Sleep analysis through electroencephalogram cyclic alternating pattern a phase detection

• Autor(a) principal: Alves, Arturo José Morais
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.13/4309
• topic: Machine learning; Neural networks; Sleep analysis; CAP cycle; EEG; Redes neuronais; Análise do sono; Ciclo CAP; Informatics Engineering; .; Faculdade de Ciências Exatas e da Engenharia; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática; Domínio/Área Científica::Ciências Médicas

Sleep deprivation leads to a raise in mortality and other unwanted con sequences, as such, the diagnosis of sleep conditions is crucial for medicine. Sleep is a complex physiological process which is subdivided into different stages of deepness. A more detailed analysis of sleep reveals that events may occur that indicate an abrupt change in the brain electrical activity. Cyclic alternating pattern is a repetitive pattern associated with sleep instability that can be detected by analysing bodily signals. This can contribute to the detection of sleep disorders such as obstructive sleep apnea, using the cyclic alternating pattern as an indicator of such conditions. These cycles tend to be hard and slow to manually identify due to the fact that extensive analysis of long recordings of physiological signals by professionals is required. So, the automatic detection of such patterns is necessary. This can be achieved using Machine Learning algorithms such as neural networks that use pa tient data to train a model that can later recognize patterns within the sleep structure. In this work features extracted from the patient electroencephalo graphic signal are used by the classifier to learn how to identify the cycles. The classification results were improved by making use of methodologies such as feature selection and model tuning. The final solution presents very good results, obtaining an accuracy of 73%, a sensitivity of 77% and a specificity of 74%.