Y Rakhshani Fatmehsari - School of Electrical and Computer Engineering University of Tehran, Tehran, Iran
F Bahrami - CIPCE, School of Electrical and Computer Engineering University of Tehran, Tehran, Iran

In this paper we compare the complexity ofgait patterns insubjects with Parkinson’s disease (PD), under twoconditions of deep brain stimulation (DBS) off and on, with that of normal subjects.To this end, for the first time in nonlinear analysis of PD gait patterns, Lempel-Ziv Complexity (LZC)values are used as the measure of complexity of gait time series.This means that greater LZC values correspond to higher complexity of the corresponding time series. The LZCvalues are computed for 9 PD patients in DBS off status and the same patients in DBS on status. It was also calculated for 10 healthy control subjects matched in age, height and weight the PD patients. We found that with applyingLZC measure, we could classify PD patients from healthy controls. To evaluate the performance this measure, receiver operating characteristic (ROC) plots have been used. For classification of PD patients inDBS off status from control subjects, it gave the sensitivity and specificity of up to 100%, and the accuracy of up to 94.74%. Also, to separate PD patients in DBS on status from control subjects, we gained the sensitivity, specificity, and accuracy of up to 100%, 90%, and 89.47%, respectively. Furthermore, we found that gait time series of PD patients have larger LZC values than those of control subjects and also, the PD patients in DBS off status havegait time series with larger LZC values thanwhen they are in DBS on status.

Parkinson’s disease (PD);Lempel-Ziv Complexity (LZC);Wearable sensors;Gait time series