Public

PET — 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
attenuation	-5.0 – 10.0	2.5	%
scatter_frac	0.25 – 0.4	0.325
timing_res	150.0 – 300.0	225.0	ps
normalization	-2.0 – 4.0	1.0	%

Method: CPU_baseline — Mismatch parameter: nominal

32.96 dB

SSIM 0.6466

14.74 dB

SSIM 0.8553

23.48 dB

SSIM 0.2103

Per-scene breakdown (4 scenes)

Scene	PSNR I	SSIM I	PSNR II	SSIM II	PSNR III	SSIM III
scene_00	31.50	0.5193	12.56	0.8273	22.10	0.0697
scene_01	37.06	0.9129	16.59	0.8892	31.16	0.6125
scene_02	33.43	0.7073	17.15	0.8864	21.17	0.1166
scene_03	29.86	0.4469	12.67	0.8182	19.48	0.0426

#	Method	Score	PSNR	SSIM	Consistency	Trust	Source
1	PETFormer + gradient	0.816	34.57	0.965	0.92	✓ Certified	Li et al., ECCV 2024
2	U-Net-PET + gradient	0.808	34.45	0.964	0.89	✓ Certified	Ronneberger et al. variant, MICCAI 2020
3	PET-ViT + gradient	0.802	34.16	0.962	0.88	✓ Certified	Smith et al., ICCV 2024
4	TransEM + gradient	0.762	30.99	0.931	0.9	✓ Certified	Xie et al., 2023
5	DeepPET + gradient	0.741	29.52	0.909	0.91	✓ Certified	Haggstrom et al., MIA 2019
6	OS-EM + gradient	0.707	27.28	0.865	0.94	✓ Certified	Hudson & Larkin, IEEE TMI 1994
7	MAPEM-RDP + gradient	0.704	27.33	0.866	0.92	✓ Certified	Nuyts et al., IEEE TMI 2002
8	ML-EM + gradient	0.668	25.8	0.826	0.9	✓ Certified	Shepp & Vardi, IEEE TPAMI 1982
9	FBP-PET + gradient	0.635	24.49	0.785	0.88	✓ Certified	Analytical baseline
10	OSEM + gradient	0.626	23.79	0.761	0.92	✓ Certified	Hudson & Larkin, IEEE TMI 1994

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%