A Unified Computational Framework for Deep Brain Stimulation at the Cellular and Network Levels

Abstract

Deep brain stimulation (DBS) has been demonstrated to be a successful therapeutic intervention for neurological disorders, yet the mechanisms underlying its effects on neuronal circuits remain incompletely understood. In this study, we propose a comprehensive phenomenological computational model that accounts for the impact of electrical stimulation parameters on neuronal circuits while incorporating experimentally-validated synaptic and cellular constraints. We investigate how DBS pulses modulate spiking activity in populations of homogeneous neurons representing stimulated nuclei, systematically examining the influence of circuitry architecture, including synaptic connectivity strength (weak vs. strong) and organization (sparse vs. rich). To characterize how DBS-modulated neuronal activity propagates through downstream networks, we develop a simple encoder that reveals distinct encoding patterns arising from different architectural configurations of stimulated nuclei. Furthermore, by connecting stimulated nuclei to recurrently connected neuronal populations, we examine the propagation of DBS-modulated neuronal synchrony across various circuit motifs. Our results demonstrate that three critical factors shape DBS-modulated neuronal activity: (a) the intrinsic synaptic and cellular properties of stimulated nuclei, (b) the architectural organization of stimulated nuclei in terms of synaptic strength and connectivity density, and (c) the circuit motifs formed by postsynaptic targets of stimulated nuclei. This unified model provides a mechanistic framework for understanding DBS representation and propagation in neuronal networks, offering insights that may inform optimization of stimulation parameters for clinical applications. Author summary Computational models of deep brain stimulation have proven to be supremely useful in disentangling the clinical benefits and adverse effects observed in the treatment of a variety of conditions. Despite this, the capacity for many of the existent computational models to account for micro/meso-circuit activation remains limited, as the major techniques rely on detailed characterization of tracts surrounding the DBS electrode, or depend on an non-physiologically constrained injected current intending to mimic the influence of electrical stimulation. The tract based methods only work for tracts that we have detailed characterization of, which are missing for many of the target structures, such as the basal ganglia. Given the restrictions of current methods we set out to define a phenomenological method that is applicable to as many simulation methods as possible, including those with missing details on tractography, while being able to readily integrate results of detailed simulations when available. Our approach is extensible and has examples implemented in some of the most popular computational neuroscience toolkits allowing for ready integration into existing network simulations. Further we demonstrate how this methodology supports interrogation of networks both for physiological responses but also computational dynamics, such as information multiplexing and delayed local evoked potentials.