A Hybrid Feature Extraction Method for Heart Disease Classification using ECG Signals

Electrocardiogram (ECG) records electrical activities of human heart that can help identify abnormalities. ECG signals are widely used to classify the signals into various categories that help medical specialists in diagnoses of heart diseases. An accurate ECG classification is a challenging problem. Different methods have been used and studied for classification of cardiac signals and classification of heart rhythm disorders in recent years. Temporal, morphological, and statistical features are commonly used for the classification of ECG signals. In this paper, we propose a hybrid feature extraction method by applying correlation and regression methods in addition to the three techniques mentioned above. Moreover, we propose Back propagation neural network (BPNN) to classify heart disorders. Furthermore, the results were compared with other traditional machine learning algorithms, including K-nearest neighbor (KNN), Decision Tree (DT) and Random Forest (RF). The obtained results demonstrate that the proposed feature extraction combination technique positively affects the BPNN model and applied traditional machine learning techniques. Proposed BPNN achieved 95.07% accuracy, DT 96.09%, KNN 97.64% and RF 96.74%.