Related Experiment Video
Updated: Aug 22, 2025
11:09
High-Resolution Cardiac Positron Emission Tomography/Computed Tomography for Small Animals
Published on: December 16, 2022
3.8K
Myocardial Perfusion SPECT Imaging Radiomic Features and Machine Learning Algorithms for Cardiac Contractile Pattern
Maziar Sabouri1,2, Ghasem Hajianfar2, Zahra Hosseini2
1Department of Medical Physics, School of Medicine, Iran University of Medical Science, Tehran, Iran.
Journal of Digital Imaging
|November 15, 2022
Summary
Machine learning models accurately identified left ventricular contractile patterns using GSPECT MPI data. This approach aids in predicting Cardiac Resynchronization Therapy (CRT) response by analyzing conventional quantitative features and radiomic features.
Area of Science:
- Cardiology
- Medical Imaging
- Machine Learning
Background:
- A U-shaped left ventricular contraction pattern is linked to better Cardiac Resynchronization Therapy (CRT) outcomes.
- Accurate identification of contractile patterns is crucial for predicting CRT response.
Purpose of the Study:
- To develop and evaluate machine learning models for automatically recognizing left ventricular contractile patterns.
- To assess the utility of conventional quantitative features (ConQuaFea) and radiomic features from GSPECT MPI for this task.
Main Methods:
- Utilized Gated single-photon emission computed tomography myocardial perfusion imaging (GSPECT MPI) data from 98 patients.
- Extracted and selected features using Recursive Feature Elimination (RFE).
- Trained seven different machine learning classifiers on ConQuaFea, radiomics, and combined feature sets.
Main Results:
- The MLP classifier showed strong performance (AUC 0.80) with ConQuaFea.
- RF achieved the best performance (AUC 0.65) with radiomic features alone.
- Combined models, particularly Gradient Boosting (GB) and Random Forest (RF), demonstrated high predictive power (AUC up to 0.78).
Conclusions:
- Machine learning models integrating ConQuaFea and radiomic features from GSPECT MPI show promise for detecting left ventricular contractile patterns.
- This approach could enhance the prediction of CRT response.
