Mode Detection in switched pursuit tracking tasks: Hybrid estimation to measure performance in Parkinson´s disease

Parkinson´s disease (PD) is a neurodegenerative disorder that impairs motor skills, speech, and other voluntary movement, and may be associated with cognitive inflexibility. Fourteen PD subjects (both on and off medication) and 10 normal subjects performed a manual pursuit tracking task, in which the dynamics of the task suddenly change without explicit enunciation. The task dynamics have three modes, in which the error (the difference between the target and the user´s cursor) is attenuated, exaggerated, or unchanged - hence we model the subject performing the tracking task as a hybrid system with arbitrary switching. Second-order stochastic LTI models of tracking performance in each mode are first obtained through system identification. We then use a multiple model adaptive estimation (MMAE) algorithm to determine a) whether each subject successfully adapted to the sudden change in tracking dynamics, and if so, b) the delay in switching to the new mode. These parameters were analyzed for all subjects, and found to be statistically significant across groups. While normal subjects consistently detected the change in task dynamics, PD subjects show considerably more difficulty in detecting the switch (especially off medication), and did not switch into the new mode as quickly as normal subjects. Our results suggest that PD subjects have considerable impairment in adapting to changing motor environments.