Heart arrhythmia diagnosis based on the combination of morphological, frequency and nonlinear features of ECG signals and metaheuristic feature selection algorithm

Abstract

Cardiac arrhythmia disorder is known as one of the most common diseases in the world. Today, this disease is considered as the leading cause of death in industrial and semi-industrial societies. Various tools and methods have been developed to study the detection of heart diseases, based on analyzing the electrocardiogram (ECG)signal. Due to the simplicity and noninvasive nature, ECG signals are vastly used by physicians to determine the heart problems and abnormalities.In this paper, a computer-aided diagnosis (CAD) system is provided for the automated classification and accurate diagnosis of seven types of cardiac arrhythmias using the ECG signal. The basis of this method is using machine learning algorithms to classify normal rhythm and six abnormal cardiac functions. In the proposed method, after the pre-processing stage, the ECG signal is segmented, and various morphological characteristics, frequency domain features, and nonlinear indices are extracted for the ECG signal. Several metaheuristic optimization algorithms are used to remove redundant or irrelevant features and reduce the feature space dimension. These are used on the combination of the extracted features in which, non-dominated sorting genetic algorithm (NSGA II) as a multi-objective optimization algorithm has the best performance. Furthermore, various machine learning algorithms include k-nearest neighbor (KNN), feed-forward neural network (FF net), fitting neural network (Fit net), radial basis function neural network (RBFNN) and pattern recognition network (Pat net) are employed for the classification. The highest accuracy obtained based on ten-fold cross-validation from the FF net is 98.75%, demonstrates the efficiency of the proposed method and the achieved improvement compared to the other similar works with the same dataset. The combination of a vast and various features from morphology, frequency, and nonlinear characteristics to demonstrate the diverse aspects of ECG signals as well as employing a multi-objective meta-heuristic optimization algorithm for selecting the more correlated features are the main contributions of this study.