ECG features and methods for automatic classification of ventricular premature and ischemic heartbeats: A...

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Accurate detection of cardiac pathological events via electrocardiogram (ECG) is crucial for patient treatment.
• Previous research has focused on individual heartbeat classifications, but a combined approach for non-ischemic, ischemic, and ventricular premature beats is less explored.

Purpose of the Study:

• To develop and evaluate an automated system for classifying various heartbeat types, including non-ischemic, ischemic (two grades), and ventricular premature beats.
• To compare the effectiveness of morphological and spectral features for heartbeat classification.
• To identify optimal machine learning models for accurate ECG analysis.

Main Methods:

• Analysis of ECGs from isolated rabbit hearts under non-ischemic and ischemic conditions.
• Extraction and testing of various commonly used and novel morphological and spectral features.
• Implementation and comparison of different classification models, including k-nearest neighbors and support vector machines.

Main Results:

• Morphological features demonstrated superior performance compared to spectral features for heartbeat classification.
• High classification accuracies were achieved (up to 98.3% for morphological and 96.2% for spectral features), even without QRS-T segment delineation.
• Reduced feature sets (3-14 features) yielded comparable or improved performance over larger sets (10-29 features).
• K-nearest neighbors (up to 98.6% accuracy) and support vector machines (up to 93.5% accuracy) were identified as the most effective models.

Conclusions:

• Automated classification of complex cardiac arrhythmias is feasible with high accuracy.
• Morphological features and machine learning models offer an efficient approach to ECG analysis, potentially simplifying clinical workflows.
• The study provides a foundation for developing more sophisticated automated diagnostic tools for cardiac conditions.