State-space optimal feedback control of optogenetically driven neural activity

Abstract Objective. The rapid acceleration of tools for recording neuronal populations and targeted optogenetic manipulation has enabled real-time, feedback control circuits in the brain. Continuously-graded measured activity poses a wide range technical challenges, which we address through combination stimulation state-space optimal framework implemented thalamocortical circuit awake mouse. Approach. Closed-loop neurons was performed real-time via channelrhodopsin-2 expressed somatosensory thalamus head-fixed A linear dynamical system model structure used to approximate light-to-spiking input-output relationship both single-neuron as well multi-neuron scenarios when from multielectrode arrays. These models were utilized design state controller gains by way quadratic also online estimation feedback, where parameter-adaptive Kalman filter provided robustness model-mismatch. Main results. This model-based scheme proved effective firing rate animals. Notably, graded optical actuation here did not synchronize simultaneously recorded neurons, but heterogeneity across population resulted varied response stimulation. Simulated multi-output better heterogeneous demonstrated how approach generalizes beyond applications. Significance. To our knowledge, this work represents first experimental application space In with control, approaches laid out tested should generalize future problems involving highly complex neural circuits. More generally, opens door adaptively interacting dynamics underlying sensory, motor, cognitive signaling, enabling deeper understanding function ultimately face injury or disease.