PET/MR

Positron Emission Tomography / Magnetic Resonance

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
🥇 PET-ViT 0.881 38.36 0.983 ✓ Certified Smith et al., ICCV 2024
🥈 U-Net-PET 0.811 34.66 0.966 ✓ Certified Ronneberger et al. variant, MICCAI 2020
🥉 PETFormer 0.810 34.64 0.965 ✓ Certified Li et al., ECCV 2024
4 TransEM 0.781 33.7 0.938 ✓ Certified Xie et al., 2023
5 DeepPET 0.749 32.4 0.918 ✓ Certified Haggstrom et al., MIA 2019
6 FBP-PET 0.707 29.96 0.916 ✓ Certified Analytical baseline
7 ML-EM 0.689 29.23 0.904 ✓ Certified Shepp & Vardi, IEEE TPAMI 1982
8 OS-EM 0.658 28.03 0.881 ✓ Certified Hudson & Larkin, IEEE TMI 1994
9 MAPEM-RDP 0.632 28.5 0.815 ✓ Certified Nuyts et al., 2002
10 OSEM 0.508 24.8 0.690 ✓ Certified Hudson & Larkin, IEEE TMI 1994

Dataset: PWM Benchmark (10 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
🥇 PET-ViT + gradient 0.778
0.848
37.28 dB / 0.979
0.769
32.81 dB / 0.951
0.717
29.63 dB / 0.911
✓ Certified Smith et al., ICCV 2024
🥈 PETFormer + gradient 0.731
0.780
32.61 dB / 0.949
0.732
30.07 dB / 0.918
0.682
27.57 dB / 0.871
✓ Certified Li et al., ECCV 2024
🥉 U-Net-PET + gradient 0.711
0.779
32.38 dB / 0.947
0.702
28.1 dB / 0.883
0.653
25.93 dB / 0.830
✓ Certified Ronneberger et al. variant, MICCAI 2020
4 TransEM + gradient 0.710
0.768
31.76 dB / 0.940
0.701
27.74 dB / 0.875
0.661
25.83 dB / 0.827
✓ Certified Xie et al., 2023
5 FBP-PET + gradient 0.690
0.703
27.6 dB / 0.872
0.701
28.23 dB / 0.886
0.666
27.16 dB / 0.862
✓ Certified Analytical baseline
6 ML-EM + gradient 0.673
0.715
27.73 dB / 0.875
0.676
26.76 dB / 0.852
0.628
25.24 dB / 0.810
✓ Certified Shepp & Vardi, IEEE TPAMI 1982
7 OS-EM + gradient 0.631
0.659
25.35 dB / 0.813
0.639
25.27 dB / 0.811
0.596
23.85 dB / 0.763
✓ Certified Hudson & Larkin, IEEE TMI 1994
8 DeepPET + gradient 0.626
0.741
29.4 dB / 0.907
0.613
24.08 dB / 0.771
0.524
20.55 dB / 0.625
✓ Certified Haggstrom et al., MIA 2019
9 OSEM + gradient 0.593
0.622
23.63 dB / 0.755
0.593
23.1 dB / 0.735
0.564
22.23 dB / 0.700
✓ Certified Hudson & Larkin, IEEE TMI 1994
10 MAPEM-RDP + gradient 0.585
0.705
27.41 dB / 0.868
0.582
22.67 dB / 0.718
0.469
19.31 dB / 0.565
✓ Certified Nuyts et al., IEEE TMI 2002

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 PET-ViT + gradient 0.848 37.28 0.979
2 PETFormer + gradient 0.780 32.61 0.949
3 U-Net-PET + gradient 0.779 32.38 0.947
4 TransEM + gradient 0.768 31.76 0.94
5 DeepPET + gradient 0.741 29.4 0.907
6 ML-EM + gradient 0.715 27.73 0.875
7 MAPEM-RDP + gradient 0.705 27.41 0.868
8 FBP-PET + gradient 0.703 27.6 0.872
9 OS-EM + gradient 0.659 25.35 0.813
10 OSEM + gradient 0.622 23.63 0.755
Spec Ranges (4 parameters)
Parameter Min Max Unit
mr_attenuation_error -25.0 50.0 %
motion_shift -6.0 12.0 pixels
b0_inhomogeneity -40.0 80.0 Hz
pet_scatter_fraction -0.25 0.5
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 PET-ViT + gradient 0.769 32.81 0.951
2 PETFormer + gradient 0.732 30.07 0.918
3 U-Net-PET + gradient 0.702 28.1 0.883
4 TransEM + gradient 0.701 27.74 0.875
5 FBP-PET + gradient 0.701 28.23 0.886
6 ML-EM + gradient 0.676 26.76 0.852
7 OS-EM + gradient 0.639 25.27 0.811
8 DeepPET + gradient 0.613 24.08 0.771
9 OSEM + gradient 0.593 23.1 0.735
10 MAPEM-RDP + gradient 0.582 22.67 0.718
Spec Ranges (4 parameters)
Parameter Min Max Unit
mr_attenuation_error -30.0 45.0 %
motion_shift -7.2 10.8 pixels
b0_inhomogeneity -48.0 72.0 Hz
pet_scatter_fraction -0.3 0.45
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 PET-ViT + gradient 0.717 29.63 0.911
2 PETFormer + gradient 0.682 27.57 0.871
3 FBP-PET + gradient 0.666 27.16 0.862
4 TransEM + gradient 0.661 25.83 0.827
5 U-Net-PET + gradient 0.653 25.93 0.83
6 ML-EM + gradient 0.628 25.24 0.81
7 OS-EM + gradient 0.596 23.85 0.763
8 OSEM + gradient 0.564 22.23 0.7
9 DeepPET + gradient 0.524 20.55 0.625
10 MAPEM-RDP + gradient 0.469 19.31 0.565
Spec Ranges (4 parameters)
Parameter Min Max Unit
mr_attenuation_error -17.5 57.5 %
motion_shift -4.2 13.8 pixels
b0_inhomogeneity -28.0 92.0 Hz
pet_scatter_fraction -0.175 0.575
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

F(k) → S(coil) → D(μ_mr) → R(θ) → D(g, η)

F MR k-Space (FFT) (k)
S Coil Sensitivities (coil)
D MR-Based Attenuation (μ_mr)
R PET Radon Transform (θ)
D Detector + Noise (g, η)

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
Δμ_mr mr_attenuation_error MR attenuation estimation error (%) 0 25.0
Δs motion_shift Inter-modality motion (pixels) 0 6.0
ΔB0 b0_inhomogeneity B0 inhomogeneity (Hz) 0 40.0
f_s pet_scatter_fraction PET scatter fraction 0 0.25

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.