Related Experiment Video
Updated: Aug 16, 2026
Retrospective Cardiac Gating with A Prototype Small-Animal X-ray Computed Tomograph
Published on: February 21, 2025
A Deep Learning Approach for Noise Suppression in Cardiac-gated SPECT Studies
Xirang Zhang1, Yongyi Yang2, Jovan G Brankov2
1Medical Imaging Research Center and Department of Electrical and Computer Engineering, Illinois Institute of Technology, Chicago, IL, USA; School of Artificial Intelligence and Robotics, Hunan University of Technology and Business, Changsha, Hunan, China.
Deep learning effectively reduces noise in cardiac-gated SPECT images, improving image quality without altering left ventricular (LV) function measurements. This noise suppression technique shows promise across different imaging systems.
Area of Science:
- Medical Imaging
- Artificial Intelligence in Medicine
- Nuclear Cardiology
Background:
- Cardiac-gated SPECT imaging often suffers from high noise levels in clinical settings.
- This noise can impact the diagnostic accuracy of myocardial and left ventricular (LV) assessments.
Purpose of the Study:
- To investigate the efficacy of a deep learning (DL) network for noise reduction in electrocardiogram-gated cardiac SPECT.
- To evaluate the impact of DL denoising on image quality and LV functional parameters.
Main Methods:
- A 3D convolutional autoencoder DL network, pre-trained on ungated SPECT data, was adapted for cardiac-gated images.
- The network was trained on 862 ungated studies and tested on 50 cardiac-gated studies from a Philips BrightView SPECT/CT system.
- Performance was assessed by quantifying noise levels and LV functional measures, with validation on a GE StarGuide SPECT/CT system.
Main Results:
- DL denoising improved LV wall uniformity by 20.6% and reduced myocardial temporal variability by 47.4%.
- High agreement (Pearson's r > 0.94) was observed for LV functional measures (wall thickness, volumes, ejection fraction) before and after denoising.
- Consistent performance was noted across different SPECT/CT systems.
Conclusions:
- The proposed DL approach effectively suppresses noise in cardiac-gated SPECT images.
- LV function measurements remain accurate after DL denoising, showing no significant distortion.
- The DL method demonstrates potential for broad applicability across various cardiac SPECT imaging platforms.