Perspective: Evolution of Control Variables and Policies for Closed-Loop Deep Brain Stimulation for Parkinson's Disease Using Bidirectional Deep-Brain-Computer Interfaces

Abstract

A deep brain stimulation system capable of closed-loop neuromodulation is a type of bidirectional deep brain-computer interface (dBCI), in which neural signals are recorded, decoded, and then used as the input commands for neuromodulation at the same site in the brain. The challenge in assuring successful implementation of bidirectional dBCIs in Parkinson's disease (PD) is to discover and decode stable, robust and reliable neural inputs that can be tracked during stimulation, and to optimize neurostimulation patterns and parameters (control policies) for motor behaviors at the brain interface, which are customized to the individual. In this perspective, we will outline the work done in our lab regarding the evolution of the discovery of neural and behavioral control variables relevant to PD, the development of a novel personalized dual-threshold control policy relevant to the individual's therapeutic window and the application of these to investigations of closed-loop STN DBS driven by neural or kinematic inputs, using the first generation of bidirectional dBCIs.

Keywords: Parkinson’s disease; beta oscillations; brain-computer interface (BCI); brain-machine interface (BMI); closed-loop neurostimulation; deep brain stimulation; kinematics; subthalamic nucleus.

Figure 1

Angular velocity traces measured during repetitive wrist flexion-extension OFF deep brain stimulation (DBS; A); the insert on the right at higher magnitude demonstrates the severe progressive bradykinesia, ON open-loop STN DBS (olDBS; B) and after 60 min of closed-loop STN DBS (clDBS) (C). Schematic of the DBS lead demonstrates the use of a triple monopole during olDBS and a single monopole during clDBS. V_rms: the root mean square angular velocity averaged over the trial.

Figure 2

Demonstration of experiments performed on the preclinical benchtop system using the Summit application programming interface. Schematic of the fully implanted bidirectional deep brain-computer interface (dBCI) with data from the benchtop experiments. Left-hand panel: the neural input was beta band burst duration from the filtered local field potential; the single threshold control policy decided whether a neural burst was normal or long (pathological), and adapted closed-loop deep brain stimulation (clDBS) by decreasing or increasing stimulation intensity respectively. Right-hand panel: the kinematic input was the shank angular velocity streamed from wearable inertial measurement units; a dual-threshold control policy was based on whether the step was determined to be normal, uncertain, or part of freezing of gait episode and adapted clDBS by either (I) decreasing, not changing, or increasing stimulation intensity, OR, (II) by switching to 140 Hz, staying unchanged or switching to 60 Hz, respectively (right panel).