PEI-LIN LEEWenbo GuWen-Chi HuangAmbrose A. Chiang2024-11-282024-11-28202497830316826299783031682636https://scholars.lib.ntu.edu.tw/handle/123456789/723354The realm of sleep medicine is experiencing a rapid evolution driven by advancements in sleep technologies. Emerging devices for obstructive sleep apnea (OSA) detection are becoming increasingly sophisticated and portable, due to the integration of innovative miniaturized sensor designs, advanced processing techniques, and artificial intelligence/machine learning/deep learning (AI/ML/DL) algorithms. AI models have become ubiquitous in sleep medicine, fundamentally altering the approach to OSA detection from screening at the clinic to at-home diagnosis utilizing cutting-edge sleep technologies. This chapter delves into ML models that leverage clinical features for OSA screening and illustrates their use with a case study. We also explore the current landscape of innovative AI/ML/DL models employing photoplethysmography and accelerometry for at-home OSA diagnosis. Finally, we offer insights on crucial considerations for model design, dataset selection, and performance evaluation and emphasize the importance of external testing using independent datasets. As the complexity of physiological signals increases with the data integration from various sensors, more advanced DL techniques might suite better for handling intricate data. This trend highlights a shift beyond traditional ML and basic DL models toward more advanced, customized, and powerful DL approaches.book part10.1007/978-3-031-68263-6_4