Artificial Neural Networks for Identification and Classification of Obstructive Sleep Apnea and Prediction of Cardiovascular Diseases in Oman

To enhance the diagnostic process in hospital daily routine and avoid misdiagnosis, artificial intelligence methods (especially computer aided diagnosis and artificial neural networks) can be employed. These adaptive learning algorithms can handle diverse types of medical data and integrate them into categorized outputs. Sleep apnea is complete or partial cessation of breathing during sleep. Obstructive sleep apnea (OSA) is the common form of apnea that occurs when the upper airways are partially or completely obstructed for 10 seconds or more due to the relaxation of the dilating muscles. Symptoms of OSA are snoring, difficulty of sleep in night, stop of breath during sleep, daily sleepiness, poor concentration on work and sleeping while drive. Different kinds of cardiovascular conditions can be caused by OSA. Hypertension is most likely to be common for OSA patients. Atrial fibrillation (irregularity of heart beat) is most likely to be accompanied with severe OSA. Patients with OSA are also most likely to have coronary artery disease, happens when the blood vessels become narrow and this may lead to heart attack. Untreated severe OSA patients are as twice likely to develop heart attack and can develop heart failure. It is also well known that the increase in myocardial oxygen consumption is the main cause of ischemia events, and since the oxygen supply decreases with OSA, severe OSA can develop ischemic heart disease. So, in brief OSA has been identified as a significant cause of and/or contributor to cardiovascular diseases. Deep Learning (DL) is a novel machine learning field with a lot of applications in the last years. It is based on learning multiple levels of representations by creating a hierarchy of features where the higher levels are defined from the lower levels. In this project, different neural networks from simple perceptron to DLNN are to be used for identification of OSA patients from normal subjects and in classification of sleep apnea into different levels: mild, moderate and severe. This will provide a non-invasive automatic simple method based on single lead ECG signal for sleep apnea diagnosis to replace classical complex and costly polysomnographic test and thus to reduce the pressure on hospitals. NN will help also in diagnosis based on short length data (minute by minute diagnosis). In addition to that, by using DNN, the classification algorithm can anticipate early stages of cardiovascular disease in patients with sleep apnea, which will help a lot in preventative care for patients at risk of heart disease. The developed approach consists of two parts: a deep neural network learning-based training model and prediction model for the presence of cardiovascular disease in patients with obstructive sleep apnea.