Adaptive Parameter Modulation of Deep Brain Stimulation in a Computational Model of Basal Ganglia-thalamic Network

Deep brain stimulation (DBS) has proven to be an effective treatment for Parkinson’s disease (PD). Adaptive control strategies offer the potential to improve efficacy, limit side effects and save battery consumption via reducing the total amount of stimulation delivered. However, the mechanisms underlying the beneficial effects of DBS for PD remain poorly understood and are still under debate, which has hindered the development of closed-loop DBS. And during the chronically implanted phase, adaptive DBS needs to be further improved to maintain its advantages. In the design of new adaptive DBS, more insights into inaccuracies when establishing mathematical basal ganglia model, unknown external disturbance signal and dynamics of focal area should be considered. A controlled auto-regressive moving average is used as the representative description of stimulus–response relationship based on recursive extended least square method, where stimulation signal is applied to subthalamic nucleus (STN) and the feedback signal is selected as local field potential signal from internal segment of the globus pallidus (GPi). The generalized minimum variance algorithm is used for online update of stimulation frequency and amplitude in a closed-loop manner. Simulation results illustrated the efficiency of the proposed closed-loop stimulation methods in disrupting the aberrant beta-band (12–35 Hz) oscillation and restoring normal firing pattern when compared with the PD state. Robustness of the adaptive algorithm was further verified through dynamic change of illness state.