Introduction: Current clinically available deep brain stimulation (DBS) therapies for Parkinson’s disease (PD) are all open-loop and are unable to adapt to the ever-changing patient, medication, and disease states. A closed-loop DBS system that utilizes appropriate physiological control based on patients’ predicted behavior may improve therapeutic results. In this study, we develop a behavior classification method robust to stimulation to label patient actions even in the presence of therapy. Furthermore, we investigate the effect of medication and stimulation on the increased beta power associated with PD.

Methods: Three participants were implanted with DBS leads in the subthalamic nucleus (STN). During two data collection sessions from the implanted leads, one “on” medication and one “off” medication, the participants were cued to perform a series of 60 “button press” then “reach” actions with and without therapeutic stimulation. We transformed the bipolar re-referenced local field potentials (LFP) into their time-frequency representation and used the beta frequency range (13-30Hz) as input to a support vector machine (SVM) classifier. Additionally, we used Welch’s power spectral density (PSD) estimate to evaluate the effect of the medication and stimulation on the beta power of LFPs.

Results: We obtained a classification accuracy of 87%, 85%, and 87% for stimulation “off”, “on”, and “combined” data sets using a SVM classifier. An analysis of variance (ANOVA) for the PSDs of the four combinations of stimulation “on/off” and medication “on/off” show beta power is suppressed significantly when the patients take medication (p-value<0.002) or receive therapeutic stimulation (p-value<0.0003).

Conclusions: The results show that STN-LFPs contain useful information for human behavior recognition. The high-frequency stimulation pulse (~140 Hz) had limited impact on the classification performance. This is a precursor for designing the next generation of closed-loop DBS systems.

Patient Care: Current clinically available deep brain stimulation (DBS) therapies for Parkinson’s disease (PD) are all open-loop and are unable to adapt to the ever-changing patient, medication, and disease states. A closed-loop DBS system that utilizes appropriate physiological control based on patients’ predicted behavior may improve therapeutic results.

Learning Objectives: The possibility of using STN-LFP signals for human behavior recognition.

References: [1] A. O. Hebb, F. Darvas, and K. J. Miller, “Transient and state modulation of beta power in human subthalamic nucleus during speech production and finger movement,” Neuroscience., vol. 202, pp. 218-233, 2012. [2] H. M. Golshan, A. O. Hebb, S. J. Hanrahan, J. Nedrud, and M. H. Mahoor, “A hierarchical structure for human behavior classification using STN local field potentials,” Journal of Neuroscience Methods., vol. 293, pp. 254-263, 2018.