Related Experiment Video
Updated: Jul 17, 2025
11:09
High-Resolution Cardiac Positron Emission Tomography/Computed Tomography for Small Animals
Published on: December 16, 2022
3.7K
Deep learning approximation of attenuation maps for myocardial perfusion SPECT with an IQ SPECT collimator
Tamino Huxohl1, Gopesh Patel1, Reinhard Zabel2
1Institute of Radiology, Nuclear Medicine and Molecular Imaging, Heart and Diabetes Center North Rhine-Westphalia, University Hospital of the Ruhr University Bochum, Bad Oeynhausen, Germany.
EJNMMI Physics
|August 28, 2023
Summary
Deep learning can approximate attenuation maps for myocardial perfusion imaging using SPECT scanners with IQSPECT collimators. This method enhances diagnostic confidence without requiring CT scans, making it more accessible.
Area of Science:
- Medical Imaging
- Nuclear Medicine
- Artificial Intelligence
Background:
- CT imaging aids SPECT myocardial perfusion imaging but is often unavailable.
- SPECT-only scanners limit attenuation correction capabilities.
- Deep learning shows promise for approximating attenuation maps.
Purpose of the Study:
- Investigate deep learning for attenuation map approximation from non-attenuation-corrected SPECT data.
- Evaluate feasibility on SPECT scanners with IQSPECT collimators.
Main Methods:
- Retrospective acquisition of 150 SPECT studies.
- U-Net model trained using conditional generative adversarial networks.
- Comparison of predicted vs. real attenuation maps (NMAE) and polar maps (APE).
Main Results:
- Predicted attenuation maps closely resemble real maps (NMAE: 0.020±0.007).
- Polar maps derived from predicted attenuation maps show high similarity to CT-based maps.
- Low pixel-wise (3.095±3.199) and segment-wise (1.155±0.769) errors achieved.
Conclusions:
- Deep learning effectively approximates attenuation maps from non-attenuation-corrected SPECT reconstructions.
- Feasible for SPECT scanners equipped with IQSPECT collimators.
- Potential to improve diagnostic confidence in myocardial perfusion imaging.
