Development Of Stochastic Dynamical Models For Optimization Of Deep Brain Stimulation In Movement And Neuropsychiatric Disorders

Abstract

The use of electrical deep brain stimulation (DBS), during approximately the last thirty years,has been proven to provide striking benefits for patients with advanced Parkinson’s disease(PD), essential tremor (ET) and dystonia who have failed conventional therapies. In the interim,extended applications of this technique for the treatment of neuropsychiatric disorders haveemerged, including treatment-refractory obsessive-compulsive disorder (OCD), Tourette‘ssyndrome (TS), major depressive disorder (MDD), drug addiction and anorexia nervosa (AN).Challenges however exist and are principally related to the optimization of the efficiency ofstimulation through refined strategies at multiple peri-operative levels. Particularly, in additionto appropriate patient selection and anatomical target determination, the outcome of DBS maybe strongly influenced by the quality of post-operative clinical management, i.e. the adjustmentof stimulation parameters and the selection of the optimal contact, usually over periods of weeksto months. This trial-and-error procedure entails important disantvantages: (a) it may notnecessarily yield the optimal therapeutic window (i.e. the ratio of the threshold for stimulation-induced side-effects to the induction threshold for therapeutic benefit) (b) it is considerablytime-consuming, while movement and neuropsychiatric disorder symptoms may fluctuate overtime-scales of seconds to days (c) the open-loop nature and monomorph pattern of standardhigh-frequency stimulation (regular, 130Hz) appears chronically to favor tolerance/habituationphenomena, while being associated with maximal energy consumption. Against thisbackground, closed-loop neuromodulation is emerging as a more robust alternative and one ofthe most promising breakthroughs in the field of DBS. Principally, any algorithm designed for amaximally efficient closed loop DBS system shall conceptually satisfy two core specifications:the reliable assessment of optimal biomarkers for feedback control that capture the patient‘sclinical state in real time and the identification of temporally alternative stimulation protocolsthat may be more therapeutically- and energy-efficient compared with the conventional patternof stimulation.In the framework of the current doctoral dissertation, stochastic dynamical models for theoptimization of the clinical outcome of DBS in movement and neuropsychiatric disorders are being developed. The ultimate goal is to algorithmically design a closed-loop DBS system foradvanced PD and treatment-refractory OCD, ensuring optimal performance in terms of bothefficiency and selectivity of stimulation, as well as in terms of computational speed. The mainhypothesis we build upon is that temporally alternative DBS waveforms hold the potential todrive the neuronal dynamics within the basal ganglia back to the normal desynchronized state,thereby outperforming the action of standard DBS waveforms, the mechanism of which hasbeen principally attributed to the reinforcement-driven regularization of neural firing patterns inthe vicinity of the stimulated nucleus. On grounds of a stochastic phase model describing anensemble of globally coupled chaotic oscillators driven by common noise, independent noise,and external forcing, and fitted to subthalamic MERs acquired during surgical interventions forPD and OCD, we first prove that the desynchronizing and probably also the therapeutic effect oflow-frequency stochastic DBS waveforms may be significantly stronger compared with theeffect of standard stimulation. Subsequently, relying upon a series of methods robust to thepresence of measurement noise, we assess the presence of significant nonlinear couplingbetween beta and high-frequency subthalamic neuronal activity, as a biomarker for feedbackcontrol in the proposed closed-loop neuromodulation scheme, and further present a strategy,incorporating the application of a modified version of the stochastic phase model (phase-reduced bursting neuron model) and a derivative-free optimization algorithm (direct searchoptimization based on quadratic modeling), through which optimal stochastic patterns andparameters of stimulation for minimum energy desynchronizing control of neuronal activity arebeing identified. Cross-frequency coupling proves to be a consistently appropriate biomarker forfeedback control in case of PD, but may display subject-specific applicability in case of OCD.We demonstrate the ability of the presented modeling approach to identify, at a relatively lowcomputational cost, stimulation settings associated with a significantly higher efficiency andselectivity compared with stimulation settings determined during post-operative clinicalmanagement of patients with advanced PD and treatment-refractory OCD. Together, our dataprovide strong evidence for the applicability of computational neurostimulation to real-time,closed-loop DBS systems for movement and neuropsychiatric disorders.