SPECT

Single Photon Emission Computed Tomography

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
🥇 PET-ViT 0.876 38.08 0.982 ✓ Certified Smith et al., ICCV 2024
🥈 PETFormer 0.873 37.9 0.982 ✓ Certified Li et al., ECCV 2024
🥉 U-Net-PET 0.794 33.86 0.960 ✓ Certified Ronneberger et al. variant, MICCAI 2020
4 TransEM 0.781 33.7 0.938 ✓ Certified Xie et al., 2023
5 DeepPET 0.749 32.4 0.918 ✓ Certified Haggstrom et al., MIA 2019
6 FBP-PET 0.711 30.1 0.918 ✓ Certified Analytical baseline
7 ML-EM 0.694 29.4 0.907 ✓ Certified Shepp & Vardi, IEEE TPAMI 1982
8 OS-EM 0.656 27.96 0.880 ✓ Certified Hudson & Larkin, IEEE TMI 1994
9 MAPEM-RDP 0.632 28.5 0.815 ✓ Certified Nuyts et al., 2002
10 OSEM 0.508 24.8 0.690 ✓ Certified Hudson & Larkin, IEEE TMI 1994

Dataset: PWM Benchmark (10 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
🥇 PET-ViT + gradient 0.785
0.844
36.67 dB / 0.977
0.781
33.6 dB / 0.958
0.730
29.72 dB / 0.912
✓ Certified Smith et al., ICCV 2024
🥈 PETFormer + gradient 0.762
0.818
34.92 dB / 0.967
0.764
31.25 dB / 0.934
0.704
27.87 dB / 0.878
✓ Certified Li et al., ECCV 2024
🥉 TransEM + gradient 0.696
0.786
31.96 dB / 0.942
0.691
27.83 dB / 0.877
0.612
23.77 dB / 0.760
✓ Certified Xie et al., 2023
4 FBP-PET + gradient 0.691
0.700
27.23 dB / 0.864
0.693
27.21 dB / 0.863
0.680
26.63 dB / 0.849
✓ Certified Analytical baseline
5 DeepPET + gradient 0.680
0.770
31.09 dB / 0.932
0.653
25.36 dB / 0.813
0.618
24.59 dB / 0.789
✓ Certified Haggstrom et al., MIA 2019
6 ML-EM + gradient 0.669
0.692
27.14 dB / 0.862
0.683
27.06 dB / 0.860
0.633
24.65 dB / 0.791
✓ Certified Shepp & Vardi, IEEE TPAMI 1982
7 U-Net-PET + gradient 0.668
0.768
31.53 dB / 0.937
0.630
25.11 dB / 0.806
0.606
23.76 dB / 0.760
✓ Certified Ronneberger et al. variant, MICCAI 2020
8 OS-EM + gradient 0.634
0.692
26.65 dB / 0.849
0.621
23.91 dB / 0.765
0.588
23.38 dB / 0.746
✓ Certified Hudson & Larkin, IEEE TMI 1994
9 MAPEM-RDP + gradient 0.633
0.669
25.71 dB / 0.824
0.643
25.08 dB / 0.805
0.588
23.44 dB / 0.748
✓ Certified Nuyts et al., IEEE TMI 2002
10 OSEM + gradient 0.546
0.578
22.08 dB / 0.693
0.553
21.81 dB / 0.682
0.508
20.52 dB / 0.623
✓ Certified Hudson & Larkin, IEEE TMI 1994

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 PET-ViT + gradient 0.844 36.67 0.977
2 PETFormer + gradient 0.818 34.92 0.967
3 TransEM + gradient 0.786 31.96 0.942
4 DeepPET + gradient 0.770 31.09 0.932
5 U-Net-PET + gradient 0.768 31.53 0.937
6 FBP-PET + gradient 0.700 27.23 0.864
7 ML-EM + gradient 0.692 27.14 0.862
8 OS-EM + gradient 0.692 26.65 0.849
9 MAPEM-RDP + gradient 0.669 25.71 0.824
10 OSEM + gradient 0.578 22.08 0.693
Spec Ranges (4 parameters)
Parameter Min Max Unit
center_offset -1.5 3.0 pixels
collimator_septal -0.02 0.04
attenuation -5.0 10.0 %
scatter 0.15 0.3
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 PET-ViT + gradient 0.781 33.6 0.958
2 PETFormer + gradient 0.764 31.25 0.934
3 FBP-PET + gradient 0.693 27.21 0.863
4 TransEM + gradient 0.691 27.83 0.877
5 ML-EM + gradient 0.683 27.06 0.86
6 DeepPET + gradient 0.653 25.36 0.813
7 MAPEM-RDP + gradient 0.643 25.08 0.805
8 U-Net-PET + gradient 0.630 25.11 0.806
9 OS-EM + gradient 0.621 23.91 0.765
10 OSEM + gradient 0.553 21.81 0.682
Spec Ranges (4 parameters)
Parameter Min Max Unit
center_offset -1.8 2.7 pixels
collimator_septal -0.024 0.036
attenuation -6.0 9.0 %
scatter 0.14 0.29
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 PET-ViT + gradient 0.730 29.72 0.912
2 PETFormer + gradient 0.704 27.87 0.878
3 FBP-PET + gradient 0.680 26.63 0.849
4 ML-EM + gradient 0.633 24.65 0.791
5 DeepPET + gradient 0.618 24.59 0.789
6 TransEM + gradient 0.612 23.77 0.76
7 U-Net-PET + gradient 0.606 23.76 0.76
8 OS-EM + gradient 0.588 23.38 0.746
9 MAPEM-RDP + gradient 0.588 23.44 0.748
10 OSEM + gradient 0.508 20.52 0.623
Spec Ranges (4 parameters)
Parameter Min Max Unit
center_offset -1.05 3.45 pixels
collimator_septal -0.014 0.046
attenuation -3.5 11.5 %
scatter 0.165 0.315
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

About the Imaging Modality

SPECT images the 3D distribution of a gamma-emitting radiotracer (e.g. 99mTc-sestamibi) by detecting single photons with rotating gamma cameras equipped with parallel-hole collimators. The collimator creates a projection of the activity distribution, and multiple angles enable tomographic reconstruction. The forward model includes collimator response (depth-dependent blurring), photon attenuation, and scatter. Reconstruction uses OSEM with corrections for attenuation (AC), scatter (SC), and resolution recovery (RR).

Principle

Single Photon Emission Computed Tomography detects single gamma-ray photons emitted by a radiotracer (⁹⁹ᵐTc, ¹²³I, ²⁰¹Tl) using a rotating gamma camera with a parallel-hole or pinhole collimator. The collimator provides directional sensitivity at the cost of low geometric efficiency (~0.01 %). Projections from multiple angles are reconstructed into 3-D activity maps.

How to Build the System

A dual-head gamma camera (e.g., Siemens Symbia, GE Discovery) with NaI(Tl) scintillator crystals (9.5 mm thick) and parallel-hole collimators rotates around the patient (typically 60-128 angular stops over 360°). For cardiac SPECT, use dedicated CZT-based cameras with pinhole or multi-pinhole collimators. Acquire in step-and-shoot or continuous rotation mode. Energy windows are set around the photopeak (e.g., 140 keV ± 10 % for ⁹⁹ᵐTc).

Common Reconstruction Algorithms

  • FBP with ramp-Butterworth filter
  • OSEM with attenuation and scatter correction
  • Resolution recovery (collimator-detector response modeling in OSEM)
  • CT-based attenuation correction (SPECT/CT)
  • Deep-learning SPECT reconstruction (dose reduction, resolution enhancement)

Common Mistakes

  • Insufficient count statistics causing noisy, unreliable reconstructions
  • Not correcting for depth-dependent collimator blur (resolution degrades with distance)
  • Attenuation artifacts in uncorrected SPECT (false defects in myocardial perfusion)
  • Patient motion during the long SPECT acquisition (15-30 minutes)
  • Incorrect energy window or scatter window setup leading to poor image quality

How to Avoid Mistakes

  • Ensure adequate injected dose and acquisition time for sufficient count statistics
  • Use resolution recovery (distance-dependent PSF modeling) in iterative reconstruction
  • Apply CT-based attenuation correction; verify CT-SPECT registration
  • Use motion detection and correction algorithms; shorter acquisitions with CZT cameras
  • Verify energy window settings match the radionuclide photopeak and scatter windows

Forward-Model Mismatch Cases

  • The widefield fallback produces a blurred (64,64) image, but SPECT acquires projections of shape (n_angles, n_detectors) using a rotating gamma camera with collimator — output shape (32,64) vs (64,64)
  • SPECT measurement involves collimated gamma-ray detection with depth-dependent spatial resolution (the collimator PSF broadens with distance) — the widefield spatially-invariant Gaussian blur cannot model this depth-dependent response

How to Correct the Mismatch

  • Use the SPECT operator that models collimated gamma-ray projection with distance-dependent resolution: y(theta,s) = integral of (h(d) * f) along projection rays for each angle
  • Reconstruct using OSEM with depth-dependent collimator-detector response modeling and attenuation correction (Chang method or CT-based mu-map)

Experimental Setup — Signal Chain

Experimental setup diagram for Single Photon Emission Computed Tomography

Experimental Setup

Instrument: Siemens Symbia Intevo / GE NM/CT 870 CZT
Matrix Size: 64x64
Projections: 64
Reconstruction: OSEM (AC+SC+RR)
Iterations: 8
Subsets: 8
Post Filter Fwhm Mm: 8.0
Isotope: 99mTc-sestamibi
Application: myocardial perfusion imaging
Acquisition Time Per View S: 20

Key References

  • Hudson & Larkin, 'Accelerated image reconstruction using ordered subsets of projection data (OSEM)', IEEE TMI 13, 601-609 (1994)

Canonical Datasets

  • Clinical SPECT benchmark collections

Spec DAG — Forward Model Pipeline

Π(parallel) → Σ_E → D(g, η₃)

Π Parallel-Hole Collimator (parallel)
Σ Energy Window Sum (E)
D Gamma Camera (g, η₃)

Mismatch Parameters

Symbol Parameter Description Nominal Perturbed
Δc center_offset Center-of-rotation offset (pixels) 0 1.5
s collimator_septal Septal penetration fraction 0 0.02
μ attenuation Attenuation coefficient error (%) 0 5.0
f_s scatter Scatter fraction error 0.2 0.25

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.