Automated screening of arrhythmia using wavelet based machine learning techniques.

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Arrhythmia is a prevalent cardiac condition requiring early detection for preventive care.
• Screening for arrhythmias necessitates robust and accurate diagnostic tools.

Purpose of the Study:

• To develop and evaluate a system for automated arrhythmia screening from electrocardiogram (ECG) signals.
• To compare the performance of different machine learning algorithms for ECG-based arrhythmia classification.

Main Methods:

• Feature extraction from ECG signals using Discrete Wavelet Transform (DWT).
• R-point detection via the Pan Tompkins algorithm for signal registration.
• Dimensionality reduction using Principal Component Analysis (PCA) and statistical validation with t-tests.
• Classification using Gaussian Mixture Model (GMM), Error Back Propagation Neural Network (EBPNN), and Support Vector Machine (SVM).

Main Results:

• The system was trained and validated using the MIT-BIH arrhythmia database.
• Support Vector Machine (SVM) achieved the highest classification accuracy at 95.60%.
• Error Back Propagation Neural Network (EBPNN) demonstrated high accuracy at 93.41%, outperforming Gaussian Mixture Model (GMM) at 87.36%.

Conclusions:

• The developed system effectively screens for arrhythmias using ECG analysis.
• Supervised machine learning classifiers, particularly SVM and EBPNN, show significant promise for accurate arrhythmia detection.