Contrast-Enhanced Ultrasound (CEUS)

Contrast-Enhanced Ultrasound (CEUS)

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
🥇 DiffusionCEUS 0.891 39.6 0.962 ✓ Certified Chen 2024
🥈 SUPER-ULM 0.868 38.5 0.953 ✓ Certified Rigo 2023
🥉 CEUSF-Transformer 0.853 37.8 0.946 ✓ Certified Huang 2023
4 PINN-CEUS 0.824 36.4 0.934 ✓ Certified Lowerison 2022
5 DeepULM 0.795 35.1 0.920 ✓ Certified van Sloun 2021
6 ULM-Net 0.758 33.5 0.900 ✓ Certified Christensen-Jeffries 2020
7 CNN-Bubble 0.682 30.2 0.858 ✓ Certified Youn 2020
8 AM-CEUS 0.571 25.8 0.781 ✓ Certified Mor-Avi 2002
9 Pulse-Inversion 0.527 24.1 0.751 ✓ Certified Simpson 1999

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
🥇 SUPER-ULM + gradient 0.774
0.849
36.8 dB / 0.977
0.738
30.96 dB / 0.930
0.734
29.43 dB / 0.908
✓ Certified Rigo et al., IEEE TUFFC 2023
🥈 CEUSF-Transformer + gradient 0.765
0.819
35.48 dB / 0.971
0.761
31.34 dB / 0.935
0.716
29.6 dB / 0.911
✓ Certified Huang et al., IEEE TMI 2023
🥉 DiffusionCEUS + gradient 0.763
0.842
37.28 dB / 0.979
0.745
31.01 dB / 0.931
0.703
27.92 dB / 0.879
✓ Certified Chen et al., Nat. Methods 2024
4 PINN-CEUS + gradient 0.719
0.803
34.21 dB / 0.962
0.715
28.92 dB / 0.899
0.640
25.15 dB / 0.807
✓ Certified Lowerison et al., Nat. Commun. 2022
5 DeepULM + gradient 0.716
0.782
32.35 dB / 0.946
0.715
28.72 dB / 0.895
0.652
26.06 dB / 0.834
✓ Certified van Sloun et al., IEEE TUFFC 2021
6 ULM-Net + gradient 0.647
0.787
32.27 dB / 0.946
0.586
23.28 dB / 0.742
0.567
22.27 dB / 0.701
✓ Certified Christensen-Jeffries et al., Nat. Biomed. Eng. 2020
7 AM-CEUS + gradient 0.611
0.642
24.27 dB / 0.778
0.629
24.49 dB / 0.785
0.561
21.88 dB / 0.685
✓ Certified Mor-Avi et al., JACC 2002
8 Pulse-Inversion + gradient 0.543
0.574
22.13 dB / 0.696
0.535
20.73 dB / 0.633
0.520
20.8 dB / 0.636
✓ Certified Simpson et al., Ultrasound Med. Biol. 1999
9 CNN-Bubble + gradient 0.542
0.704
27.68 dB / 0.874
0.474
18.75 dB / 0.537
0.448
18.21 dB / 0.510
✓ Certified Youn et al., IEEE TUFFC 2020

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 SUPER-ULM + gradient 0.849 36.8 0.977
2 DiffusionCEUS + gradient 0.842 37.28 0.979
3 CEUSF-Transformer + gradient 0.819 35.48 0.971
4 PINN-CEUS + gradient 0.803 34.21 0.962
5 ULM-Net + gradient 0.787 32.27 0.946
6 DeepULM + gradient 0.782 32.35 0.946
7 CNN-Bubble + gradient 0.704 27.68 0.874
8 AM-CEUS + gradient 0.642 24.27 0.778
9 Pulse-Inversion + gradient 0.574 22.13 0.696
Spec Ranges (3 parameters)
Parameter Min Max Unit
bubble_concentration -1.0 2.0 relative
nonlinear_harmonic_extraction -2.0 4.0 -
motion_between_frames -1.0 2.0 mm
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 CEUSF-Transformer + gradient 0.761 31.34 0.935
2 DiffusionCEUS + gradient 0.745 31.01 0.931
3 SUPER-ULM + gradient 0.738 30.96 0.93
4 PINN-CEUS + gradient 0.715 28.92 0.899
5 DeepULM + gradient 0.715 28.72 0.895
6 AM-CEUS + gradient 0.629 24.49 0.785
7 ULM-Net + gradient 0.586 23.28 0.742
8 Pulse-Inversion + gradient 0.535 20.73 0.633
9 CNN-Bubble + gradient 0.474 18.75 0.537
Spec Ranges (3 parameters)
Parameter Min Max Unit
bubble_concentration -1.2 1.8 relative
nonlinear_harmonic_extraction -2.4 3.6 -
motion_between_frames -1.2 1.8 mm
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 SUPER-ULM + gradient 0.734 29.43 0.908
2 CEUSF-Transformer + gradient 0.716 29.6 0.911
3 DiffusionCEUS + gradient 0.703 27.92 0.879
4 DeepULM + gradient 0.652 26.06 0.834
5 PINN-CEUS + gradient 0.640 25.15 0.807
6 ULM-Net + gradient 0.567 22.27 0.701
7 AM-CEUS + gradient 0.561 21.88 0.685
8 Pulse-Inversion + gradient 0.520 20.8 0.636
9 CNN-Bubble + gradient 0.448 18.21 0.51
Spec Ranges (3 parameters)
Parameter Min Max Unit
bubble_concentration -0.7 2.3 relative
nonlinear_harmonic_extraction -1.4 4.6 -
motion_between_frames -0.7 2.3 mm
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

P → R → D

P Propagation
R Rotation
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
b_c bubble_concentration Bubble concentration (relative) 0.0 1.0
n_h nonlinear_harmonic_extraction Nonlinear harmonic extraction (-) 0.0 2.0
m_b motion_between_frames Motion between frames (mm) 0.0 1.0

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.