CLOSED-LOOP OPTOGENETIC CONTROL OF CORTICAL POPULATION DYNAMICS USING HIGH-RESOLUTION PATTERNED LIGHT STIMULATION IN AWAKE MICE

Cognitive functions emerge from precisely coordinated neuronal population activity unfolding on millisecond timescales. To causally probe and manipulate these spatiotemporal dynamics, we developed a closed-loop optogenetic platform combining high-resolution patterned photostimulation with real-time electrophysiological feedback. Using a digital micromirror device (DMD), we delivered grid-patterned illumination with 20 μm spatial and 0.2 ms temporal resolution across a 2-mm cortical area in Thy1-ChR2-EYFP mice while simultaneously recording local field potentials (LFPs) from the primary somatosensory cortex.

Spatially distinct grid patterns (3×3 configuration) evoked distinguishable LFP responses with ~75% classification accuracy using convolutional neural networks. Classification errors decreased with increasing spatial separation between stimulation sites, confirming spatial specificity of evoked responses. Rapid temporal sequences (100×100 grids at 50 Hz) generated more consistent trial-to-trial responses than conventional constant photostimulation, with forward versus reverse sequences discriminable at >90% accuracy. Sequential stimulation produced higher trial-to-trial correlations compared to static patterns, indicating that temporal structure constrains network variability.

We then implemented a closed-loop feedback system that iteratively adjusts stimulation patterns based on evoked LFP responses. The algorithm minimizes discrepancies between evoked and target waveforms through gradient-based optimization of the spatial stimulation pattern. Closed-loop optimization demonstrated progressive convergence toward target patterns across iterations, with optimized patterns producing lower error than randomly generated controls. Overall, this platform provides a method for testing how specific spatiotemporal activity patterns contribute to cortical function.