الفهرس | Only 14 pages are availabe for public view |
Abstract 1. Introduction Classification of electrocardiogram (ECG) signals plays an important role in diagnoses of heart diseases. An accurate ECG classification is a challenging problem. Automated classification of heart diseases will increase the effectiveness of diagnosis. ECG classification into arrhythmia types early is important in detecting heart diseases and choosing appropriate treatment for a patient. Recent machine learning methods have shown significant advantages, for detecting arrhythmia types. Most machine learning experimentation starts from understanding the ECG data on local computer and does not require that much computation power. However, very quickly we will run into the need more than our local CPU. The cloud is by far the more scalable place to do machine learning. We will get access to the latest GPUs or, even TPUs that we would not be able to afford and maintain on ourselves. Using machine learning in a Collaborative manner in which the intelligence is distributed across Edge/Fog layer, and Cloud layer. The edge-computing layer is used as a server with computation power and the cloud layer will use as a storage layer. Cardiovascular disease has been increasing and become a great threat to human life in many recent years. According to statistics from the World Health Organization (WHO), up to 17.9 million people died each year from cardiovascular diseases and the corresponding number of deaths is 31% per year. These statistics can be a great impact on the research, diagnosis and treatment of cardiovascular diseases. Early diagnosis and treatment of cardiovascular diseases can significantly reduce the number of deaths. Moreover, with the development of science and technology, homebased and remote disease monitoring and warnings with classification technology about cardiovascular disease is very useful for both doctors and patients. In addition, real applications of the health care technology for remote monitoring will save a lot of time and money for society. 2. Objectives Detect the early abnormal indicators for cardiac failures that an physicians cannot discover. The aim of the work is to detect and categorize ECG signal patient heart arrhythmia. An adequately developed Computer-aided detection method for cardiac recognition, based on ECG, is likely to eliminate subjectivity and provide an informed decision-making quantitative evaluation. We discuss the existing CAD for automatic detection of cardiac and highlight the creation of an ECG-based diagnostic system that automatically detects cardiac using signal processing methodologies, deep learning algorithms, and optimization techniques. We aimed to examine five forms of micro-classification heartbeats, i.e. normal (N), Supraventricular ectopic beat (S), Ventricular-ectopic beats (VEBs), and Fusion (F). Considering the recommended practice of the Association for the Advancement of Medical Instrumentation (AAMI). 3. The Methodology Our thesis achieved its goal through pioneering and harnessing the state-of-art ICT technologies, such as computing, machine learning and data fusion. The following is a summary of the used methodology: 1. Presenting a comprehensive review to study state-of-the-art challenges and recommended techniques for enhancing the Cardiac health recognition based machine learning and optimization techniques. This thesis surveys the current techniques and presents several studies to reflect promising enhancements. 2. To reduce the number of features in a data-set by creating new features from the existing ones, different features extraction and descriptors were investigated and adapted for multi-class arrhythmia detection. Cross features are extracted of ECG beats of different sizes with multiple descriptors. Additionally, new morphological descriptor is proposed and utilized for further mining the important features from the signals. 3. Introducing the use of the novel optimization technique called Manta Ray Foraging Optimization (MRFO) to solve feature selection problem and configure support vector machine to enhance the classification process by adapting its hyper-parameters. 4. Proposing an enhanced version of optimization technique namely improved marine predators algorithm (MPA) hybridized with convolution neural network (CNN) to tackle the model configuration and tuning its parameters to enhance the classification process. 4. Results The experimental results of ECG arrhythmia classification using the proposed MRFO-SVM revealed with evidence its superiority with overall classification accuracy of 98.26% over seven well-known metaheuristic algorithms. The experimental results of classification of heart diseases using the proposed IMPA-CNN also showed its superiority over several known improvement algorithms with a general classification accuracy exceeding 99%. 5. Recommendations It is clear in this thesis the important role that optimization algorithms and machine learning algorithms play in resolving many science and engineering problems and in this study we specifically looked at the high capacity for optimization algorithms and machine learning algorithms in solving cardiac health recognition. As a result, we urge that more scientific progress be made in proposing and enhancing these types of algorithms because they help solve and optimize difficult problems in science and engineering. |