Brachytherapy 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
🥇	DiffusionSeed DiffusionSeed Gao et al., Med. Phys. 2024 40.3 dB SSIM 0.968 Checkpoint unavailable	0.906	40.3	0.968	✓ Certified	Gao et al., Med. Phys. 2024
🥈	CTFormer CTFormer Wang et al., MICCAI 2023 39.1 dB SSIM 0.957 Checkpoint unavailable	0.880	39.1	0.957	✓ Certified	Wang et al., MICCAI 2023
🥉	DuDoTrans DuDoTrans Wang et al., IEEE TMI 2022 38.2 dB SSIM 0.948 Checkpoint unavailable	0.861	38.2	0.948	✓ Certified	Wang et al., IEEE TMI 2022
4	Learned Primal-Dual Learned Primal-Dual Adler & Oktem, IEEE TMI 2018 37.0 dB SSIM 0.935 Checkpoint unavailable	0.834	37.0	0.935	✓ Certified	Adler & Oktem, IEEE TMI 2018
5	Metal-AR-Net Metal-AR-Net Zhang & Yu, IEEE TMI 2018 36.4 dB SSIM 0.928 Checkpoint unavailable	0.821	36.4	0.928	✓ Certified	Zhang & Yu, IEEE TMI 2018
6	RED-CNN RED-CNN Chen et al., IEEE TMI 2017 35.1 dB SSIM 0.912 Checkpoint unavailable	0.791	35.1	0.912	✓ Certified	Chen et al., IEEE TMI 2017
7	FBPConvNet FBPConvNet Jin et al., IEEE TIP 2017 34.2 dB SSIM 0.895 Checkpoint unavailable	0.768	34.2	0.895	✓ Certified	Jin et al., IEEE TIP 2017
8	TV-ADMM TV-ADMM Boyd et al., Found. Trends Mach. Learn. 2011 31.8 dB SSIM 0.861 Try in SpecLab →	0.711	31.8	0.861	✓ Certified	Boyd et al., Found. Trends Mach. Learn. 2011
9	FDK FDK Feldkamp et al., J. Opt. Soc. Am. A 1984 28.5 dB SSIM 0.812 Try in SpecLab →	0.631	28.5	0.812	✓ Certified	Feldkamp et al., J. Opt. Soc. Am. A 1984

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
🥇	DiffusionSeed + gradient DiffusionSeed + gradient Gao et al., Med. Phys. 2024 Score 0.785 Correct & Reconstruct →	0.785	0.849 38.22 dB / 0.983	0.777 32.87 dB / 0.951	0.729 30.38 dB / 0.922	✓ Certified	Gao et al., Med. Phys. 2024
🥈	CTFormer + gradient CTFormer + gradient Wang et al., MICCAI 2023 Score 0.764 Correct & Reconstruct →	0.764	0.837 36.99 dB / 0.978	0.765 31.59 dB / 0.938	0.691 27.62 dB / 0.873	✓ Certified	Wang et al., MICCAI 2023
🥉	DuDoTrans + gradient DuDoTrans + gradient Wang et al., IEEE TMI 2022 Score 0.763 Correct & Reconstruct →	0.763	0.825 35.97 dB / 0.973	0.758 31.95 dB / 0.942	0.705 28.72 dB / 0.895	✓ Certified	Wang et al., IEEE TMI 2022
4	Learned Primal-Dual + gradient Learned Primal-Dual + gradient Adler & Oktem, IEEE TMI 2018 Score 0.762 Correct & Reconstruct →	0.762	0.830 35.38 dB / 0.970	0.750 31.08 dB / 0.932	0.707 29.18 dB / 0.903	✓ Certified	Adler & Oktem, IEEE TMI 2018
5	Metal-AR-Net + gradient Metal-AR-Net + gradient Zhang & Yu, IEEE TMI 2018 Score 0.725 Correct & Reconstruct →	0.725	0.802 33.99 dB / 0.961	0.712 28.55 dB / 0.892	0.660 25.59 dB / 0.820	✓ Certified	Zhang & Yu, IEEE TMI 2018
6	RED-CNN + gradient RED-CNN + gradient Chen et al., IEEE TMI 2017 Score 0.722 Correct & Reconstruct →	0.722	0.809 34.05 dB / 0.961	0.687 27.8 dB / 0.877	0.671 26.21 dB / 0.838	✓ Certified	Chen et al., IEEE TMI 2017
7	FBPConvNet + gradient FBPConvNet + gradient Jin et al., IEEE TIP 2017 Score 0.697 Correct & Reconstruct →	0.697	0.771 31.89 dB / 0.941	0.695 27.59 dB / 0.872	0.625 24.74 dB / 0.794	✓ Certified	Jin et al., IEEE TIP 2017
8	FDK + gradient FDK + gradient Feldkamp et al., J. Opt. Soc. Am. A 1984 Score 0.635 Correct & Reconstruct →	0.635	0.672 26.03 dB / 0.833	0.639 24.64 dB / 0.791	0.594 23.87 dB / 0.764	✓ Certified	Feldkamp et al., J. Opt. Soc. Am. A 1984
9	TV-ADMM + gradient TV-ADMM + gradient Boyd et al., Found. Trends Mach. Learn. 2011 Score 0.626 Correct & Reconstruct →	0.626	0.733 29.3 dB / 0.906	0.597 23.56 dB / 0.753	0.549 21.6 dB / 0.673	✓ Certified	Boyd et al., Found. Trends Mach. Learn. 2011

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	DiffusionSeed + gradient	0.849	38.22	0.983
2	CTFormer + gradient	0.837	36.99	0.978
3	Learned Primal-Dual + gradient	0.830	35.38	0.97
4	DuDoTrans + gradient	0.825	35.97	0.973
5	RED-CNN + gradient	0.809	34.05	0.961
6	Metal-AR-Net + gradient	0.802	33.99	0.961
7	FBPConvNet + gradient	0.771	31.89	0.941
8	TV-ADMM + gradient	0.733	29.3	0.906
9	FDK + gradient	0.672	26.03	0.833

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
source_position_error	-0.4	0.8	mm
attenuation_coefficient	0.19	0.22	1/cm
detector_gain_drift	0.99	1.02	-
scatter_fraction	0.13	0.19	-

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	DiffusionSeed + gradient	0.777	32.87	0.951
2	CTFormer + gradient	0.765	31.59	0.938
3	DuDoTrans + gradient	0.758	31.95	0.942
4	Learned Primal-Dual + gradient	0.750	31.08	0.932
5	Metal-AR-Net + gradient	0.712	28.55	0.892
6	FBPConvNet + gradient	0.695	27.59	0.872
7	RED-CNN + gradient	0.687	27.8	0.877
8	FDK + gradient	0.639	24.64	0.791
9	TV-ADMM + gradient	0.597	23.56	0.753

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
source_position_error	-0.48	0.72	mm
attenuation_coefficient	0.188	0.218	1/cm
detector_gain_drift	0.988	1.018	-
scatter_fraction	0.126	0.186	-

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	DiffusionSeed + gradient	0.729	30.38	0.922
2	Learned Primal-Dual + gradient	0.707	29.18	0.903
3	DuDoTrans + gradient	0.705	28.72	0.895
4	CTFormer + gradient	0.691	27.62	0.873
5	RED-CNN + gradient	0.671	26.21	0.838
6	Metal-AR-Net + gradient	0.660	25.59	0.82
7	FBPConvNet + gradient	0.625	24.74	0.794
8	FDK + gradient	0.594	23.87	0.764
9	TV-ADMM + gradient	0.549	21.6	0.673

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
source_position_error	-0.28	0.92	mm
attenuation_coefficient	0.193	0.223	1/cm
detector_gain_drift	0.993	1.023	-
scatter_fraction	0.136	0.196	-

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
s_p	source_position_error	Source position error (mm)	0.0	0.4
a_c	attenuation_coefficient	Attenuation coefficient (1/cm)	0.2	0.21
d_g	detector_gain_drift	Detector gain drift (-)	1.0	1.01
s_f	scatter_fraction	Scatter fraction (-)	0.15	0.17

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.