Myocardial Ischemia Detection Using Body Surface Potential Mappings and Machine Learning.

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Acute myocardial ischemia detection is crucial for timely intervention.
• Traditional electrocardiograms (ECG) have limitations in ischemia detection.
• Body surface potential mapping (BSPM) offers a more comprehensive, noninvasive approach.

Purpose of the Study:

• To develop and evaluate machine learning models for acute myocardial ischemia detection using BSPM data.
• To compare the performance of logistic regression and XGBoost classifiers.
• To identify key electrodes for efficient ischemia detection.

Main Methods:

• Acquisition of experimental BSPM data with ground truth ischemia measurements.
• Development and training of logistic regression and XGBoost models.
• Visualization and ranking of electrode contributions.
• Retraining models with reduced electrode subsets.

Main Results:

• XGBoost classifier achieved 97.63% mean accuracy and 0.9972 mean AUC.
• Logistic regression achieved 96.46% mean accuracy and 0.9927 mean AUC.
• Models trained on subsets of top electrodes maintained high performance with reduced training times.

Conclusions:

• Machine learning models applied to BSPM data show high efficacy in detecting acute myocardial ischemia.
• BSPM-based models offer superior performance compared to traditional ECG methods.
• Reduced electrode sets can provide efficient and accurate ischemia detection, enabling faster diagnoses.