Recognition Of Obstreperous Sleep Apnea And Cardiovascular Diseases Using Machine And Deep Learning Models

This study examines the SHHS database and analyses and contrasts the performance of several ML and DL techniques for OSA and CVD detection. The dataset is primarily focused on diagnosing sleep apnea and associated breathing disorders, as well as evaluating the potential influence of sleep apnea on cardiovascular events. It contains polysomnography recordings from over 6,000 people. The current study evaluated ten top-notch ML and DL models, including SVM, DT, RF, NB, ANN, 1D-Convolutional Neural Networks (1D-CNN), ResNet, ResNet-100, ResNet-50, and ResNet-18 as well as Artificial Neural Networks (ANN). They suggested a 50:80:20 split for training and testing; however we evaluated the models with data splits ranging from 50:80:20 to see how well they predicted CVD risks in OSA patients. When compared to the other ML models, DT's 86.5 accuracy and 84.4 sensitivity were the best. Similarly, ResNet101 surpassed competing DL models, with an accuracy of 85% and a sensitivity of 85.5%.