الفهرس 
Chapter 1 IntroductionOnly 14 pages are availabe for public view
 |  | from | 105 |  |  |
| | | from | 105 | | |
Abstract
Cardiac disease is one of the leading causes of death around the world. All abnormal heart conditions are caused by various types of diseases such as arrhythmia. Electrocardiogram (ECG) plays a critical role in the medical field, as it records the electrical activity of the heart over time. An efficient analysis of the ECG is essential to detect numerous heart diseases. The main aim of this thesis is to propose an effective ECG data analytics approach to improve cardiac healthcare by achieving high classification accuracy for many cardiac diseases. The proposed approach includes two modules: the ECG data stream analytics module using machine learning and the ECG data analytics module using deep learning.
The real-time ECG signals are submitted to the first module, then preprocessing and feature extraction techniques are applied to the ECG streaming data. The first module proposes an effective detection approach for Paroxysmal Atrial Fibrillation (PAF) events using the Extreme Learning Machine (ELM) classification technique. Detecting PAF events from ECG data streams plays a vital role in generating real-time alerts for cardiac disorders. These alerts are extremely important to cardiologists in taking precautions to prevent their patients from having a stroke.
The second module of our proposed approach is focused on classifying the ECG into five heartbeat classes using deep learning to identify cardiac arrhythmias. An optimized Convolutional Neural Network (CNN) model for ECG classification is proposed. We present an optimization step for the proposed CNN model using a customized genetic algorithm. It provides an automatic suggestion for the best hyperparameter settings of the proposed CNN. The CNN model is designed with an optimal configuration to classify cardiac arrhythmias quickly and effectively. Experiments show promising results of our proposed ECG data analytics approach. Our approach achieved classification accuracy of 97.2% and 94.4% for detecting PAF events in non-streaming and streaming modes. Additionally, the performance of our proposed CNN is better than other existing methods both in terms of higher classification accuracy (98.45%), and lower computational complexity.