Arrhythmia ventricular fibrillation classification on ECG signal using ensemble feature selection and deep neural network

Electrocardiogram (ECG) signal monitors the heart rate of the patients for long term periods and helps to analyse heart conditions of patients. Arrhythmia cardiac conduction disorder is characterized by irregular heartbeats observed in ECG signal. Arrhythmia classification helps to analyze the heart condition to support the doctor in the decision-making of correct treatment selection. Various existing methods such as feature extraction and deep learning method were applied for the arrhythmia classification. In this research, the ensemble feature selection method with Deep Neural Network (DNN) is applied for arrhythmia classification. The MIT-BIH arrhythmia database and MIT-BIH ventricular arrhythmia database were used to evaluate the performance of ensemble feature selection. The proposed ensemble feature selection method has the advantage of selecting the relevant features from the best solution of whale optimization, grasshopper optimization and Grey Wolf Optimization (GWO) methods. The proposed ensemble method finds the correlation among selected features of feature selection method and features with higher correlation are selected as relevant features. The proposed ensemble method has sensitivity of 96.95% and whale-DNN model has 96.47% sensitivity in arrhythmia classification. The proposed ensemble method has sensitivity of 92.71% and whale-DNN model has 91.21% sensitivity in ventricular arrhythmia classification. The result shows that the proposed ensemble feature selection has 98.11% accuracy and the existing CNN has 95.1% accuracy.