Sit-to-stand or stand-to-sit: Which movement can classify better Parkinsonian patients from healthy elderly subjects?

In this paper, we proposed Approximate Entropy (ApEn) measure for nonlinear analysis of patterns of two daily common movements, that is sit-to-stand and stand-to-sit, in patients with Parkinson´s disease (PD) using deep brain stimulation (DBS). ApEn value of a signal indicates its complexity and irregularity, such that larger the ApEn measure is, more stochastic and less regular the signal is. Here, this nonlinear measure is used for classifying three groups of 1) healthy subjects, 2) PD patients with DBS off, and 3) the same PD patients with DBS on, based on their sit-to-stand and stand-to-sit patterns. To this end, the area under receiver operating characteristic (ROC) plots (AUC) is used to evaluate the capability of ApEn. For stand-to-sit patterns and for discrimination between two groups of normal vs. PD with DBS off, normal vs. PD with DBS on, and PD with DBS off vs. PD with DBS on, we achieved the AUC values of up to 0.9625, 0.9, and 0.9583, respectively. Also, there was high correlation between ApEn values of PD patients obtained from different sensors and the severity of their Parkinson´s disease i.e. UPDRS scores (up to 0.7671 for PD with DBS off and 0.774 for PD with DBS on). For sit-to-stand patterns and to distinguish between normal vs. PD with DBS off, normal vs. PD with DBS on, and PD with DBS off vs. PD with DBS on, we achieved the AUC values of up to 1, 0.9778, and 0.9167, respectively. Moreover, for this pattern, there was higher correlation between ApEn values of PD patients and their UPDRS scores (up to 0.9573 for PD with DBS off and 0.7145 for PD with DBS on). Furthermore, in both sit-to-stand and stand-to-sit patterns, it was observed that PD patients have larger complexity values than healthy controls and PD subjects in DBS on state had smaller irregularity values than PD subjects in DBS off state. In general, we may conclude that sit-to-stand patterns can distinguish PD patients from healthy subjects and also be used to estimate the severi- y of Parkinson´s disease better than stand-to-sit patterns.