Public

CT — Public Tier

(11 scenes)

Full-access tier: 11 real patient CT slices from LoDoPaB-CT (LIDC/IDRI test split).

What you get

Measured sinogram (y), ideal forward operator (H), spec ranges, ground truth x_true, and true mismatch spec per sample.

How to use

Load ct_challenge_public.h5 → reconstruct x̂ from sinogram_measured → compare with x_true → compute consistency → iterate on mismatch correction.

What to submit

Reconstructed images (x_hat) and corrected mismatch spec as HDF5.

Parameter Specifications

✓

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

Parameter Spec Range True Value Unit
center_offset_px -4.0 – 6.0 1.0 px
angle_error_deg -6.5 – 9.5 1.5 deg
beam_hardening_beta -0.1 – 0.2 0.05
detector_tilt_deg -2.5 – 3.5 0.5 deg

InverseNet Baseline Scores

Method: CPU_baseline — Mismatch parameter: nominal

Scenario I (Ideal)

13.95 dB

SSIM 0.1104

Scenario II (Mismatch)

11.23 dB

SSIM 0.0542

Scenario III (Oracle)

14.82 dB

SSIM 0.1569

Per-scene breakdown (4 scenes)
Scene PSNR I SSIM I PSNR II SSIM II PSNR III SSIM III
scene_00 13.02 0.1100 10.90 0.0538 14.55 0.1630
scene_01 13.63 0.1059 10.91 0.0520 14.53 0.1501
scene_02 14.80 0.1169 11.75 0.0561 15.25 0.1561
scene_03 14.34 0.1087 11.36 0.0549 14.96 0.1586

Public Tier Leaderboard

# Method Score PSNR SSIM Consistency Trust Source
1 CT-FM + gradient 0.910 42.34 0.992 0.95 ✓ Certified Wang et al., Nature MI 2026
2 CT-MAE + gradient 0.902 41.95 0.992 0.93 ✓ Certified Chen et al., MICCAI 2024
3 DiffusionMBIR + gradient 0.894 41.38 0.991 0.92 ✓ Certified Song et al., arXiv 2024
4 PINER-CT + gradient 0.886 41.79 0.991 0.86 ✓ Certified Sun et al., CVPR 2025
5 Score-CT + gradient 0.877 40.38 0.989 0.89 ✓ Certified Gao et al., IEEE TMI 2024
6 Eformer + gradient 0.871 39.27 0.986 0.92 ✓ Certified Wang et al., AAAI 2022
7 TransCT + gradient 0.865 38.72 0.984 0.92 ✓ Certified Xia et al., MICCAI 2021
8 CTformer + gradient 0.861 39.08 0.985 0.88 ✓ Certified Wang et al., MICCAI 2023
9 DuDoRNet + gradient 0.849 37.14 0.979 0.93 ✓ Certified Zhou et al., CVPR 2020
10 LEARN + gradient 0.831 35.77 0.972 0.92 ✓ Certified Chen et al., IEEE TPAMI 2018
11 iCT-Net + gradient 0.814 34.76 0.966 0.9 ✓ Certified Li et al., IEEE TMI 2019
12 RED-CNN + gradient 0.801 33.88 0.96 0.89 ✓ Certified Chen et al., IEEE TMI 2017
13 FBPConvNet + gradient 0.774 32.33 0.946 0.86 ✓ Certified Jin et al., IEEE TMI 2017
14 WGAN-CT + gradient 0.767 31.61 0.938 0.88 ✓ Certified Wolterink et al., IEEE TMI 2017
15 CT-U-Net + gradient 0.763 31.3 0.935 0.88 ✓ Certified Han et al., Phys. Med. Biol. 2016
16 DLCT + gradient 0.762 30.6 0.926 0.93 ✓ Certified Xu et al., IEEE TMI 2012
17 BM3D-CT + gradient 0.754 29.97 0.916 0.94 ✓ Certified Dabov et al., IEEE TIP 2007; Chen 2014
18 PnP-ADMM + gradient 0.747 30.38 0.922 0.87 ✓ Certified Venkatakrishnan et al., GlobalSIP 2013
19 TV-ADMM + gradient 0.736 28.86 0.898 0.94 ✓ Certified Sidky & Pan, Phys. Med. Biol. 2008
20 ART-TV + gradient 0.697 27.28 0.865 0.89 ✓ Certified Li et al., Med. Phys. 2004
21 SART + gradient 0.679 26.38 0.842 0.89 ✓ Certified Andersen & Kak, Ultrason. Imaging 1984
22 OSEM + gradient 0.654 25.31 0.812 0.88 ✓ Certified Hudson & Larkin, IEEE TMI 1994
23 CGLS + gradient 0.650 25.33 0.812 0.86 ✓ Certified Bjorck, SIAM 1996
24 FBP + gradient 0.588 22.38 0.706 0.91 ✓ Certified Kak & Slaney, IEEE Press 1988

Visible Data Fields

y H_ideal spec_ranges x_true true_spec

Dataset

Format: HDF5
Scenes: 11

Scoring Formula

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

PSNR: 40% SSIM: 40% Consistency: 20%
Back to CT