Automatic Model of Sleep Staging using Convolution-1D Network from EEG signals

Sleep is vital for physical and mental health, making accurate sleep disorder diagnostics essential. Sleep disturbances can lead to numerous health issues, one of the most common being Sleep Apnea (SA). SA occurs when breathing is obstructed, leading to malfunctions in the brain and heart. In severe cases, it can cause strokes and cardiovascular diseases. Additionally, SA increases the risk of secondary issues, such as daytime vehicular accidents. Due to the complexity of the condition, early diagnosis and treatment are crucial. Polysomnography (PSG) is commonly used for diagnosing SA, and ongoing research aims to enhance diagnostic accuracy. PSG data, which is recorded overnight, is large and rapidly changing, making manual analysis by healthcare professionals a challenging and time-consuming task. As a result, automated SA detection using machine learning has become a significant area of research. Given the vast volume of medical data, deep learning is essential to quickly and effectively extract meaningful features. Initially, EEG signals are collected from a standard data source, and the collected signals undergo preprocessing, which decomposes them into five components: delta, theta, alpha, beta, and gamma. Manual sleep staging is complex, time-consuming, and expertise-dependent, necessitating automated solutions derived from EEG data and deep learning models. This study explores CNN architecture for sleep stage classification, leveraging its ability to learn spatial features from EEG data effectively. CNN achieves notable accuracy in classifying sleep stages, showcasing its capability in spatial feature extraction. Future work will enhance model scalability, integrate multi-modal data, and explore real-time applications in wearable devices for advanced sleep monitoring.