CT + Fluorescence (FLIT)

CT + Fluorescence (FLIT)

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
🥇 DiffusionXRF 0.896 40.1 0.955 ✓ Certified Song 2021
🥈 PhysXRF-Net 0.862 38.5 0.941 ✓ Certified Raissi 2019
🥉 SwinXRF 0.846 37.8 0.932 ✓ Certified Liu 2021
4 PnP-XRF 0.805 35.9 0.914 ✓ Certified Chan 2016
5 U-Net-XRF 0.777 34.6 0.901 ✓ Certified Ronneberger 2015
6 DnCNN-XRF 0.726 32.4 0.872 ✓ Certified Zhang 2017
7 TV-XRFCT 0.660 29.7 0.831 ✓ Certified Larsson 2020
8 MLEM-XRF 0.570 26.3 0.764 ✓ Certified Jaszczak 1981
9 FBP-XRF 0.480 22.8 0.701 ✓ Certified Boisseau 1987

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
🥇 DiffusionXRF + gradient 0.795
0.867
38.4 dB / 0.983
0.774
32.32 dB / 0.946
0.745
31.46 dB / 0.937
✓ Certified Song et al., ICLR 2021 (XRF adapt.)
🥈 PnP-XRF + gradient 0.771
0.817
34.39 dB / 0.964
0.760
30.99 dB / 0.931
0.736
29.72 dB / 0.912
✓ Certified Chan et al., IEEE TIP 2016 (XRF adapt.)
🥉 SwinXRF + gradient 0.760
0.841
36.25 dB / 0.975
0.748
31.48 dB / 0.937
0.692
28.32 dB / 0.887
✓ Certified Liu et al., ICCV 2021 (XRF adapt.)
4 PhysXRF-Net + gradient 0.731
0.828
35.68 dB / 0.972
0.711
28.48 dB / 0.891
0.653
25.99 dB / 0.832
✓ Certified Raissi et al., J. Comput. Phys. 2019 (XRF)
5 U-Net-XRF + gradient 0.680
0.802
33.47 dB / 0.957
0.651
25.08 dB / 0.805
0.587
22.73 dB / 0.720
✓ Certified Ronneberger et al., MICCAI 2015 (XRF adapt.)
6 DnCNN-XRF + gradient 0.623
0.768
30.74 dB / 0.927
0.606
23.16 dB / 0.737
0.494
19.56 dB / 0.577
✓ Certified Zhang et al., IEEE TIP 2017 (XRF adapt.)
7 MLEM-XRF + gradient 0.606
0.654
24.78 dB / 0.795
0.600
23.61 dB / 0.754
0.563
22.54 dB / 0.713
✓ Certified Jaszczak et al., IEEE TNS 1981 (XRF adapt.)
8 TV-XRFCT + gradient 0.569
0.724
28.32 dB / 0.887
0.529
20.61 dB / 0.628
0.453
18.84 dB / 0.542
✓ Certified Larsson et al., Phys. Med. Biol. 2020
9 FBP-XRF + gradient 0.489
0.531
20.47 dB / 0.621
0.478
18.87 dB / 0.543
0.458
18.19 dB / 0.509
✓ Certified Boisseau & Grodzins, Hyperfine Int. 1987

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 DiffusionXRF + gradient 0.867 38.4 0.983
2 SwinXRF + gradient 0.841 36.25 0.975
3 PhysXRF-Net + gradient 0.828 35.68 0.972
4 PnP-XRF + gradient 0.817 34.39 0.964
5 U-Net-XRF + gradient 0.802 33.47 0.957
6 DnCNN-XRF + gradient 0.768 30.74 0.927
7 TV-XRFCT + gradient 0.724 28.32 0.887
8 MLEM-XRF + gradient 0.654 24.78 0.795
9 FBP-XRF + gradient 0.531 20.47 0.621
Spec Ranges (3 parameters)
Parameter Min Max Unit
optical_property_assignment_error -6.0 12.0 -
autofluorescence -10.0 20.0 -
registration_(ct_to_optical) -0.6 1.2 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 DiffusionXRF + gradient 0.774 32.32 0.946
2 PnP-XRF + gradient 0.760 30.99 0.931
3 SwinXRF + gradient 0.748 31.48 0.937
4 PhysXRF-Net + gradient 0.711 28.48 0.891
5 U-Net-XRF + gradient 0.651 25.08 0.805
6 DnCNN-XRF + gradient 0.606 23.16 0.737
7 MLEM-XRF + gradient 0.600 23.61 0.754
8 TV-XRFCT + gradient 0.529 20.61 0.628
9 FBP-XRF + gradient 0.478 18.87 0.543
Spec Ranges (3 parameters)
Parameter Min Max Unit
optical_property_assignment_error -7.2 10.8 -
autofluorescence -12.0 18.0 -
registration_(ct_to_optical) -0.72 1.08 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 DiffusionXRF + gradient 0.745 31.46 0.937
2 PnP-XRF + gradient 0.736 29.72 0.912
3 SwinXRF + gradient 0.692 28.32 0.887
4 PhysXRF-Net + gradient 0.653 25.99 0.832
5 U-Net-XRF + gradient 0.587 22.73 0.72
6 MLEM-XRF + gradient 0.563 22.54 0.713
7 DnCNN-XRF + gradient 0.494 19.56 0.577
8 FBP-XRF + gradient 0.458 18.19 0.509
9 TV-XRFCT + gradient 0.453 18.84 0.542
Spec Ranges (3 parameters)
Parameter Min Max Unit
optical_property_assignment_error -4.2 13.8 -
autofluorescence -7.0 23.0 -
registration_(ct_to_optical) -0.42 1.38 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

(Π → D) + (M → R → P → D) → ⊕

Π Projection
D Detector (CT)
M Modulation
R Rotation
P Propagation
D Detector (FLI)
Fusion Fusion

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
o_p optical_property_assignment_error Optical property assignment error (-) 0.0 6.0
a autofluorescence Autofluorescence (-) 0.0 10.0
r_( registration_(ct_to_optical) Registration (CT to optical) (mm) 0.0 0.6

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.