Stephen L. Schmidt

As a step towards advancing adaptive DBS control for Parkinson’s disease, we have developed an automated algorithm that detects tremor continuously on a seconds-resolution time scale from a wearable accelerometer and present the feasibility study test results. Triaxial acceleration data were wirelessly streamed from an Apple Watch as well as acquired from an internal accelerometer in the implanted DBS device itself. The algorithm first determines if there is any high-power voluntary activity, such as walking, using the arm, or transitioning from sitting to standing; then, it identifies peaks in the 4–7 Hz Parkinsonian tremor frequency band. Peak detection for tremor activity was more accurate using the Apple Watch than the IPG. Peak and harmonic detection were also more accurate using continuous wavelet transforms than short-time Fourier transform. According to the repeated measures correlation, our detection algorithm correlated strongly with DBS intensity (Subject RZCH: r = −0.93, p = 3.6 × 10⁻⁵; 6KOZ: r = −0.97, p = 1.6 × 10⁻⁵, NU5U: r = −0.94, p = 0.057). Pearson’s correlation coefficient between our tremor detection algorithm and the currently accepted industry metric was found to be 0.57 (t-value = 8.5, dof = 148, p < 1 × 10⁻⁶). Our algorithm is distinctive in the capability to detect Parkinsonian tremor, with a high degree of clinical relevance, during daily living activities and is able to discriminate tremor from walking, using a convenient, commercial wrist-worn accelerometer.

Background Deep brain stimulation (DBS) reduces the motor symptoms of Parkinson’s disease. The two most common targets are the subthalamic nucleus and the globus pallidus. Dual target deep brain stimulation may better reduce symptoms and minimize side effects, but the optimal parameters of dual target deep brain stimulation and their potential interactions are unknown. Objective Our purpose was to quantify the frequency response of dual target DBS on bradykinesia and beta oscillations in participants with Parkinson’s disease, and to explore intrahemispheric pulse delays as a means to reduce total energy delivered. Methods We applied dual target DBS using the Summit RC+S in six participants, varying deep brain stimulation frequency. Results Dual target DBS at 50 Hz was effective at reducing bradykinesia, whereas increasing deep brain stimulation frequency up to 125 Hz also significantly reduced beta power. This frequency effect on beta power was replicated in a biophysical model. The model suggested that 22 Hz dual target deep brain stimulation, with an intrahemispheric delay of 40 ms, can reduce beta power by 87%. Conclusion We conclude that dual target DBS at 125 Hz best reduced bradykinesia. However, low frequency DBS with an appropriate intrahemispheric delay could improve symptom relief.

Continuous deep brain stimulation (cDBS) of the subthalamic nucleus (STN) or globus pallidus (GP) is an effective treatment for the motor symptoms of Parkinson’s Disease (PD). The relative benefit of one region over the other is of great interest but cannot usually be compared in the same patient. Simultaneous DBS of both regions may synergistically increase the therapeutic benefit. CDBS is limited by a lack of responsiveness to dynamic, fluctuating symptoms intrinsic to the disease. Adaptive DBS (aDBS) adjusts stimulation in response to biomarkers to improve efficacy, side effects, and efficiency. We combined bilateral DBS of both STN and GP (dual target DBS) in a prospective within-participant, clinical trial in six PD patients (n = 6, 55 to 65 years, two women). Dual target cDBS was tested for PD symptom control annually over 2 years, measured by motor rating scales, ON time without dyskinesia, and medication reduction. Random amplitude experiments probed system dynamics to estimate parameters for aDBS. We then implemented proportional-plus-integral (PI) aDBS using a novel distributed (off-implant) architecture. In the home setting we collected tremor and dyskinesia scores as well as individualized beta and DBS amplitudes. Dual target cDBS reduced motor symptoms as measured by UPDRS to a greater degree than either region alone (p < 0.05, Linear Mixed Model) in the cohort. The amplitude of beta oscillations in the STN correlated to the speed of hand grasp movements for five of six participants (p < 0.05, Pearson correlation). Random amplitude experiments provided insight into temporal windowing to avoid stimulation artifacts and demonstrated a correlation between STN beta amplitude and DBS amplitude. PI control of aDBS reduced average power, while preserving Unified Parkinson’s Disease Rating Scale Motor Examination (UPDRS III) scores in the clinic (p = 0.28, Wilcoxon signed rank), and tremor and dyskinesia scores during blinded testing at home (n =3, p > 0.05, Wilcoxon ranked sum). In the home setting, DBS power reductions were slight but significant. Dual target cDBS may offer an improvement in treatment of motor symptoms of PD over DBS of either the STN or GP alone. When combined with PI aDBS, stimulation power may be reduced while preserving the increased benefit of dual target DBS.

Deep brain stimulation (DBS) has shown great promise toward treating motor symptoms caused by Parkinson's disease (PD), by delivering electrical pulses to the Basal Ganglia (BG) region of the brain. However, DBS devices approved by the U.S. Food and Drug Administration (FDA) can only deliver continuous DBS (cDBS) at a fixed amplitude; this energy inefficient operation reduces battery lifetime of the device, cannot adapt treatment dynamically for activity, and may cause significant side-effects (e.g., gait impairment). In this work, we introduce an offline reinforcement learning (RL) framework, allowing the use of past clinical data to train an RL policy to adjust the stimulation amplitude in real time, with the goal of reducing energy use while maintaining the same level of treatment (i.e., control) efficacy as cDBS. Moreover, clinical protocols require the safety and performance of such RL controllers to be demonstrated ahead of deployments in patients. Thus, we also introduce an offline policy evaluation (OPE) method to estimate the performance of RL policies using historical data, before deploying them on patients. We evaluated our framework on four PD patients equipped with the RC+S DBS system, employing the RL controllers during monthly clinical visits, with the overall control efficacy evaluated by severity of symptoms (i.e., bradykinesia and tremor), changes in PD biomakers (i.e., local field potentials), and patient ratings. The results from clinical experiments show that our RL-based controller maintains the same level of control efficacy as cDBS, but with significantly reduced stimulation energy. Further, the OPE method is shown effective in accurately estimating and ranking the expected returns of RL controllers.

Deep brain stimulation (DBS) has shown great promise toward treating motor symptoms caused by Parkinson's disease (PD), by delivering electrical pulses to the Basal Ganglia (BG) region of the brain. However, DBS devices approved by the U.S. Food and Drug Administration (FDA) can only deliver continuous DBS (cDBS) stimuli at a fixed amplitude; this energy inefficient operation reduces battery lifetime of the device, cannot adapt treatment dynamically for activity, and may cause significant side-effects (e.g., gait impairment). In this work, we introduce an offline reinforcement learning (RL) framework, allowing the use of past clinical data to train an RL policy to adjust the stimulation amplitude in real time, with the goal of reducing energy use while maintaining the same level of treatment (i.e., control) efficacy as cDBS. Moreover, clinical protocols require the safety and performance of such RL controllers to be demonstrated ahead of deployments in patients. Thus, we also introduce an offline policy evaluation (OPE) method to estimate the performance of RL policies using historical data, before deploying them on patients. We evaluated our framework on four PD patients equipped with the RC+S DBS system, employing the RL controllers during monthly clinical visits, with the overall control efficacy evaluated by severity of symptoms (i.e., bradykinesia and tremor), changes in PD biomakers (i.e., local field potentials), and patient ratings. The results from clinical experiments show that our RL-based controller maintains the same level of control efficacy as cDBS, but with significantly reduced stimulation energy. Further, the OPE method is shown effective in accurately estimating and ranking the expected returns of RL controllers.