Normal and Abnormal Classification of Electrocardiogram: A Primary Screening Tool Kit

Area of Science:

• Biomedical Engineering
• Computational Cardiology
• Machine Learning in Healthcare

Background:

• Cardiovascular diseases (CVDs) are a leading cause of mortality globally.
• Cardiac arrhythmia, a significant CVD, is detectable via electrocardiogram (ECG).
• Automated ECG analysis offers potential for early arrhythmia identification and prevention of sudden death.

Purpose of the Study:

• To present a simple, automated model for classifying ECG recordings into normal and abnormal categories.
• To evaluate the efficacy of various machine learning classifiers and feature selection algorithms for ECG analysis.
• To provide a tool for mass screening and primary detection of cardiac arrhythmias.

Main Methods:

• Signal quality analysis (SQA) was performed to exclude poor-quality ECG signals.
• Morphological and heart rate variability (HRV) features were extracted from ECG recordings.
• Multiple machine learning classifiers (SVM, Adaboost, RF, ET, DT, ANN, KNN, LR, NB, GB) were explored.
• Feature selection algorithms (F-test, LASSO, mRMR) were applied to optimize feature space.
• The model was validated on a dataset of 2648 normal and 2518 abnormal ECG recordings.

Main Results:

• The study explored the performance of ten different machine learning classifiers on extracted ECG features.
• Feature selection algorithms were utilized to enhance classification accuracy.
• The best-performing classifier achieved an accuracy of 95.25% in distinguishing normal from abnormal ECG recordings.
• Comparative analysis of classifiers and feature selection methods was conducted.

Conclusions:

• The developed automated ECG analysis model demonstrates high accuracy in classifying normal and abnormal heart rhythms.
• The proposed model shows promise as a cost-effective tool for mass screening and preliminary diagnosis of cardiac arrhythmias in clinical settings.
• Integration of signal quality analysis and advanced machine learning techniques is crucial for reliable automated ECG interpretation.