Remoção de artefatos e análise de parâmetros espectrais em sinais de EEG: efeitos do fármaco flunitrazepam

Abstract

Removing various artifact types of electroencephalographic (EEG) signals is essential because the analysis of raw data may lead to wrong interpretations. The recorded signals used in this research have strong artifact influence. The reason for that comes from the experimental protocol, that involved simultaneous perform of psychometric assessment task DSST (Digit Symbol Substitution Test) and signal recording (in research about flunitrazepam memory influence). In this study, two techniques are presented: first of all, Local SSA (Local Singular Spectrum Analysis) that removes ocular artifacts without losing signal samples; and then is applied the method that compares the samples with a threshold based on the standard deviation of the reference signal considered free of high magnitude artifacts. Combining both methods, 89 % of the recorded signal, on average, was considered free from artifacts. In contrast, if the simple use of the Standard Deviation method is applied, only 50% of the signal would remain. With the resulting signals, the normalized power spectral density (PSD) was estimated for obtaining the individual defined energy bands, using IBIW (Individual Band and Individual Width) criteria. The statistic test (ANOVA) was performed between flunitrazepam and placebo sessions that exhibited significant difference (p < 0.05), mostly in lower-1-Alpha, lower-2-Alpha, Beta-1, Beta-2 e Beta-3 bands. These results are consistent with the ones found in the literature, validating the cognitive effects of flunitrazepam and DSST. Moreover they detail the behavior of the various energy bands in all derivations of the 10-20 system.