2026-04-17T00:34:20Zhttps://recercat.cat/oai/request

oai:recercat.cat:2117/1737812025-07-22T23:49:39Zcom_2072_1033col_2072_452951

urn:hdl:2117/173781 Electroencephalogram characterization by Multiscale Entropy for sleepiness detection Borràs Argemí, Marta Àrees temàtiques de la UPC::Enginyeria biomèdica Son -- Aspectes sanitaris The excessive daytime sleepiness (EDS) is one of the symptoms of several sleep-related disorders. Several investigations sustain that there are more affected people than the clinically diagnosed, while many studies have been carried out to assess daytime sleepiness, automatic EDS detection remains an open problem. The protocols for sleepiness detection are complicated, long and annoying for the patients. Thereby, the electroencephalogram (EEG) signal has been introduced to provide additional, valuable and non-intrusive information about the state of the patient. In this project, two methodologies based on linear frequency domain analysis (Power Spectral Density (PSD)) and nonlinear time domain analysis (Multiscale entropy (MSE)) has been developed and applied to the pre-processed EEG signals (once eliminated the EEG artifacts with two Butterworth filters (band-pass (0.3-70Hz) and Notch) and applied the independent component analysis (ICA) ), for sleepiness characterization. In this way, different indices have been defined in order to best describe the different EEG behavior between EDS and without daytime sleepiness (WDS) patients. It has been found that both MSE and PSD techniques contribute to highlight the EEG characteristics of both groups of patients (EDS and WDS), and therefore, these methodologies can help to find EEG indices which define this sleep disease and thus, it can be detected more easily. 2018-06-21 Bachelor thesis Restricted access - author's decision Universitat Politècnica de Catalunya