Identificação dos estados de alerta e sonolência usando Teste F Espectral aplicado a sinais pupilométricos

• Ano de defesa: 2014
• Autor(a) principal: Valciene Caroline Goncalves Soares
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais
• Palavras-chave em Português: Engenharia elétrica; Pupilometria; Sonolência; Alerta; Potência; Teste F Espectral
• Link de acesso: https://hdl.handle.net/1843/BUOS-9TZHLT

The use of pupillometry has been reported in several studies as an alternative method to current complex and expensive exams for sleepiness evaluation. This exam is based on the analysis of pupillogram which consists of recording pupil diameter variation (Time Serie of Pupil Diameter - TSPD). Due to its behavior at different conditions of sleep deprivation, the TSPD has been suggested as objective physiological measure for assessing the alertness level of an individual. During drowsiness, pupil diameter tends to decrease and the power of low frequency oscillations increases. Thus, this work aims at investigating the ability of Spectral F-Test (SFT), a frequency-domain statistical signal processing technique, to differentiate alertness and sleepiness states. For this purpose, pupillometric signals were recorded in the Psychobiology Department of UNIFESP) from 13 volunteers in 10 minutes-sessions every 4 hours, for a 36 hours-period of sleep deprivation, resulting in 10 TSPD records. The power of TSPD were calculated based on Welch Periodogram and compared with the first session power (obtained at 8 am of the first recording day after a normal night of sleep), considered as alertness reference. For inter-session analysis, the percentage of volunteers for whom the power of TSPD analysis (sessions 2-10) increased compared to the power of TSPD reference (session 1) was investigated. Then, the most suitable frequency range to differentiate the sleepiness and alertness states was established comparing the median power of TSPD in the analysis sessions with the reference by means of the Wilcoxon signed-rank test for frequencies up to 0.8 Hz. Besides of inter-session study, an intra-session investigation was performed to evaluate SFT time evolution, calculating three SFT estimates for the same 10 minutes-session. The inter-session analysis results showed that from session 7 onwards occurred the highest percentages of volunteers for whom the power was statistically higher than in session 1, with rates above 90 % for the most frequencies below 0.2 Hz. The Wilcoxon test showed statistical differences between median powers (p0.001), obtained from the seventh session and frequencies below 0.2 Hz. Hence, the results suggest the use of the Spectral F-Test (SFT) as a technique to differentiate the states of alertness and sleepiness during periods of sleep deprivation. For the time evolution analysis, the percentage of subjects for whom there were statistical difference for each session compared with the first for the frequency 0.0439 Hz ( within the range of better differentiating between conditions) did not show any consistent pattern of changes in alertness / drowsiness state for the whole casuistry.