Introduction: Motor deficit in Parkinson’s Disease is typically assessed by subjective measures such as the Unified Parkinson's Disease Rating Scale (UPDRS) [1] that measure un-naturalistic and discontinuous movements, and may not generalize to continuous limb movements. In this study, we present a simple joystick task that overcomes these limitations by monitoring motor errors in continuous, naturalistic movements with second-timescale precision. We also investigated local field potentials (LFP) in the subthalamic nucleus (STN) to investigate the relationship between behavioral performance and neural states.

Methods: The task we designed required subjects to follow a target that moves in one of several invisible patterns as closely as possible with a cursor controlled by a joystick (Fig 1a). Each trial lasts 10 to 20 seconds depending on the pattern. From four patients, we collected the X and Y positions from joystick trajectories and LFPs in the STN during awake Deep Brain Stimulation (DBS) surgery.

Results: Two performance metrics were identified: tremor magnitude (TM), which quantifies physiological tremor amplitude, and vector error (VE), which quantifies how well a patient tracks the moving target by measuring the vector error between the target and cursor movements (Fig 1b). Motor behavior was analyzed in a sliding 2-second epoch. We found epochs with higher VE or TM were associated with broadly elevated LFP power (Fig 1c). Using frequency bands from 2-400Hz with a 3Hz step as features, machine learning algorithm (Support Vector Machine) based on VE or TM was able to predict behavioral motor performance with 78.6% and 82.9% accuracy, respectively (Fig 1d).

Conclusions: These results demonstrate that motor performance can be predicted on a short time scale using broadband LFP power. This approach could be useful for implementing an adaptive, closed-loop DBS system [2].

Patient Care: Our task design may serve as an objective quantification of Parkinson’s patients’ motor performance caused by tremor, rigidity, bradykinesia and/or dystonia over a continuous time scale. In addition, our LFP analysis approach may contribute to developing an adaptive, closed-loop DBS system that maximize efficiency while sufficiently reducing symptoms and side effects.

Learning Objectives: By the conclusion of this session, participants should be able to: 1) Describe the importance of using a continuous task to quantify motor performance in Parkinson’s patients. 2) Discuss, in small groups, how neuronal features are related to different symptoms such as tremor, rigidity, bradykinesia and dyskinesia. 3) Identify an effective strategy to capture neuronal features associated with motor error.

References: 1. Pal G, Goetz CG: Assessing bradykinesia in Parkinsonian disorders. Mov Disord 2013;4:54. 2. Beudel M, Brown P: Adaptive deep brain stimulation in Parkinson’s disease. Parkinsonism Relat Disord 2016 Jan;22, Supplement 1:S123–S126.