In recent years the location of brain structures as the basal ganglia has shown to be a useful tool in surgery of patients suffering from neurodegenerative diseases like Parkinson's disease, where it is needed the implantation of a neurostimulator in the subthalamic nucleus (STN). Despite the different existing approaches that seek to address this problem, it is still possible to improve classification rates because during the stage of preprocessing of the signals from microelectrodes (which capture the neuronal activity of the brain), usually filtering a known signal portion: background neuronal activity. This preprocessing is done under the assumption that this part of the signal is not relevant for the classification process. However, in no event processing is performed to validate this assumption. This paper develops a methodology for the detection of determinism in background neuronal activity in the microelectrode recording signals (MER): application to Parkinson's disease. The proposal is based on the method of surrogate data, nonlinear tool has now been widely used for time series analysis, especially in physiological series.