Developing A Sophisticated Sleep Monitoring Platform Capable Of Analyzing Polysomnography (Psg) Signals To Effectively Detect Sleep Disorders

Abstract

Background and objective: The goal of this project is to develop a new framework and application that can automatically detect sleep phases and sleep problems. We will achieve this by utilizing a machine learning method based on single-channel EEG data. The electroencephalogram (EEG) is an essential tool for detecting sleep issues since it detects and records brain activity. The electroencephalogram (EEG) is a vital diagnostic tool for identifying sleep problems since it detects and records brain activity. The accuracy of the data, as well as the stability of both the patient and the measuring apparatus throughout sleep, are crucial factors in monitoring patients with sleep disorders. Methods: We use the CAP dataset, a subset of Physionet consisting of a significant collection of sleep-related data. To extract the wave properties, we divide the data into frequency ranges ranging from 0.5 to 30 Hz. Then the ANOVA techniques, normalization, and collapse aid in processing the waves. We train the model using processed waves after removing any background noise and patient movements from each EEG. Using their sleep data, we internally design a user interface that allows users to input, evaluate, and get predictions for illnesses linked to insufficient sleep, as well as other research on sleep. Outcome: The approach achieves two different goals. The first model evaluates the many phases of sleep, while the second model categorizes sleep disorders. After assessing the influence of various diseases on sleep phases, we ultimately chose the sleep stages model, which consistently achieved 96% accuracy in classifying sleep phases. Unlike the diseases, the model incorporates all the specific characteristics of each condition, effectively classifying sleep disorders into multiple groups with a classification accuracy exceeding 95%. After completing all these processes, we created a reliable and user-friendly interface to assist medical professionals in eliminating their biases. Moreover, it provides the foundation for therapy and prognosis for those who want to evaluate their sleeping quality, as well as for patients receiving treatment. Conclusions: The study ultimately found that each person's unique brain signaling characteristics limit the model's effectiveness, preventing the entire system from fully synchronizing. The model serves as a diagnostic tool to accurately determine the nature of the ailment.