DESI Mass Spectrometry Imaging

DESI Mass Spectrometry 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
🥇 DiffusionMSI 0.858 38.2 0.942 ✓ Certified Palmer 2024
🥈 MSIFormer 0.812 36.1 0.921 ✓ Certified Kalinichenko 2023
🥉 SpaMSI-Net 0.782 34.8 0.904 ✓ Certified Rappez 2021
4 MSI-GAN 0.756 33.7 0.888 ✓ Certified Yang 2021
5 DeepMSI 0.726 32.4 0.871 ✓ Certified Gruber 2021
6 MSI-TV 0.642 28.9 0.821 ✓ Certified Fonville 2012
7 MSI-NMF 0.579 26.3 0.782 ✓ Certified Blanco 2013
8 MSI-PCA 0.538 24.8 0.749 ✓ Certified Alexandrov 2010
9 MSI-Hotelling 0.469 22.1 0.701 ✓ Certified Deininger 2011

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
🥇 MSIFormer + gradient 0.773
0.799
34.04 dB / 0.961
0.775
32.55 dB / 0.948
0.744
30.72 dB / 0.927
✓ Certified Kalinichenko et al., Nat. Methods 2023
🥈 DiffusionMSI + gradient 0.768
0.824
35.52 dB / 0.971
0.756
31.11 dB / 0.932
0.723
30.4 dB / 0.923
✓ Certified Palmer et al., Nat. Methods 2024
🥉 MSI-GAN + gradient 0.699
0.764
31.39 dB / 0.936
0.677
26.42 dB / 0.843
0.656
25.97 dB / 0.831
✓ Certified Yang et al., Anal. Chem. 2021
4 DeepMSI + gradient 0.645
0.769
31.06 dB / 0.932
0.630
24.87 dB / 0.798
0.535
21.03 dB / 0.647
✓ Certified Gruber et al., Anal. Chem. 2021
5 SpaMSI-Net + gradient 0.644
0.777
31.84 dB / 0.941
0.628
24.59 dB / 0.789
0.528
21.33 dB / 0.661
✓ Certified Rappez et al., Nat. Methods 2021
6 MSI-TV + gradient 0.611
0.709
27.51 dB / 0.870
0.600
23.37 dB / 0.745
0.523
20.34 dB / 0.615
✓ Certified Fonville et al., Bioinformatics 2012
7 MSI-NMF + gradient 0.605
0.659
25.09 dB / 0.805
0.607
23.58 dB / 0.753
0.548
21.83 dB / 0.683
✓ Certified Blanco et al., Anal. Chem. 2013
8 MSI-PCA + gradient 0.571
0.585
22.48 dB / 0.710
0.585
22.93 dB / 0.728
0.542
21.2 dB / 0.655
✓ Certified Alexandrov et al., J. Bioinform. Comput. Biol. 2010
9 MSI-Hotelling + gradient 0.499
0.542
20.44 dB / 0.620
0.516
20.22 dB / 0.609
0.438
17.59 dB / 0.480
✓ Certified Deininger et al., Proteomics 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 5 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 DiffusionMSI + gradient 0.824 35.52 0.971
2 MSIFormer + gradient 0.799 34.04 0.961
3 SpaMSI-Net + gradient 0.777 31.84 0.941
4 DeepMSI + gradient 0.769 31.06 0.932
5 MSI-GAN + gradient 0.764 31.39 0.936
6 MSI-TV + gradient 0.709 27.51 0.87
7 MSI-NMF + gradient 0.659 25.09 0.805
8 MSI-PCA + gradient 0.585 22.48 0.71
9 MSI-Hotelling + gradient 0.542 20.44 0.62
Spec Ranges (4 parameters)
Parameter Min Max Unit
spray_angle_error -1.0 2.0 deg
solvent_flow_variation -3.0 6.0 -
ion_suppression_(matrix_effect) -10.0 20.0 -
spatial_resolution_degradation -10.0 20.0 -
View Details →
Dev 5 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 MSIFormer + gradient 0.775 32.55 0.948
2 DiffusionMSI + gradient 0.756 31.11 0.932
3 MSI-GAN + gradient 0.677 26.42 0.843
4 DeepMSI + gradient 0.630 24.87 0.798
5 SpaMSI-Net + gradient 0.628 24.59 0.789
6 MSI-NMF + gradient 0.607 23.58 0.753
7 MSI-TV + gradient 0.600 23.37 0.745
8 MSI-PCA + gradient 0.585 22.93 0.728
9 MSI-Hotelling + gradient 0.516 20.22 0.609
Spec Ranges (4 parameters)
Parameter Min Max Unit
spray_angle_error -1.2 1.8 deg
solvent_flow_variation -3.6 5.4 -
ion_suppression_(matrix_effect) -12.0 18.0 -
spatial_resolution_degradation -12.0 18.0 -
Submit Reconstruction View Details →
Hidden 5 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 MSIFormer + gradient 0.744 30.72 0.927
2 DiffusionMSI + gradient 0.723 30.4 0.923
3 MSI-GAN + gradient 0.656 25.97 0.831
4 MSI-NMF + gradient 0.548 21.83 0.683
5 MSI-PCA + gradient 0.542 21.2 0.655
6 DeepMSI + gradient 0.535 21.03 0.647
7 SpaMSI-Net + gradient 0.528 21.33 0.661
8 MSI-TV + gradient 0.523 20.34 0.615
9 MSI-Hotelling + gradient 0.438 17.59 0.48
Spec Ranges (4 parameters)
Parameter Min Max Unit
spray_angle_error -0.7 2.3 deg
solvent_flow_variation -2.1 6.9 -
ion_suppression_(matrix_effect) -7.0 23.0 -
spatial_resolution_degradation -7.0 23.0 -
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

S → D

S Sampling
D Detector

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
s_a spray_angle_error Spray angle error (deg) 0.0 1.0
s_f solvent_flow_variation Solvent flow variation (-) 0.0 3.0
i_s ion_suppression_(matrix_effect) Ion suppression (matrix effect) (-) 0.0 10.0
s_r spatial_resolution_degradation Spatial resolution degradation (-) 0.0 10.0

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.