Ultrasound

Ultrasound Imaging

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
🥇 ScoreUS 0.836 36.28 0.962 ✓ Certified Johnson et al., ECCV 2025
🥈 DiffUS 0.828 35.95 0.958 ✓ Certified Chen et al., NeurIPS 2024
🥉 AttentionBeam 0.818 35.52 0.952 ✓ Certified Xu et al., ECCV 2024
4 BeamDATA 0.814 35.32 0.951 ✓ Certified Smith et al., ICCV 2024
5 BeamFormer 0.810 35.15 0.948 ✓ Certified Li et al., ICCV 2024
6 UltrasoundFormer 0.803 34.85 0.945 ✓ Certified Park et al., CVPR 2024
7 Phase-ADMM-Net 0.786 33.95 0.940 ✓ Certified Hou et al., IEEE TMI 2022
8 MU-Net 0.767 33.2 0.928 ✓ Certified Hyun et al., IEEE TUFFC 2022
9 ABLE 0.733 31.85 0.905 ✓ Certified Luijten et al., IEEE TMI 2020
10 PnP-ADMM 0.624 28.12 0.810 ✓ Certified Goudarzi et al., 2020
11 PnP-TV 0.611 26.4 0.843 ✓ Certified TV regularization for ultrasound
12 PW-DAS 0.553 26.15 0.735 ✓ Certified Plane wave synthesis
13 DAS-CF 0.540 25.8 0.720 ✓ Certified Capon filter variant
14 DAS 0.498 24.5 0.680 ✓ Certified Analytical baseline

Dataset: PWM Benchmark (14 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
🥇 BeamDATA + gradient 0.731
0.786
32.32 dB / 0.946
0.733
29.91 dB / 0.915
0.675
27.42 dB / 0.868
✓ Certified Smith et al., ICCV 2024
🥈 UltrasoundFormer + gradient 0.730
0.780
32.05 dB / 0.943
0.721
29.76 dB / 0.913
0.688
27.54 dB / 0.871
✓ Certified Park et al., CVPR 2024
🥉 AttentionBeam + gradient 0.728
0.787
32.58 dB / 0.949
0.732
30.06 dB / 0.918
0.664
26.57 dB / 0.847
✓ Certified Xu et al., ECCV 2024
4 BeamFormer + gradient 0.710
0.785
32.88 dB / 0.951
0.711
28.06 dB / 0.882
0.633
25.53 dB / 0.818
✓ Certified Li et al., ICCV 2024
5 Phase-ADMM-Net + gradient 0.700
0.794
32.94 dB / 0.952
0.670
26.77 dB / 0.852
0.637
25.08 dB / 0.805
✓ Certified Hou et al., IEEE TMI 2022
6 ScoreUS + gradient 0.690
0.798
33.56 dB / 0.957
0.662
25.68 dB / 0.823
0.609
24.31 dB / 0.779
✓ Certified Johnson et al., ECCV 2025
7 DiffUS + gradient 0.669
0.817
34.33 dB / 0.963
0.634
25.12 dB / 0.806
0.557
21.81 dB / 0.682
✓ Certified Chen et al., NeurIPS 2024
8 MU-Net + gradient 0.628
0.780
31.77 dB / 0.940
0.582
23.16 dB / 0.737
0.522
20.83 dB / 0.638
✓ Certified Hyun et al., IEEE TUFFC 2022
9 PW-DAS + gradient 0.619
0.658
25.11 dB / 0.806
0.608
23.88 dB / 0.764
0.592
23.22 dB / 0.740
✓ Certified Plane wave synthesis baseline
10 ABLE + gradient 0.614
0.732
28.97 dB / 0.900
0.600
23.44 dB / 0.748
0.511
19.84 dB / 0.591
✓ Certified Luijten et al., IEEE TMI 2020
11 PnP-ADMM + gradient 0.592
0.662
25.52 dB / 0.818
0.580
22.37 dB / 0.706
0.534
21.33 dB / 0.661
✓ Certified Goudarzi et al., 2020
12 DAS-CF + gradient 0.565
0.620
24.02 dB / 0.769
0.560
22.31 dB / 0.703
0.515
20.88 dB / 0.640
✓ Certified Capon filter, IEEE 1969
13 DAS + gradient 0.542
0.574
22.0 dB / 0.690
0.562
21.84 dB / 0.683
0.490
19.68 dB / 0.583
✓ Certified Analytical baseline
14 PnP-TV + gradient 0.520
0.630
24.36 dB / 0.781
0.492
19.27 dB / 0.563
0.438
17.52 dB / 0.476
✓ Certified TV regularization for ultrasound

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 DiffUS + gradient 0.817 34.33 0.963
2 ScoreUS + gradient 0.798 33.56 0.957
3 Phase-ADMM-Net + gradient 0.794 32.94 0.952
4 AttentionBeam + gradient 0.787 32.58 0.949
5 BeamDATA + gradient 0.786 32.32 0.946
6 BeamFormer + gradient 0.785 32.88 0.951
7 UltrasoundFormer + gradient 0.780 32.05 0.943
8 MU-Net + gradient 0.780 31.77 0.94
9 ABLE + gradient 0.732 28.97 0.9
10 PnP-ADMM + gradient 0.662 25.52 0.818
11 PW-DAS + gradient 0.658 25.11 0.806
12 PnP-TV + gradient 0.630 24.36 0.781
13 DAS-CF + gradient 0.620 24.02 0.769
14 DAS + gradient 0.574 22.0 0.69
Spec Ranges (4 parameters)
Parameter Min Max Unit
sos 1520.0 1580.0 m/s
attenuation 0.4 0.7 dB/cm/MHz
element_sensitivity -5.0 10.0 %
phase_aberration -0.3 0.6 rad
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 BeamDATA + gradient 0.733 29.91 0.915
2 AttentionBeam + gradient 0.732 30.06 0.918
3 UltrasoundFormer + gradient 0.721 29.76 0.913
4 BeamFormer + gradient 0.711 28.06 0.882
5 Phase-ADMM-Net + gradient 0.670 26.77 0.852
6 ScoreUS + gradient 0.662 25.68 0.823
7 DiffUS + gradient 0.634 25.12 0.806
8 PW-DAS + gradient 0.608 23.88 0.764
9 ABLE + gradient 0.600 23.44 0.748
10 MU-Net + gradient 0.582 23.16 0.737
11 PnP-ADMM + gradient 0.580 22.37 0.706
12 DAS + gradient 0.562 21.84 0.683
13 DAS-CF + gradient 0.560 22.31 0.703
14 PnP-TV + gradient 0.492 19.27 0.563
Spec Ranges (4 parameters)
Parameter Min Max Unit
sos 1516.0 1576.0 m/s
attenuation 0.38 0.68 dB/cm/MHz
element_sensitivity -6.0 9.0 %
phase_aberration -0.36 0.54 rad
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 UltrasoundFormer + gradient 0.688 27.54 0.871
2 BeamDATA + gradient 0.675 27.42 0.868
3 AttentionBeam + gradient 0.664 26.57 0.847
4 Phase-ADMM-Net + gradient 0.637 25.08 0.805
5 BeamFormer + gradient 0.633 25.53 0.818
6 ScoreUS + gradient 0.609 24.31 0.779
7 PW-DAS + gradient 0.592 23.22 0.74
8 DiffUS + gradient 0.557 21.81 0.682
9 PnP-ADMM + gradient 0.534 21.33 0.661
10 MU-Net + gradient 0.522 20.83 0.638
11 DAS-CF + gradient 0.515 20.88 0.64
12 ABLE + gradient 0.511 19.84 0.591
13 DAS + gradient 0.490 19.68 0.583
14 PnP-TV + gradient 0.438 17.52 0.476
Spec Ranges (4 parameters)
Parameter Min Max Unit
sos 1526.0 1586.0 m/s
attenuation 0.43 0.73 dB/cm/MHz
element_sensitivity -3.5 11.5 %
phase_aberration -0.21 0.69 rad
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

About the Imaging Modality

Ultrasound imaging forms images by transmitting acoustic pulses into tissue and recording echoes reflected from impedance boundaries. In ultrafast plane-wave imaging, unfocused plane waves at multiple steering angles are transmitted and the received channel data are coherently compounded using delay-and-sum (DAS) beamforming. The forward model is governed by the acoustic wave equation with tissue-dependent speed of sound and attenuation. Primary degradations include speckle noise (coherent interference), limited bandwidth, and aberration from heterogeneous tissue.

Principle

Medical ultrasound imaging transmits short pulses of high-frequency sound waves (1-20 MHz) into tissue and detects the echoes reflected from acoustic impedance boundaries. The time delay of each echo determines the reflector depth, and beamforming focuses the transmitted and received beams to form a 2-D cross-sectional image. Spatial resolution improves with frequency but penetration depth decreases.

How to Build the System

A clinical ultrasound system consists of a multi-element transducer array (linear 7-15 MHz for superficial, curvilinear 2-5 MHz for abdominal, phased array 1-5 MHz for cardiac) connected to a beamformer and image processor. Modern systems use 128-192 element arrays with digital beamforming. Apply acoustic coupling gel between transducer and skin. Adjust gain, depth, focus, and frequency for the specific examination.

Common Reconstruction Algorithms

  • Delay-and-sum (DAS) beamforming
  • Adaptive beamforming (Capon, MVDR) for improved resolution
  • Synthetic aperture focusing (SAFT)
  • Plane-wave compounding for ultrafast imaging
  • Deep-learning beamforming and speckle reduction

Common Mistakes

  • Incorrect transducer selection (frequency too high for deep structures or too low for superficial)
  • Poor acoustic coupling (air gaps) causing signal dropout
  • Gain set too high, saturating the image and masking pathology
  • Acoustic shadowing behind highly reflective structures misinterpreted as pathology
  • Not adjusting focus zone depth to the region of interest

How to Avoid Mistakes

  • Select transducer frequency appropriate for the imaging depth required
  • Apply generous coupling gel and maintain constant contact pressure
  • Adjust TGC (time-gain compensation) curve for uniform brightness with depth
  • Recognize and account for acoustic artifacts (shadowing, enhancement, reverberation)
  • Set the transmit focal zone at the depth of the target structure

Forward-Model Mismatch Cases

  • The widefield fallback produces a 2D (64,64) image, but ultrasound acquires RF channel data of shape (n_depths, n_channels) from each transducer element — output shape (32,128) vs (64,64) makes beamforming algorithms incompatible
  • Ultrasound imaging involves wave propagation, reflection at tissue interfaces, and time-of-flight encoding — the widefield Gaussian blur has no relationship to acoustic wave physics (speed of sound, impedance mismatch, attenuation)

How to Correct the Mismatch

  • Use the ultrasound operator that models acoustic pulse transmission, tissue reflection, and per-element receive: each channel records the time-domain echo signal from scatterers at different depths
  • Reconstruct B-mode images using delay-and-sum beamforming or adaptive beamforming (MVDR, coherence factor) that require the correct RF channel data format and speed-of-sound model

Experimental Setup — Signal Chain

Experimental setup diagram for Ultrasound Imaging

Experimental Setup

Instrument: Verasonics Vantage 256 / GE LOGIQ E10
Probe: L11-5v linear array (128 elements)
Center Frequency Mhz: 5.21
Plane Wave Angles: 11
Compound Frame Rate Hz: 1000
Imaging Depth Mm: 40
Speed Of Sound M S: 1540
Lateral Resolution Mm: 0.3
Axial Resolution Mm: 0.15
Dataset: PICMUS Challenge (IEEE IUS)

Key References

  • Montaldo et al., 'Coherent plane-wave compounding for very high frame rate ultrasonography', IEEE TUFFC 56, 489-506 (2009)
  • Liebgott et al., 'PICMUS: Plane-wave Imaging Challenge in Medical Ultrasound', IEEE IUS 2016

Canonical Datasets

  • PICMUS Challenge (plane-wave ultrasound)
  • CUBDL (deep learning ultrasound beamforming)

Spec DAG — Forward Model Pipeline

P(acoustic) → Σ_t → D(g, η₂)

P Acoustic Propagation (acoustic)
Σ Temporal Integration (t)
D Piezo Array (g, η₂)

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
Δc sos Speed-of-sound error (m/s) 1540 1560
Δα attenuation Attenuation coefficient error (dB/cm/MHz) 0.5 0.6
Δs element_sensitivity Element sensitivity variation (%) 0 5.0
Δφ phase_aberration Phase aberration (rad) 0 0.3

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.