Análise da estacionariedade e gaussianidade da atividade elétrica neural, do ruído biológico e do ruído de instrumentação

Abstract

This work analyzes signals of spontaneous neuronal electrical activity, measured by devices known as Multi-Electrode Arrays (MEA), which are the substrate of neuronal cultures of rat hippocampus. The database under investigation consists of five signals: (i) instrumentation noise, (ii) biological noise, (iii) inactive culture and (iv) two active culture experiments. Beyond the four primary statistical measures of the MEA amplitude histogram, the Mean Ratio Test and Jarque-Bera Test were proposed for estimating the degree of wide-sense stationarity and Gaussianity of MEA recordings respectively. The results indicate that spectral-density-based stationarity analysis is able to distinguish between experiments of biological noise, inactive culture and noise instrumentation. Meanwhile, JB statistic discriminates active culture experiments from others, i.e., whereas application of MR leads to a remarkable conclusion: although biological noise and instrumentation noise present more or less the same amplitudes and gaussianity, the former is essentially non-stationary. The quantitative indices proposed in this study showed good statistical correlation with those of classical spike analysis for active culture experiments. Thus, results demonstrate that neuronal electrical activity of spikes and bursts showed a remarkable non-stationarity and non-Gaussianity character when compared to biological noise and instrumentation noise, that are very close to stationarity and Gaussianity.