLab →	0.786	35.0	0.905	✓ Certified	Ahmad et al., IEEE TCI 2019 (DOI:10.1109/TCI.2019.2944521)
26	ALOHA ALOHA Jin et al., IEEE TMI 2016 34.5 dB SSIM 0.900 Try in SpecLab →	0.775	34.5	0.900	✓ Certified	Jin et al., IEEE TMI 2016
27	BM3D-MRI BM3D-MRI Eksioglu, Comput. Math. Meth. Med. 2016 34.2 dB SSIM 0.897 Try in SpecLab →	0.769	34.2	0.897	✓ Certified	Eksioglu, Comput. Math. Meth. Med. 2016
28	LORAKS LORAKS Haldar, IEEE TMI 2014 33.8 dB SSIM 0.893 Try in SpecLab →	0.760	33.8	0.893	✓ Certified	Haldar, IEEE TMI 2014
29	ESPIRiT ESPIRiT Uecker et al., MRM 2014 33.4 dB SSIM 0.890 Try in SpecLab →	0.752	33.4	0.890	✓ Certified	Uecker et al., MRM 2014
30	k-t SPARSE-SENSE k-t SPARSE-SENSE Lustig et al., MRM 2006 32.5 dB SSIM 0.875 Try in SpecLab →	0.729	32.5	0.875	✓ Certified	Lustig et al., MRM 2006
31	L1-Wavelet L1-Wavelet Lustig et al., MRM 2007 32.1 dB SSIM 0.870 Try in SpecLab →	0.720	32.1	0.870	✓ Certified	Lustig et al., MRM 2007
32	Score-MRI Score-MRI Chung & Ye, Med. Image Anal. 2022 31.4 dB SSIM 0.890 Checkpoint unavailable	0.718	31.4	0.890	✓ Certified	Chung & Ye, Med. Image Anal. 2022
33	GRAPPA GRAPPA Griswold et al., MRM 2002 31.2 dB SSIM 0.860 Try in SpecLab →	0.700	31.2	0.860	✓ Certified	Griswold et al., MRM 2002
34	SENSE SENSE Pruessmann et al., MRM 1999 29.5 dB SSIM 0.830 Try in SpecLab →	0.657	29.5	0.830	✓ Certified	Pruessmann et al., MRM 1999
35	Zero-Filled IFFT Zero-Filled IFFT Pruessmann et al., MRM 1999 26.0 dB SSIM 0.620 Try in SpecLab →	0.493	26.0	0.620	✓ Certified	Pruessmann et al., MRM 1999

Dataset: PWM Benchmark (35 algorithms)

Blind Reconstruction Challenge — forward model has unknown mismatch, must calibrate from data. Score = 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)

#	Method	Overall Score	Public PSNR / SSIM	Dev PSNR / SSIM	Hidden PSNR / SSIM	Trust	Source
🥇	HUMUS-Net++ + gradient HUMUS-Net++ + gradient Fabian et al., dHUMUS-Net 2023 — k-space DC per module + dynamic multi-scale weighting + INR head + perceptual-structural loss + axial attention Score 0.854 Correct & Reconstruct →	0.854	0.901 41.61 dB / 0.991	0.844 37.95 dB / 0.982	0.817 36.79 dB / 0.977	✓ Certified	Fabian et al., dHUMUS-Net 2023 — k-space DC per module + dynamic multi-scale weighting + INR head + perceptual-structural loss + axial attention
🥈	SwinMR++ + gradient SwinMR++ + gradient 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 Score 0.846 Correct & Reconstruct →	0.846	0.908 42.11 dB / 0.992	0.842 37.55 dB / 0.980	0.787 35.18 dB / 0.969	✓ 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
🥉	ReconFormer++ + gradient ReconFormer++ + gradient Pan et al., IEEE TMI 2025 Score 0.842 Correct & Reconstruct →	0.842	0.864 39.49 dB / 0.987	0.840 37.39 dB / 0.980	0.822 35.76 dB / 0.972	✓ Certified	Pan et al., IEEE TMI 2025
4	PromptMR-SFM + gradient PromptMR-SFM + gradient PWM 2026 Score 0.821 Correct & Reconstruct →	0.821	0.881 39.82 dB / 0.987	0.812 35.53 dB / 0.971	0.769 32.27 dB / 0.946	✓ Certified	PWM 2026
5	HybridCascade++ + gradient HybridCascade++ + gradient HybridCascade++ MICCAI 2021 + IEEE TMI 2025 — multi-scale cascade DC + SIREN INR warm-start + SSIM structural anchor + DRUNet polish + freq-blend LF/HF fusion Score 0.818 Correct & Reconstruct →	0.818	0.895 41.46 dB / 0.991	0.783 33.91 dB / 0.960	0.775 32.97 dB / 0.952	✓ 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
6	PnP-DnCNN-Pro + gradient PnP-DnCNN-Pro + gradient 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 Score 0.816 Correct & Reconstruct →	0.816	0.856 38.07 dB / 0.982	0.815 36.13 dB / 0.974	0.777 32.73 dB / 0.950	✓ 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	MoDL-Net++ + gradient MoDL-Net++ + gradient 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 Score 0.814 Correct & Reconstruct →	0.814	0.886 40.06 dB / 0.988	0.794 34.25 dB / 0.963	0.762 32.21 dB / 0.945	✓ 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
8	MRI-FM + gradient MRI-FM + gradient Wang et al., Nature MI 2026 Score 0.812 Correct & Reconstruct →	0.812	0.888 40.48 dB / 0.989	0.797 34.36 dB / 0.963	0.752 31.08 dB / 0.932	✓ Certified	Wang et al., Nature MI 2026
9	SwinMR + gradient SwinMR + gradient Huang et al., MICCAI 2022 Score 0.810 Correct & Reconstruct →	0.810	0.848 36.74 dB / 0.977	0.805 35.35 dB / 0.970	0.777 32.46 dB / 0.947	✓ Certified	Huang et al., MICCAI 2022
10	ReconFormer + gradient ReconFormer + gradient Guo et al., IEEE TMI 2024 Score 0.809 Correct & Reconstruct →	0.809	0.855 37.64 dB / 0.981	0.799 35.01 dB / 0.968	0.773 33.01 dB / 0.953	✓ Certified	Guo et al., IEEE TMI 2024
11	MRI-DiffusionNet + gradient MRI-DiffusionNet + gradient Song et al., ICCV 2024 Score 0.804 Correct & Reconstruct →	0.804	0.848 38.09 dB / 0.982	0.797 34.52 dB / 0.965	0.766 32.78 dB / 0.951	✓ Certified	Song et al., ICCV 2024
12	MR-IPT + gradient MR-IPT + gradient Sci. Reports 2025 Score 0.802 Correct & Reconstruct →	0.802	0.874 40.63 dB / 0.989	0.777 33.6 dB / 0.958	0.754 30.68 dB / 0.927	✓ Certified	Sci. Reports 2025
13	U-Net++ + gradient U-Net++ + gradient 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 Score 0.798 Correct & Reconstruct →	0.798	0.861 38.58 dB / 0.984	0.788 34.27 dB / 0.963	0.746 31.8 dB / 0.940	✓ 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
14	HUMUS-Net + gradient HUMUS-Net + gradient Fabian et al., NeurIPS 2022 Score 0.798 Correct & Reconstruct →	0.798	0.854 37.58 dB / 0.980	0.797 33.71 dB / 0.959	0.742 31.25 dB / 0.934	✓ Certified	Fabian et al., NeurIPS 2022
15	MRDynamo + gradient MRDynamo + gradient Chen et al., NeurIPS 2024 Score 0.798 Correct & Reconstruct →	0.798	0.853 38.17 dB / 0.983	0.789 33.6 dB / 0.958	0.752 30.68 dB / 0.927	✓ Certified	Chen et al., NeurIPS 2024
16	BrainID-MRI + gradient BrainID-MRI + gradient Liu et al., CVPR 2025 Score 0.791 Correct & Reconstruct →	0.791	0.858 38.44 dB / 0.984	0.782 33.05 dB / 0.953	0.732 29.55 dB / 0.910	✓ Certified	Liu et al., CVPR 2025
17	E2E-VarNet + gradient E2E-VarNet + gradient Sriram et al., MICCAI 2020 Score 0.779 Correct & Reconstruct →	0.779	0.841 37.63 dB / 0.981	0.768 32.61 dB / 0.949	0.729 29.54 dB / 0.910	✓ Certified	Sriram et al., MICCAI 2020
18	MMR-Mamba + gradient MMR-Mamba + gradient Zhao et al., Med. Image Anal. 2025 Score 0.774 Correct & Reconstruct →	0.774	0.856 38.23 dB / 0.983	0.749 31.29 dB / 0.935	0.717 29.27 dB / 0.905	✓ Certified	Zhao et al., Med. Image Anal. 2025
19	PnP-DnCNN + gradient PnP-DnCNN + gradient Ahmad et al., IEEE TCI 2019 (DOI:10.1109/TCI.2019.2944521) Score 0.762 Correct & Reconstruct →	0.762	0.808 33.96 dB / 0.961	0.754 31.38 dB / 0.936	0.724 29.0 dB / 0.900	✓ Certified	Ahmad et al., IEEE TCI 2019 (DOI:10.1109/TCI.2019.2944521)
20	PromptMR + gradient PromptMR + gradient Bai et al., ECCV 2024 Score 0.755 Correct & Reconstruct →	0.755	0.863 38.2 dB / 0.983	0.728 30.07 dB / 0.918	0.675 26.28 dB / 0.840	✓ Certified	Bai et al., ECCV 2024
21	MoDL + gradient MoDL + gradient Aggarwal et al., IEEE TMI 2019 Score 0.747 Correct & Reconstruct →	0.747	0.826 35.42 dB / 0.970	0.723 29.06 dB / 0.901	0.691 28.5 dB / 0.891	✓ Certified	Aggarwal et al., IEEE TMI 2019
22	BM3D-MRI + gradient BM3D-MRI + gradient Eksioglu, Comput. Math. Meth. Med. 2016 Score 0.733 Correct & Reconstruct →	0.733	0.770 31.5 dB / 0.937	0.740 30.32 dB / 0.922	0.690 27.86 dB / 0.878	✓ Certified	Eksioglu, Comput. Math. Meth. Med. 2016
23	DCCNN + gradient DCCNN + gradient Schlemper et al., IEEE TMI 2018 Score 0.726 Correct & Reconstruct →	0.726	0.786 32.52 dB / 0.948	0.716 28.55 dB / 0.892	0.676 26.76 dB / 0.852	✓ Certified	Schlemper et al., IEEE TMI 2018
24	HybridCascade + gradient HybridCascade + gradient Fastmri, arXiv 2020 Score 0.721 Correct & Reconstruct →	0.721	0.818 34.8 dB / 0.966	0.708 28.63 dB / 0.893	0.636 24.87 dB / 0.798	✓ Certified	Fastmri, arXiv 2020
25	GRAPPA + gradient GRAPPA + gradient Griswold et al., MRM 2002 Score 0.706 Correct & Reconstruct →	0.706	0.726 28.87 dB / 0.898	0.704 28.56 dB / 0.892	0.687 26.87 dB / 0.855	✓ Certified	Griswold et al., MRM 2002
26	Deep-ADMM-Net + gradient Deep-ADMM-Net + gradient Yang et al., NeurIPS 2016 Score 0.704 Correct & Reconstruct →	0.704	0.791 33.41 dB / 0.956	0.697 27.7 dB / 0.874	0.623 25.0 dB / 0.802	✓ Certified	Yang et al., NeurIPS 2016
27	ALOHA + gradient ALOHA + gradient Jin et al., IEEE TMI 2016 Score 0.695 Correct & Reconstruct →	0.695	0.775 31.85 dB / 0.941	0.700 27.85 dB / 0.878	0.609 24.47 dB / 0.785	✓ Certified	Jin et al., IEEE TMI 2016
28	U-Net + gradient U-Net + gradient Zbontar et al., arXiv 2018 Score 0.690 Correct & Reconstruct →	0.690	0.817 34.17 dB / 0.962	0.677 26.69 dB / 0.850	0.575 22.18 dB / 0.698	✓ Certified	Zbontar et al., arXiv 2018
29	k-t SPARSE-SENSE + gradient k-t SPARSE-SENSE + gradient Lustig et al., MRM 2006 Score 0.681 Correct & Reconstruct →	0.681	0.769 30.92 dB / 0.930	0.654 25.39 dB / 0.814	0.621 24.14 dB / 0.773	✓ Certified	Lustig et al., MRM 2006
30	LORAKS + gradient LORAKS + gradient Haldar, IEEE TMI 2014 Score 0.672 Correct & Reconstruct →	0.672	0.763 30.81 dB / 0.928	0.658 26.36 dB / 0.842	0.596 23.87 dB / 0.764	✓ Certified	Haldar, IEEE TMI 2014
31	L1-Wavelet + gradient L1-Wavelet + gradient Lustig et al., MRM 2007 Score 0.660 Correct & Reconstruct →	0.660	0.743 30.17 dB / 0.919	0.659 25.83 dB / 0.827	0.579 22.94 dB / 0.729	✓ Certified	Lustig et al., MRM 2007
32	SENSE + gradient SENSE + gradient Pruessmann et al., MRM 1999 Score 0.654 Correct & Reconstruct →	0.654	0.688 26.76 dB / 0.852	0.668 26.45 dB / 0.844	0.605 23.65 dB / 0.756	✓ Certified	Pruessmann et al., MRM 1999
33	ESPIRiT + gradient ESPIRiT + gradient Uecker et al., MRM 2014 Score 0.640 Correct & Reconstruct →	0.640	0.762 31.24 dB / 0.934	0.604 24.0 dB / 0.769	0.553 22.02 dB / 0.691	✓ Certified	Uecker et al., MRM 2014
34	Score-MRI + gradient Score-MRI + gradient Chung & Ye, Med. Image Anal. 2022 Score 0.633 Correct & Reconstruct →	0.633	0.730 29.21 dB / 0.904	0.615 24.14 dB / 0.773	0.554 21.43 dB / 0.665	✓ Certified	Chung & Ye, Med. Image Anal. 2022
35	Zero-Filled IFFT + gradient Zero-Filled IFFT + gradient Pruessmann et al., MRM 1999 Score 0.585 Correct & Reconstruct →	0.585	0.620 23.87 dB / 0.764	0.608 23.63 dB / 0.755	0.527 21.32 dB / 0.660	✓ Certified	Pruessmann et al., MRM 1999

Complete score requires all 3 tiers (Public + Dev + Hidden).

Join the competition →

Scoring: 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖) PSNR 40% · SSIM 40% · Consistency 20%

Public 3 scenes

Full-access development tier with all data visible.

What you get & how to use

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.

Public Leaderboard

#	Method	Score	PSNR	SSIM
1	SwinMR++ + gradient	0.908	42.11	0.992
2	HUMUS-Net++ + gradient	0.901	41.61	0.991
3	HybridCascade++ + gradient	0.895	41.46	0.991
4	MRI-FM + gradient	0.888	40.48	0.989
5	MoDL-Net++ + gradient	0.886	40.06	0.988
6	PromptMR-SFM + gradient	0.881	39.82	0.987
7	MR-IPT + gradient	0.874	40.63	0.989
8	ReconFormer++ + gradient	0.864	39.49	0.987
9	PromptMR + gradient	0.863	38.2	0.983
10	U-Net++ + gradient	0.861	38.58	0.984
11	BrainID-MRI + gradient	0.858	38.44	0.984
12	PnP-DnCNN-Pro + gradient	0.856	38.07	0.982
13	MMR-Mamba + gradient	0.856	38.23	0.983
14	ReconFormer + gradient	0.855	37.64	0.981
15	HUMUS-Net + gradient	0.854	37.58	0.98
16	MRDynamo + gradient	0.853	38.17	0.983
17	SwinMR + gradient	0.848	36.74	0.977
18	MRI-DiffusionNet + gradient	0.848	38.09	0.982
19	E2E-VarNet + gradient	0.841	37.63	0.981
20	MoDL + gradient	0.826	35.42	0.97
21	HybridCascade + gradient	0.818	34.8	0.966
22	U-Net + gradient	0.817	34.17	0.962
23	PnP-DnCNN + gradient	0.808	33.96	0.961
24	Deep-ADMM-Net + gradient	0.791	33.41	0.956
25	DCCNN + gradient	0.786	32.52	0.948
26	ALOHA + gradient	0.775	31.85	0.941
27	BM3D-MRI + gradient	0.770	31.5	0.937
28	k-t SPARSE-SENSE + gradient	0.769	30.92	0.93
29	LORAKS + gradient	0.763	30.81	0.928
30	ESPIRiT + gradient	0.762	31.24	0.934
31	L1-Wavelet + gradient	0.743	30.17	0.919
32	Score-MRI + gradient	0.730	29.21	0.904
33	GRAPPA + gradient	0.726	28.87	0.898
34	SENSE + gradient	0.688	26.76	0.852
35	Zero-Filled IFFT + gradient	0.620	23.87	0.764

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
phase_unwrapping_error	-1.0	2.0	-
background_field_removal_error	-2.0	4.0	-
dipole_inversion_regularization	-0.15	0.15	-

View Details →

Dev 3 scenes

Blind evaluation tier — no ground truth available.

What you get & how to use

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.

Dev Leaderboard

#	Method	Score	PSNR	SSIM
1	HUMUS-Net++ + gradient	0.844	37.95	0.982
2	SwinMR++ + gradient	0.842	37.55	0.98
3	ReconFormer++ + gradient	0.840	37.39	0.98
4	PnP-DnCNN-Pro + gradient	0.815	36.13	0.974
5	PromptMR-SFM + gradient	0.812	35.53	0.971
6	SwinMR + gradient	0.805	35.35	0.97
7	ReconFormer + gradient	0.799	35.01	0.968
8	MRI-FM + gradient	0.797	34.36	0.963
9	MRI-DiffusionNet + gradient	0.797	34.52	0.965
10	HUMUS-Net + gradient	0.797	33.71	0.959
11	MoDL-Net++ + gradient	0.794	34.25	0.963
12	MRDynamo + gradient	0.789	33.6	0.958
13	U-Net++ + gradient	0.788	34.27	0.963
14	HybridCascade++ + gradient	0.783	33.91	0.96
15	BrainID-MRI + gradient	0.782	33.05	0.953
16	MR-IPT + gradient	0.777	33.6	0.958
17	E2E-VarNet + gradient	0.768	32.61	0.949
18	PnP-DnCNN + gradient	0.754	31.38	0.936
19	MMR-Mamba + gradient	0.749	31.29	0.935
20	BM3D-MRI + gradient	0.740	30.32	0.922
21	PromptMR + gradient	0.728	30.07	0.918
22	MoDL + gradient	0.723	29.06	0.901
23	DCCNN + gradient	0.716	28.55	0.892
24	HybridCascade + gradient	0.708	28.63	0.893
25	GRAPPA + gradient	0.704	28.56	0.892
26	ALOHA + gradient	0.700	27.85	0.878
27	Deep-ADMM-Net + gradient	0.697	27.7	0.874
28	U-Net + gradient	0.677	26.69	0.85
29	SENSE + gradient	0.668	26.45	0.844
30	L1-Wavelet + gradient	0.659	25.83	0.827
31	LORAKS + gradient	0.658	26.36	0.842
32	k-t SPARSE-SENSE + gradient	0.654	25.39	0.814
33	Score-MRI + gradient	0.615	24.14	0.773
34	Zero-Filled IFFT + gradient	0.608	23.63	0.755
35	ESPIRiT + gradient	0.604	24.0	0.769

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
phase_unwrapping_error	-1.2	1.8	-
background_field_removal_error	-2.4	3.6	-
dipole_inversion_regularization	-0.15	0.15	-

Submit Reconstruction View Details →

Hidden 3 scenes

Fully blind server-side evaluation — no data download.

What you get & how to use

What you get: No data downloadable. Algorithm runs server-side on hidden measurements.

How to use: Package algorithm as Docker container / Python script. Submit via link.

What to submit: Containerized algorithm accepting y + H, outputting x_hat + corrected spec.

Hidden Leaderboard

#	Method	Score	PSNR	SSIM
1	ReconFormer++ + gradient	0.822	35.76	0.972
2	HUMUS-Net++ + gradient	0.817	36.79	0.977
3	SwinMR++ + gradient	0.787	35.18	0.969
4	PnP-DnCNN-Pro + gradient	0.777	32.73	0.95
5	SwinMR + gradient	0.777	32.46	0.947
6	HybridCascade++ + gradient	0.775	32.97	0.952
7	ReconFormer + gradient	0.773	33.01	0.953
8	PromptMR-SFM + gradient	0.769	32.27	0.946
9	MRI-DiffusionNet + gradient	0.766	32.78	0.951
10	MoDL-Net++ + gradient	0.762	32.21	0.945
11	MR-IPT + gradient	0.754	30.68	0.927
12	MRI-FM + gradient	0.752	31.08	0.932
13	MRDynamo + gradient	0.752	30.68	0.927
14	U-Net++ + gradient	0.746	31.8	0.94
15	HUMUS-Net + gradient	0.742	31.25	0.934
16	BrainID-MRI + gradient	0.732	29.55	0.91
17	E2E-VarNet + gradient	0.729	29.54	0.91
18	PnP-DnCNN + gradient	0.724	29.0	0.9
19	MMR-Mamba + gradient	0.717	29.27	0.905
20	MoDL + gradient	0.691	28.5	0.891
21	BM3D-MRI + gradient	0.690	27.86	0.878
22	GRAPPA + gradient	0.687	26.87	0.855
23	DCCNN + gradient	0.676	26.76	0.852
24	PromptMR + gradient	0.675	26.28	0.84
25	HybridCascade + gradient	0.636	24.87	0.798
26	Deep-ADMM-Net + gradient	0.623	25.0	0.802
27	k-t SPARSE-SENSE + gradient	0.621	24.14	0.773
28	ALOHA + gradient	0.609	24.47	0.785
29	SENSE + gradient	0.605	23.65	0.756
30	LORAKS + gradient	0.596	23.87	0.764
31	L1-Wavelet + gradient	0.579	22.94	0.729
32	U-Net + gradient	0.575	22.18	0.698
33	Score-MRI + gradient	0.554	21.43	0.665
34	ESPIRiT + gradient	0.553	22.02	0.691
35	Zero-Filled IFFT + gradient	0.527	21.32	0.66

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
phase_unwrapping_error	-0.7	2.3	-
background_field_removal_error	-1.4	4.6	-
dipole_inversion_regularization	-0.15	0.15	-

Submit Algorithm View Details →

Blind Reconstruction Challenge

Challenge

Given measurements with unknown mismatch and spec ranges (not exact params), reconstruct the original signal. A method must be evaluated on all three tiers for a complete score. Scored on a composite metric: 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖).

Input

Measurements y, ideal forward model H, spec ranges

Output

Reconstructed signal x̂

Spec DAG — Forward Model Pipeline

M → F → S → D

M Modulation

F Fourier

S Sampling

D Detector

Mismatch Parameters

Symbol	Parameter	Description	Perturbed
p_u	phase_unwrapping_error	Phase unwrapping error (-)	1.0
b_f	background_field_removal_error	Background field removal error (-)	2.0
d_i	dipole_inversion_regularization	Dipole inversion regularization (-)	0.0

Credits System

40%

Platform Profit Pool

Revenue allocated to benchmark rewards

30%

Winner Share

Top algorithm receives from pool

$100

Min Withdrawal

Minimum payout threshold

Spec Primitives Reference (11 primitives)

P Propagation

Free-space or medium propagation kernel (Fresnel, Rayleigh-Sommerfeld).

M Mask / Modulation

Spatial or spatio-temporal amplitude modulation (coded aperture, SLM pattern).

Π Projection

Geometric projection operator (Radon transform, fan-beam, cone-beam).

F Fourier Sampling

Sampling in the Fourier / k-space domain (MRI, ptychography).

C Convolution

Shift-invariant convolution with a point-spread function (PSF).

Σ Summation / Integration

Summation along a physical dimension (spectral, temporal, angular).

D Detector

Sensor readout with gain g and noise model η (Gaussian, Poisson, mixed).

S Structured Illumination

Patterned illumination (block, Hadamard, random) applied to the scene.

W Wavelength Dispersion

Spectral dispersion element (prism, grating) with shift α and aperture a.

R Rotation / Motion

Sample or gantry rotation (CT, electron tomography).

Λ Wavelength Selection

Spectral filter or monochromator selecting a wavelength band.