Arterial Spin Labeling (ASL) MRI

Arterial Spin Labeling (ASL) MRI

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
🥇 Score-MRI (ASL) 0.833 36.7 0.942 ✓ Certified Chung & Ye, Med. Image Anal. 93:102689, 2022
🥈 PromptMR 0.819 36.1 0.934 ✓ Certified Xin et al., ECCV 2024
🥉 ReconFormer 0.801 35.4 0.922 ✓ Certified Guo et al., IEEE TMI 41(5):1297, 2024
4 E2E-VarNet 0.781 34.6 0.908 ✓ Certified Sriram et al., MICCAI 2020
5 Kinetic-CS 0.749 33.2 0.891 ✓ Certified Zhao et al., JMRI 60(4):1204, 2024
6 U-Net (ASL) 0.723 32.1 0.876 ✓ Certified Tian et al., MRM 89(4):1616, 2023
7 PnP-DnCNN 0.668 29.8 0.843 ✓ Certified Ahmad et al., IEEE SPM 2020
8 L1-Wavelet (ESPIRiT) 0.632 28.3 0.820 ✓ Certified Lustig et al., MRM 2007; Uecker et al., MRM 2014
9 Zero-Filled IFFT 0.448 24.5 0.580 ✓ Certified Zbontar et al., fastMRI, arXiv 2018

Dataset: PWM Benchmark (9 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
🥇 Score-MRI (ASL) + gradient 0.749
0.804
33.97 dB / 0.961
0.729
29.29 dB / 0.905
0.713
29.27 dB / 0.905
✓ Certified Chung & Ye, Med. Image Anal. 93:102689, 2022
🥈 PromptMR + gradient 0.733
0.820
34.76 dB / 0.966
0.724
29.22 dB / 0.904
0.654
25.68 dB / 0.823
✓ Certified Xin et al., ECCV 2024
🥉 ReconFormer + gradient 0.721
0.790
33.23 dB / 0.955
0.714
29.19 dB / 0.904
0.660
26.34 dB / 0.841
✓ Certified Guo et al., IEEE TMI 41(5):1297, 2024
4 E2E-VarNet + gradient 0.708
0.800
33.24 dB / 0.955
0.703
28.01 dB / 0.881
0.621
24.3 dB / 0.779
✓ Certified Sriram et al., MICCAI 2020
5 U-Net (ASL) + gradient 0.678
0.764
30.59 dB / 0.925
0.648
25.03 dB / 0.803
0.622
24.45 dB / 0.784
✓ Certified Tian et al., MRM 89(4):1616, 2023
6 PnP-DnCNN + gradient 0.660
0.704
27.97 dB / 0.880
0.656
25.51 dB / 0.818
0.620
24.12 dB / 0.773
✓ Certified Ahmad et al., IEEE SPM 2020
7 Kinetic-CS + gradient 0.627
0.755
30.67 dB / 0.926
0.600
23.88 dB / 0.764
0.526
20.72 dB / 0.633
✓ Certified Zhao et al., JMRI 60(4):1204, 2024
8 Zero-Filled IFFT + gradient 0.534
0.567
21.6 dB / 0.673
0.551
21.09 dB / 0.650
0.485
19.84 dB / 0.591
✓ Certified Zbontar et al., fastMRI, arXiv 2018
9 L1-Wavelet (ESPIRiT) + gradient 0.504
0.668
25.75 dB / 0.825
0.468
18.32 dB / 0.516
0.376
15.44 dB / 0.375
✓ Certified Lustig et al., MRM 2007; Uecker et al., MRM 2014

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 PromptMR + gradient 0.820 34.76 0.966
2 Score-MRI (ASL) + gradient 0.804 33.97 0.961
3 E2E-VarNet + gradient 0.800 33.24 0.955
4 ReconFormer + gradient 0.790 33.23 0.955
5 U-Net (ASL) + gradient 0.764 30.59 0.925
6 Kinetic-CS + gradient 0.755 30.67 0.926
7 PnP-DnCNN + gradient 0.704 27.97 0.88
8 L1-Wavelet (ESPIRiT) + gradient 0.668 25.75 0.825
9 Zero-Filled IFFT + gradient 0.567 21.6 0.673
Spec Ranges (3 parameters)
Parameter Min Max Unit
labeling_efficiency 0.83 0.89 -
transit_delay 1.2 2.1 s
t1_blood_error -2.0 4.0 -
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 Score-MRI (ASL) + gradient 0.729 29.29 0.905
2 PromptMR + gradient 0.724 29.22 0.904
3 ReconFormer + gradient 0.714 29.19 0.904
4 E2E-VarNet + gradient 0.703 28.01 0.881
5 PnP-DnCNN + gradient 0.656 25.51 0.818
6 U-Net (ASL) + gradient 0.648 25.03 0.803
7 Kinetic-CS + gradient 0.600 23.88 0.764
8 Zero-Filled IFFT + gradient 0.551 21.09 0.65
9 L1-Wavelet (ESPIRiT) + gradient 0.468 18.32 0.516
Spec Ranges (3 parameters)
Parameter Min Max Unit
labeling_efficiency 0.826 0.886 -
transit_delay 1.14 2.04 s
t1_blood_error -2.4 3.6 -
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 Score-MRI (ASL) + gradient 0.713 29.27 0.905
2 ReconFormer + gradient 0.660 26.34 0.841
3 PromptMR + gradient 0.654 25.68 0.823
4 U-Net (ASL) + gradient 0.622 24.45 0.784
5 E2E-VarNet + gradient 0.621 24.3 0.779
6 PnP-DnCNN + gradient 0.620 24.12 0.773
7 Kinetic-CS + gradient 0.526 20.72 0.633
8 Zero-Filled IFFT + gradient 0.485 19.84 0.591
9 L1-Wavelet (ESPIRiT) + gradient 0.376 15.44 0.375
Spec Ranges (3 parameters)
Parameter Min Max Unit
labeling_efficiency 0.836 0.896 -
transit_delay 1.29 2.19 s
t1_blood_error -1.4 4.6 -
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 Nominal Perturbed
l_e labeling_efficiency Labeling efficiency (-) 0.85 0.87
t_d transit_delay Transit delay (s) 1.5 1.8
t_b t1_blood_error T1 blood error (-) 0.0 2.0

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.