Interpretable Assessment of ST-Segment Deviation in ECG Time Series

Area of Science:

• Biomedical Engineering
• Cardiology
• Machine Learning in Healthcare

Background:

• Cardiovascular diseases (CVD) remain a leading cause of mortality despite medical advancements.
• ST-segment deviation on electrocardiography (ECG) is a key indicator of myocardial infarction (MI), often diagnosed manually.
• Accurate and automated detection of ST-segment deviation is needed for timely diagnosis.

Purpose of the Study:

• To develop and validate an automated methodology for detecting and quantifying ST-segment deviation in single-lead ECG.
• To apply automated machine learning for optimizing classification and regression models for ST-segment deviation analysis.
• To assess the performance of the proposed method using a standard cardiovascular database.

Main Methods:

• Extraction of statistical features, point-to-point beat characteristics, and signal quality indexes (SQIs) from single-lead ECG.
• Application of automated machine learning (AutoML) to identify optimal hyperparameters for predictive models.
• Utilizing the Physionet ST-T database for method validation.

Main Results:

• The proposed automated method achieved 98.30% accuracy for multiclass classification of ST-segment deviation.
• The method demonstrated 99.87% accuracy for binary classification (presence or absence of deviation).
• Quantification of ST-segment deviation scale up to 1 mV was achieved.

Conclusions:

• Automated detection and quantification of ST-segment deviation using machine learning is highly accurate.
• This methodology offers a promising tool for objective and efficient myocardial infarction diagnosis.
• The approach has the potential to enhance cardiovascular disease screening and management.