In this paper, we present a novel Bayesian adaptive dual controller (ADC) for autonomously programming deep brain stimulation devices. We evaluated the Bayesian ADC’s performance in the context of reducing beta power in a computational model of Parkinson’s disease, in which it was tasked with finding the set of stimulation parameters which optimally reduced beta power as fast as possible. Here, the Bayesian ADC has dual goals: (a) to minimize beta power by exploiting the best parameters found so far, and (b) to explore the space to find better parameters, thus allowing for better control in the future. The Bayesian ADC is composed of two parts: an inner parameterized feedback stimulator and an outer parameter adjustment loop. The inner loop operates on a short time scale, delivering stimulus based upon the phase and power of the beta oscillation. The outer loop operates on a long time scale, observing the effects of the stimulation parameters and using Bayesian optimization to intelligently select new parameters to minimize the beta power. We show that the Bayesian ADC can efficiently optimize stimulation parameters, and is superior to other optimization algorithms. The Bayesian ADC provides a robust and general framework for tuning stimulation parameters, can be adapted to use any feedback signal, and is applicable across diseases and stimulator designs.

No abstract

No abstract

No abstract

Background Deep brain stimulation (DBS) is a treatment option for Parkinson’s disease patients when medication does not sufficiently manage their symptoms. DBS can be a highly effect therapy, but only after a time-consuming trial-and-error stimulation parameter adjustment process that is susceptible to clinician bias. This trial-and-error process will be further prolonged with the introduction of segmented electrodes that are now commercially available. New approaches to optimizing a patient’s stimulation parameters, that can also handle the increasing complexity of new electrode and stimulator designs, is needed. Methods To improve DBS parameter programming, we explored two semi-automated optimization approaches: a Bayesian optimization (BayesOpt) algorithm to efficiently determine a patient’s optimal stimulation parameter for minimizing rigidity, and a probit Gaussian process (pGP) to assess patient’s preference. Quantified rigidity measurements were obtained using a robotic manipulandum in two participants over two visits. Rigidity was measured, in 5Hz increments, between 10–185Hz (total 30–36 frequencies) on the first visit and at eight BayesOpt algorithm-selected frequencies on the second visit. The participant was also asked their preference between the current and previous stimulation frequency. First, we compared the optimal frequency between visits with the participant’s preferred frequency. Next, we evaluated the efficiency of the BayesOpt algorithm, comparing it to random and equal interval selection of frequency. Results The BayesOpt algorithm estimated the optimal frequency to be the highest tolerable frequency, matching the optimal frequency found during the first visit. However, the participants’ pGP models indicate a preference at frequencies between 70–110 Hz. Here the stimulation frequency is lowest that achieves nearly maximal suppression of rigidity. BayesOpt was efficient, estimating the rigidity response curve to stimulation that was almost indistinguishable when compared to the longer brute force method. Conclusions These results provide preliminary evidence of the feasibility to use BayesOpt for determining the optimal frequency, while pGP patient’s preferences include more difficult to measure outcomes. Both novel approaches can shorten DBS programming and can be expanded to include multiple symptoms and parameters.