Automated Classification of Cardiac Arrhythmia using Short-Duration ECG Signals and Machine Learning.

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Accurate cardiac arrhythmia detection is vital for preventing sudden cardiac death.
• Existing methods for electrocardiogram (ECG) signal analysis face challenges with noise and complex arrhythmias.
• Automated classification systems are needed to improve diagnostic efficiency in clinical settings.

Purpose of the Study:

• To develop and evaluate a robust automated system for classifying five types of cardiac arrhythmias.
• To assess the efficacy of signal preprocessing techniques and machine learning classifiers for arrhythmia detection.

Main Methods:

• ECG signals were preprocessed using a dual-stage Discrete Wavelet Transform (DWT) and a median filter to remove noise.
• QRS regions were extracted after signal segmentation for feature extraction.
• Nine temporal features were extracted and used to train six different machine learning classifiers, including Support Vector Machine (SVM) and Ensemble Tree.

Main Results:

• The Support Vector Machine (SVM) and Ensemble Tree classifiers showed superior performance in classifying normal beats and four types of arrhythmias (PVC, PAC, R-BBB, L-BBB).
• The SVM classifier, utilizing a Gaussian kernel, achieved high performance metrics: 97.44% sensitivity, 99.36% specificity, 97.44% positive predictive value, and 98.97% accuracy.
• The MIT-BIH Arrhythmia Database (MIT-BIH AD) was used for performance evaluation.

Conclusions:

• The proposed integrated approach of signal preprocessing and feature extraction effectively classifies cardiac arrhythmias.
• The high performance of the SVM classifier indicates its potential for reliable automated cardiac arrhythmia detection in clinical applications and trials.