HEART DISEASE ANALYSIS BASED ON DATA MINING, DEEP LEARNING AND WEARABLE TECHNOLGIES

Abstract

Information technology has virtually altered every aspect of human life in the present era. The application of informatics in the health sector is rapidly gaining prominence and the bene ts of this innovative paradigm are being realized across the globe. However, this evolution has a real challenge of how to get maximum out of the patient's data wherever already available, which can employed by high computer technologies such as data mining, deep learning and wearable health technologies, and turned into useful information and knowledge. This data can be used to develop expert systems to save expert clinicians, help in diagnosing some life-threating diseases such as heart diseases and predicting a di erent diseases before that happening like heart attack, all of this with less cost, processing time and improved diagnosis accuracy by decreasing the number of misdiagnosis. In this study, we have chosen the heart disease as a case of study due to its direct in uence on the human life and the high number of death caused by this type of disease. This study aims to develop a health informatics system for the prediction of heart diseases using data mining, deep learning and wearable health technologies. We worked on the ECG signal and risk factors of the heart state regarding to its capabilities to classify a di erent heartbeats, some heart diseases based on ECG Arrhythmia and detecting Myocardial Infarction disease as known as heart attack. Data used are gathered from three sources, the rst one is the e-Health platform mounted on Raspberry Pi3, which allows to generate and record ECG signals which is a sequence time series, the second source are records provided by Physionet website and the source are patient's life data that are the main risk factors of the heart state similar as heart disease data provided by UCI Machine learning website. We propose in this memory to use CNN, LSTM and k neighbors in the training and testing steps, in order to classify heartbeat types. some CVDs and predict a heart disease. The scored results are quite acceptable, however some adjustments can be introduced to the way of collecting data from patients. Despite this, the trained model still improve high capabilities on classifying heart beat types and Arrhythmias.