Sleep Apnoea Detection with Smart Internet of Things Technology

Abstract

Sleep apnoea (SA) is a hazardous condition characterized by interrupted breathing during sleep. This prevalent medical issue affects individuals of all ages, potentially leading to severe complications when untreated including, cardiovascular problems, diabetes, and daytime fatigue etc. Unfortunately, SA often remains undiagnosed due to the costly and inconvenient diagnostic procedures associated with it. It stands as a significant global health concern, impacting nearly one billion people worldwide, with a prevalence of 17 to 23% in women and 34 to 50% in men. SA is recognized as a risk factor for cardiovascular disorders (CVD) and carries substantial individual, societal, and economic burdens. The economic costs of SA diagnosis and treatment services run into billions of dollars annually. The reference standard for diagnosing SA is polysomnography (PSG), conducted in a laboratory setting by trained professionals. However, this process is time-consuming, susceptible to human error, and demands technical expertise for both execution and interpretation. The inconvenience of in-lab PSG has spurred the need for new, simplified methods. This thesis posits that Computer-Aided Diagnosis (CAD) systems can enhance diagnostic efficacy. To explore this hypothesis, the thesis introduces innovative real-time detection techniques for Obstructive Sleep Apnoea (OSA) and the development of a high-performance OSA detection system. This system, offering continuous OSA detection, addresses the practical challenges associated with traditional diagnostic approaches. The integration of Internet of Things (IoT) and advanced Artificial Intelligence (AI) technologies, with a focus on the Lifetouch sensor, represents a novel approach to improve the accuracy of OSA detection. This innovative strategy aims to overcome barriers to timely and reliable diagnosis and monitoring of sleep disorders. To thoroughly assess the algorithm, a clinical study enrolled 15 patients with a history of OSA. Simultaneously, standard PSG monitoring and diagnosis were conducted, serving as the benchmark for comparison. This dual approach ensured a robust evaluation of the DL algorithm's performance against established PSG methods, providing a comprehensive understanding of its capabilities in OSA detection. The trial results highlight the potential of the proposed technology model, showing a high level of patient acceptance and satisfaction with Lifetouch wearables. However, the identification of only two OSA cases among the 15 patients studied was lower than anticipated. These findings emphasize the need for improved detection methods, as addressed by the novel techniques introduced in this thesis. The results presented here also highlight the efficacy of the developed methods, showcasing their ability to deliver quick, reliable, and standardized analyses an essential step forward in overcoming the limitations of conventional diagnostic approaches.