Digital Breast Tomosynthesis (DBT)

Digital Breast Tomosynthesis (DBT)

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
🥇 DiffusionDBT 0.885 39.4 0.956 ✓ Certified Gao et al., MICCAI 2024
🥈 PhysDBT 0.857 38.1 0.945 ✓ Certified Nett et al., IEEE TMI 2024
🥉 SwinDBT 0.839 37.2 0.938 ✓ Certified Li et al., Med. Phys. 2023
4 TransDBT 0.807 35.8 0.921 ✓ Certified Wang et al., MICCAI 2022
5 DuDoRNet-DBT 0.754 33.5 0.891 ✓ Certified Zhou et al., CVPR 2020
6 DnCNN-DBT 0.677 30.2 0.848 ✓ Certified Chen et al., IEEE TMI 2018
7 SART-DBT 0.607 27.4 0.801 ✓ Certified Andersen & Kak, Ultrason. Imaging 1984
8 TV-DBT 0.564 25.8 0.768 ✓ Certified Sidky et al., Med. Phys. 2014
9 FBP-DBT 0.496 23.1 0.721 ✓ Certified Sechopoulos, Med. Phys. 2013

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
🥇 PhysDBT + gradient 0.769
0.844
36.67 dB / 0.977
0.744
30.11 dB / 0.918
0.719
29.52 dB / 0.909
✓ Certified Nett et al., IEEE TMI 2024
🥈 DiffusionDBT + gradient 0.749
0.860
38.09 dB / 0.982
0.723
28.74 dB / 0.895
0.664
25.95 dB / 0.831
✓ Certified Gao et al., MICCAI 2024
🥉 SwinDBT + gradient 0.745
0.833
35.81 dB / 0.972
0.730
30.48 dB / 0.924
0.672
26.26 dB / 0.839
✓ Certified Li et al., Med. Phys. 2023
4 TransDBT + gradient 0.704
0.792
32.9 dB / 0.952
0.712
28.39 dB / 0.889
0.608
23.69 dB / 0.757
✓ Certified Wang et al., MICCAI 2022
5 DuDoRNet-DBT + gradient 0.668
0.760
30.82 dB / 0.929
0.649
25.55 dB / 0.819
0.594
23.44 dB / 0.748
✓ Certified Zhou et al., CVPR 2020
6 DnCNN-DBT + gradient 0.623
0.710
28.22 dB / 0.885
0.600
23.23 dB / 0.740
0.559
22.47 dB / 0.710
✓ Certified Chen et al., IEEE TMI 2018
7 SART-DBT + gradient 0.610
0.655
25.46 dB / 0.816
0.602
23.15 dB / 0.737
0.573
22.66 dB / 0.717
✓ Certified Andersen & Kak, Ultrason. Imaging 1984
8 FBP-DBT + gradient 0.529
0.570
21.48 dB / 0.667
0.514
20.57 dB / 0.626
0.502
20.29 dB / 0.613
✓ Certified Sechopoulos, Med. Phys. 2013
9 TV-DBT + gradient 0.447
0.604
22.98 dB / 0.730
0.410
16.39 dB / 0.420
0.326
13.62 dB / 0.294
✓ Certified Sidky et al., Med. Phys. 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 DiffusionDBT + gradient 0.860 38.09 0.982
2 PhysDBT + gradient 0.844 36.67 0.977
3 SwinDBT + gradient 0.833 35.81 0.972
4 TransDBT + gradient 0.792 32.9 0.952
5 DuDoRNet-DBT + gradient 0.760 30.82 0.929
6 DnCNN-DBT + gradient 0.710 28.22 0.885
7 SART-DBT + gradient 0.655 25.46 0.816
8 TV-DBT + gradient 0.604 22.98 0.73
9 FBP-DBT + gradient 0.570 21.48 0.667
Spec Ranges (3 parameters)
Parameter Min Max Unit
angular_range_error -0.4 0.8 degtotal
detector_motion_blur -0.1 0.2 px
scatter_fraction 0.24 0.42 -
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 PhysDBT + gradient 0.744 30.11 0.918
2 SwinDBT + gradient 0.730 30.48 0.924
3 DiffusionDBT + gradient 0.723 28.74 0.895
4 TransDBT + gradient 0.712 28.39 0.889
5 DuDoRNet-DBT + gradient 0.649 25.55 0.819
6 SART-DBT + gradient 0.602 23.15 0.737
7 DnCNN-DBT + gradient 0.600 23.23 0.74
8 FBP-DBT + gradient 0.514 20.57 0.626
9 TV-DBT + gradient 0.410 16.39 0.42
Spec Ranges (3 parameters)
Parameter Min Max Unit
angular_range_error -0.48 0.72 degtotal
detector_motion_blur -0.12 0.18 px
scatter_fraction 0.228 0.408 -
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 PhysDBT + gradient 0.719 29.52 0.909
2 SwinDBT + gradient 0.672 26.26 0.839
3 DiffusionDBT + gradient 0.664 25.95 0.831
4 TransDBT + gradient 0.608 23.69 0.757
5 DuDoRNet-DBT + gradient 0.594 23.44 0.748
6 SART-DBT + gradient 0.573 22.66 0.717
7 DnCNN-DBT + gradient 0.559 22.47 0.71
8 FBP-DBT + gradient 0.502 20.29 0.613
9 TV-DBT + gradient 0.326 13.62 0.294
Spec Ranges (3 parameters)
Parameter Min Max Unit
angular_range_error -0.28 0.92 degtotal
detector_motion_blur -0.07 0.23 px
scatter_fraction 0.258 0.438 -
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

Π → D

Π Projection
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
a_r angular_range_error Angular range error (deg total) 0.0 0.4
d_m detector_motion_blur Detector motion blur (px) 0.0 0.1
s_f scatter_fraction Scatter fraction (-) 0.3 0.36

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.