Machine Learning Approach on High Risk Treadmill Exercise Test to Predict Obstructive Coronary Artery Disease by...

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Treadmill Exercise Tests (TET) guide decisions for invasive Coronary Angiography (CAG), but high false positive rates can lead to unnecessary procedures.
• Optimizing non-invasive diagnostic accuracy is crucial to improve patient outcomes and reduce healthcare costs.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for enhancing TET performance using electrocardiography (ECG) signal characteristics.
• To compare the diagnostic capabilities of ML models against cardiologists' interpretations of ECG data.

Main Methods:

• A dataset of 294 patients undergoing CAG after high-risk TET was analyzed.
• ECG signals were processed into time-series data, extracting P, QRS, and T wave features.
• Five ML algorithms were trained and validated using 5-fold cross-validation, with performance compared to cardiologists' V5 signal analysis.

Main Results:

• ML models demonstrated significantly superior performance compared to cardiologists' interpretation of the V5 signal (P < 0.0001).
• The XGBoost model achieved the highest accuracy (80.92±6.42%) and Area Under the Curve (AUC) (0.78±0.06).
• ML models effectively diagnosed critical coronary artery disease (CAD) using only V5 ECG signal markers, without requiring clinical data from TET reports.

Conclusions:

• Machine learning models offer a powerful, non-invasive approach to improve the accuracy of cardiac diagnostics.
• Utilizing ML models based on ECG signal markers can optimize TET interpretation, potentially reducing the need for invasive CAG procedures.