MALDI 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
🥇	CalibFormer CalibFormer Transformer calibration, 2024 32.8 dB SSIM 0.920 Checkpoint unavailable	0.757	32.8	0.920	✓ Certified	Transformer calibration, 2024
🥈	MassSpecFormer MassSpecFormer Mass spectrometry transformer, 2024 30.55 dB SSIM 0.925 Checkpoint unavailable	0.722	30.55	0.925	✓ Certified	Mass spectrometry transformer, 2024
🥉	DiffusionInstrumentation DiffusionInstrumentation Zhang et al., 2024 30.54 dB SSIM 0.925 Checkpoint unavailable	0.722	30.54	0.925	✓ Certified	Zhang et al., 2024
4	ResNet-Calib ResNet-Calib ResNet for calibration, 2022 31.3 dB SSIM 0.892 Checkpoint unavailable	0.718	31.3	0.892	✓ Certified	ResNet for calibration, 2022
5	Instrument-CNN Instrument-CNN Instrument-specific CNN 29.61 dB SSIM 0.911 Checkpoint unavailable	0.699	29.61	0.911	✓ Certified	Instrument-specific CNN
6	ScoreInstrumentation ScoreInstrumentation Wei et al., 2025 29.06 dB SSIM 0.901 Checkpoint unavailable	0.685	29.06	0.901	✓ Certified	Wei et al., 2025
7	PnP-NLM PnP-NLM Non-local means filter 26.67 dB SSIM 0.850 Try in SpecLab →	0.620	26.67	0.850	✓ Certified	Non-local means filter
8	PnP-BM3D PnP-BM3D Danielyan et al., 2012 27.6 dB SSIM 0.790 Try in SpecLab →	0.605	27.6	0.790	✓ Certified	Danielyan et al., 2012
9	Peak Fitting Peak Fitting Gaussian peak fitting 25.91 dB SSIM 0.829 Try in SpecLab →	0.596	25.91	0.829	✓ Certified	Gaussian peak fitting
10	Calibration-Lookup Calibration-Lookup Look-up table calibration 24.12 dB SSIM 0.773 Try in SpecLab →	0.538	24.12	0.773	✓ Certified	Look-up table calibration
11	Deconv Deconv Analytical baseline 24.1 dB SSIM 0.660 Try in SpecLab →	0.482	24.1	0.660	✓ Certified	Analytical baseline

Dataset: PWM Benchmark (11 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
🥇	CalibFormer + gradient CalibFormer + gradient Transformer calibration, 2024 Score 0.727 Correct & Reconstruct →	0.727	0.775 31.14 dB / 0.933	0.733 29.98 dB / 0.917	0.672 26.51 dB / 0.846	✓ Certified	Transformer calibration, 2024
🥈	MassSpecFormer + gradient MassSpecFormer + gradient Mass spectrometry transformer, 2024 Score 0.595 Correct & Reconstruct →	0.595	0.714 28.32 dB / 0.887	0.610 24.1 dB / 0.772	0.461 18.45 dB / 0.522	✓ Certified	Mass spectrometry transformer, 2024
🥉	Peak Fitting + gradient Peak Fitting + gradient Gaussian peak fitting Score 0.589 Correct & Reconstruct →	0.589	0.643 24.24 dB / 0.777	0.570 21.85 dB / 0.684	0.553 21.73 dB / 0.678	✓ Certified	Gaussian peak fitting
4	ResNet-Calib + gradient ResNet-Calib + gradient ResNet for calibration, 2022 Score 0.587 Correct & Reconstruct →	0.587	0.727 28.91 dB / 0.899	0.550 21.79 dB / 0.681	0.485 19.16 dB / 0.558	✓ Certified	ResNet for calibration, 2022
5	PnP-BM3D + gradient PnP-BM3D + gradient Danielyan et al., 2012 Score 0.569 Correct & Reconstruct →	0.569	0.685 26.41 dB / 0.843	0.530 20.55 dB / 0.625	0.492 19.75 dB / 0.587	✓ Certified	Danielyan et al., 2012
6	DiffusionInstrumentation + gradient DiffusionInstrumentation + gradient Zhang et al., 2024 Score 0.549 Correct & Reconstruct →	0.549	0.712 28.12 dB / 0.883	0.506 20.23 dB / 0.610	0.429 17.17 dB / 0.459	✓ Certified	Zhang et al., 2024
7	PnP-NLM + gradient PnP-NLM + gradient Non-local means filter Score 0.546 Correct & Reconstruct →	0.546	0.660 24.96 dB / 0.801	0.539 21.32 dB / 0.660	0.438 17.6 dB / 0.480	✓ Certified	Non-local means filter
8	Calibration-Lookup + gradient Calibration-Lookup + gradient Look-up table calibration Score 0.545 Correct & Reconstruct →	0.545	0.597 22.42 dB / 0.708	0.527 20.59 dB / 0.627	0.510 20.7 dB / 0.632	✓ Certified	Look-up table calibration
9	Deconv + gradient Deconv + gradient Analytical baseline Score 0.538 Correct & Reconstruct →	0.538	0.557 21.25 dB / 0.657	0.525 20.97 dB / 0.644	0.531 21.01 dB / 0.646	✓ Certified	Analytical baseline
10	Instrument-CNN + gradient Instrument-CNN + gradient Instrument-specific CNN Score 0.534 Correct & Reconstruct →	0.534	0.722 28.1 dB / 0.883	0.489 19.52 dB / 0.575	0.392 16.05 dB / 0.404	✓ Certified	Instrument-specific CNN
11	ScoreInstrumentation + gradient ScoreInstrumentation + gradient Wei et al., 2025 Score 0.521 Correct & Reconstruct →	0.521	0.682 26.55 dB / 0.847	0.491 19.51 dB / 0.575	0.391 16.55 dB / 0.428	✓ Certified	Wei et al., 2025

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	CalibFormer + gradient	0.775	31.14	0.933
2	ResNet-Calib + gradient	0.727	28.91	0.899
3	Instrument-CNN + gradient	0.722	28.1	0.883
4	MassSpecFormer + gradient	0.714	28.32	0.887
5	DiffusionInstrumentation + gradient	0.712	28.12	0.883
6	PnP-BM3D + gradient	0.685	26.41	0.843
7	ScoreInstrumentation + gradient	0.682	26.55	0.847
8	PnP-NLM + gradient	0.660	24.96	0.801
9	Peak Fitting + gradient	0.643	24.24	0.777
10	Calibration-Lookup + gradient	0.597	22.42	0.708
11	Deconv + gradient	0.557	21.25	0.657

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
laser_fluence_drift	0.96	1.08	-
mass_accuracy	-1.0	2.0	ppm
extraction_delay	96.0	108.0	ns
matrix_crystallization	0.94	1.12	-

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	CalibFormer + gradient	0.733	29.98	0.917
2	MassSpecFormer + gradient	0.610	24.1	0.772
3	Peak Fitting + gradient	0.570	21.85	0.684
4	ResNet-Calib + gradient	0.550	21.79	0.681
5	PnP-NLM + gradient	0.539	21.32	0.66
6	PnP-BM3D + gradient	0.530	20.55	0.625
7	Calibration-Lookup + gradient	0.527	20.59	0.627
8	Deconv + gradient	0.525	20.97	0.644
9	DiffusionInstrumentation + gradient	0.506	20.23	0.61
10	ScoreInstrumentation + gradient	0.491	19.51	0.575
11	Instrument-CNN + gradient	0.489	19.52	0.575

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
laser_fluence_drift	0.952	1.072	-
mass_accuracy	-1.2	1.8	ppm
extraction_delay	95.2	107.2	ns
matrix_crystallization	0.928	1.108	-

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	CalibFormer + gradient	0.672	26.51	0.846
2	Peak Fitting + gradient	0.553	21.73	0.678
3	Deconv + gradient	0.531	21.01	0.646
4	Calibration-Lookup + gradient	0.510	20.7	0.632
5	PnP-BM3D + gradient	0.492	19.75	0.587
6	ResNet-Calib + gradient	0.485	19.16	0.558
7	MassSpecFormer + gradient	0.461	18.45	0.522
8	PnP-NLM + gradient	0.438	17.6	0.48
9	DiffusionInstrumentation + gradient	0.429	17.17	0.459
10	Instrument-CNN + gradient	0.392	16.05	0.404
11	ScoreInstrumentation + gradient	0.391	16.55	0.428

Spec Ranges (4 parameters)

Parameter	Min	Max	Unit
laser_fluence_drift	0.972	1.092	-
mass_accuracy	-0.7	2.3	ppm
extraction_delay	97.2	109.2	ns
matrix_crystallization	0.958	1.138	-

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
l_f	laser_fluence_drift	Laser fluence drift (-)	1.0	1.04
m_a	mass_accuracy	Mass accuracy (ppm)	0.0	1.0
e_d	extraction_delay	Extraction delay (ns)	100.0	104.0
m_c	matrix_crystallization	Matrix crystallization (-)	1.0	1.06

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.