Classification of ECG Signal Using Hybrid Deep Neural Network

Abstract

This dissertation studies a comprehensive approach to evaluating the performance of different machine learning and deep learning algorithms to classify five ECG signal categories. A novel algorithm is also proposed to achieve the same objective efficiently. Cardiovascular disease is responsible for a prominent amount of mortality among humankind is detected by analyzing ECG signals. ECG signal classification is an arduous task since sometimes the abnormal heartbeats are too similar to categorize. Most of the patients with heart diseases come to the doctor when the person is severely attacked. Therefore, doctors or medical persons cannot take much time to start the treatment. The heart is the most sensitive organ of the body, a misapprehension in classification can cause death to the patient. Machine learning and deep learning can be handy tools for the classification of the ECG signal quickly and efficiently. A Famous MIT-BIH ECG signal dataset was utilized to train and test the models. Six machine learning algorithms and five deep learning algorithms were studied with efficient hyperparameter optimization technique, and their performance was evaluated. Finally, a novel Hybrid Deep Neural Network (HDNN) was proposed which provided the best accuracy of 99.23% among all the algorithms studied for the classification of ECG signal. A detailed comparative analysis of performance with all other algorithms was carried out in terms of accuracy, precision, recall, and F-1 score