Sleeping Abnormalities Detection using Deep Learning Techniques

Abstract

Insomnia is a sleep disorder characterized by difficulty falling or staying asleep, and it can significantly impact an individual's quality of life. It affects one person’s mental health and physical health. The manual detection mechanism is quite time-consuming and prone to human errors. Artificial intelligence (AI) has the potential to be useful in multiple aspects of sleep medicine, such as sleep and respiratory event scoring, sleep disorder diagnosis and management, and improving public health. Despite being in its early stages, AI faces various challenges that hinder its widespread use and generalizability in clinical settings. Nevertheless, AI can be a powerful tool in healthcare, as it can enhance patient care, diagnostic capabilities, and sleep disorder management. However, regulating and standardizing existing machine learning and deep learning algorithms is crucial before integrating them into sleep clinics. In this paper, the Convolutional Neural Network (CNN) algorithm is used to analyze EEG recordings of multiple individuals, which are collected along with the EDF file. Pre-processing is conducted to remove artifacts and improve the quality of the data. The EEG recordings are segmented without overlapping, using a sliding time window. The raw EEG signals are then converted into a set of features that serve as input for the CNN. The CNN is trained using a loss function and optimization algorithm. Readings are matched with the CNN model file , then the sleeping stage classification is performed and precautions are provided.