Early Detection And Prediction Of Sleep Apnoea Using Deep Learning Techniques

Sleep Apnoea is a sleep disorder where inhaling and exhaling repeatedly pauses and continues; it can be termed breathing difficulties, where the throat muscles irregularly relax and block the airway during sleep. Other signs may include tiredness during the day and improper sleep. In this study, sleep apnoea is detected and predicted using deep learning and machine learning techniques from electrocardiogram (ECG) signals. The dataset includes 70 patients' ECG records, and ten features were collected using a label encoder; the ECG signals are encoded into the numerical format. A deep learning technique called Restricted Boltzmann Machine (RBM) was used to program the detection of SA for the patients in the data set. A few algorithms in Machine Learning, specifically K-Nearest Neighbors, Random Forests, and Decision Trees, form the ensemble technique used to predict SA for the input in and out of the dataset. Then, the performance of the proposed models is compared using classification evaluation measures. The ensemble technique proved to be the best, with an accuracy of 96.4%, whereas RBM models give an accuracy of 94.11%. This proves that combining techniques works better than individual techniques for predicting SA. An interface is built to show the output. Now, the model developed here is ready for validation and clinical use. The proposed work is related to the latest sustainable growth in the field of Industry, Innovation, and Infrastructure, which helps to identify Early Detection and Prediction of Sleep apnea, especially while driving vehicles.