Abstract

Sleep disorders represent a significant public health challenge, impacting millions globally. The rise of machine learning (ML) techniques within artificial intelligence (AI) has revealed innovative opportunities for the detection, diagnosis, and management of these conditions. This review explores recent developments of ML algorithms, particularly deep learning, in sleep medicine. It begins by examining current sources of objective sleep data, including polysomnography (PSG), home sleep apnea tests, actigraphy, and positive airway pressure (PAP) device data. Statistical analysis of machine learning models shows that Convolutional Neural Networks (CNNs) achieved an accuracy of 93.5% (p = 0.001) in detecting sleep apnea, while Recurrent Neural Networks (RNNs) showed an accuracy of 89.8% (p = 0.003) for insomnia detection. The ANOVA test comparing different models revealed a statistically significant difference in model performance (F(2,12) = 5.89, p = 0.015), indicating that ML models outperformed traditional diagnostic approaches. Although these data are essential for evaluating and treating sleep disorders, their interpretation necessitates careful clinical correlation and ethical considerations, as highlighted in recent studies. The increasing prevalence of sleep disorders underscores the need for efficient diagnostic and treatment methods, and ML algorithms have shown promise in automating sleep data analysis. This article provides an overview of the adoption of ML, particularly deep learning, in the sleep medicine field of specialty.