Recognition and Classification of Cardiac Arrhythmias Using Discrete Wavelet Transform (DWT) and Machine Learning Techniques

Abstract

Cardiac arrhythmias are heart rhythm problems that usually occur when the electrical impulses coordinated with the heartbeat do not work correctly. For this reason, detecting abnormalities in an electrocardiogram (ECG) plays a vital role in patient follow-up. Due to the presence of noise, the irregularity of the heartbeat, and the nonstationary nature of ECG signals, their interpretation can be difficult, requiring the use of advanced computer systems to support the diagnosis of cardiac disorders. Therefore, the development of assisted ECG analysis systems is a current topic of study, and the main challenge is to achieve adequate accuracy for application in the clinical setting. Therefore, this article describes a software tool for classifying ECG samples into the main classes of cardiac arrhythmias by removing noise from the ECG signal at the preprocessing stage using conventional digital filters, the location of the QRS complex is essential for the identification of the ECG signal. Therefore, the position and amplitude of the R peaks are determined in the segmentation stage. Then the selection of the most relevant features of the ECG signal is performed using the discrete wavelet transform (DWT). The ability of the extracted features to differentiate between different classes of data is tested using machine learning techniques such as k-Nearest Neighbors, Neural Networks, and Decision Trees with 10-fold cross-validation. These methods are evaluated and tested with the MIT-BIH arrhythmia database, achieving the best accuracy of 98.54% using the k-Nearest Neighbors classifier. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.