Public

CEST MRI — Public Tier

(3 scenes)

Full-access development tier with all data visible.

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

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

Reconstructed signals (x_hat) and corrected spec as HDF5.

✓

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

Parameter	Spec Range	True Value	Unit
b0_inhomogeneity	-10.0 – 20.0	5.0	Hz
b1_inhomogeneity	-4.0 – 8.0	2.0	-
saturation_power_error	-2.0 – 4.0	1.0	-
mt_contamination	-6.0 – 12.0	3.0	-

Method: CPU_baseline — Mismatch parameter: nominal

17.61 dB

SSIM 0.3262

13.39 dB

SSIM 0.0553

16.66 dB

SSIM 0.1359

Per-scene breakdown (4 scenes)

Scene	PSNR I	SSIM I	PSNR II	SSIM II	PSNR III	SSIM III
scene_00	16.58	0.3251	12.85	0.0515	16.16	0.1317
scene_01	18.88	0.3252	14.12	0.0562	17.37	0.1299
scene_02	18.30	0.3298	13.74	0.0603	16.95	0.1425
scene_03	16.68	0.3248	12.85	0.0534	16.15	0.1394

#	Method	Score	PSNR	SSIM	Consistency	Trust	Source
1	DiffusionCEST + gradient	0.843	37.67	0.981	0.87	✓ Certified	Chen et al., NeurIPS 2024
2	PromptCEST + gradient	0.830	36.35	0.975	0.88	✓ Certified	Liu et al., MRM 2024
3	CESTFormer + gradient	0.817	35.62	0.971	0.86	✓ Certified	Wu et al., IEEE TMI 2023
4	PINN-CEST + gradient	0.793	33.19	0.954	0.9	✓ Certified	Cohen et al., MRM 2022
5	U-Net-CEST + gradient	0.783	32.82	0.951	0.87	✓ Certified	Zhao et al., MRM 2021
6	DnCNN-CEST + gradient	0.739	29.62	0.911	0.89	✓ Certified	Zhang et al., IEEE TIP 2017 (CEST adapted)
7	WASSR + gradient	0.668	25.68	0.823	0.91	✓ Certified	Kim et al., MRM 2009
8	Lorentzian-Fit + gradient	0.652	25.43	0.815	0.86	✓ Certified	Zaiss & Bachert, NMR Biomed. 2013
9	MTR-asym + gradient	0.590	22.67	0.718	0.88	✓ Certified	Zhou et al., Nat. Med. 2003

y H_ideal spec_ranges x_true true_spec

Format: HDF5

Scenes: 3

0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)

PSNR: 40% SSIM: 40% Consistency: 20%