Utilizing ECG-Based Heartbeat Classification for Hypertrophic Cardiomyopathy Identification.

Area of Science:

• Cardiology
• Biomedical Signal Processing
• Machine Learning

Background:

• Hypertrophic cardiomyopathy (HCM) is a serious heart condition characterized by thickened heart muscle and potential blood flow obstruction.
• Electrocardiograms (ECG) offer a non-invasive method for assessing heart electrical activity and can aid in HCM detection.

Purpose of the Study:

• To develop and evaluate a cardiovascular-patient classifier for identifying hypertrophic cardiomyopathy (HCM) using 12-lead ECG signals.
• To assess the efficacy of machine learning classifiers in distinguishing HCM patients based on ECG-derived heartbeat characteristics.

Main Methods:

• Extracted 504 morphological and temporal features from 10-second, 12-lead ECG signals for heartbeat classification.
• Trained and evaluated random forest (RF) and support vector machine (SVM) classifiers using 5-fold cross-validation.
• Compared RF and SVM performance against a logistic regression baseline.

Main Results:

• Both RF and SVM classifiers demonstrated high performance, with precision near 0.85 and recall/specificity around 0.90.
• Feature selection experiments revealed that a reduced set of 264 features achieved comparable performance to the full feature set.
• RF and SVM outperformed logistic regression in classifying HCM patients.

Conclusions:

• Machine learning classifiers, particularly RF and SVM, are effective tools for detecting hypertrophic cardiomyopathy using ECG data.
• A subset of highly informative features can maintain high classification accuracy, suggesting potential for streamlined diagnostic tools.
• This ECG-based approach shows promise for the early and accurate identification of HCM patients.