Modeling and characterization of signals recorded in basal ganglia of Parkinson's disease patients

This paper shows the results of the analysis, characterization and modeling of basal ganglia intracerebral signals recorded during surgical intervention of patients with Parkinson's disease (PD). Statistical tests are used, both graphical and numerical analysis of normality and stationarity of signals sampled at varying time windows (0.5 s, 1 s, 2 s, 3 s, 4 s and 5 s). Subsequently characterization is performed in the frequency domain, obtaining values of local coherence for different windows, thereby allowing identifying of a suitable temporal window to perform parametric modeling through both linear and nonlinear approaches. Linear and nonlinear parametric models have been tested with lower orders than 20, finding that AR (13) model satisfies Akaike Information Criterion (AIC) and minimum computational work.