TOWARD RELIABLE VISUAL PROSTHETIC STIMULATION UNDER HIGHLY VARIABLE SPONTANEOUS ACTIVITY

State-dependent excitability and highly variable spontaneous activity pose a fundamental challenge for neuroprostheses: they reduce the reliability of responses, which are likely to lead to percepts with lower effective contrast. Identical external stimuli can evoke variable or absent percepts due to altered cortical excitability, which is amplified in blindness [1]. Here, we present a biophysically grounded neural and optogenetic model within a closed-loop control framework, designed to evaluate the effect of noise on the efficiency of driving the firing rate of a cortical column in the model along a predetermined time course.
We use a large-scale spiking network model of higher mammalian primary visual cortex (V1) [2], coupled with a detailed model of optogenetic stimulation [3], incorporating cortical light dispersion and channelrhodopsin dynamics. We use this framework to compare open-loop stimulation with several representative closed-loop regulator algorithms, such as PID, LMS, LQR. For all regulators, we observe lower tracking error, response variability and response latency compared to the open loop protocol, while differences between the individual selected regulators appear minimal. These results suggest that closed-loop control control can regulate cortical population activity under high spontaneous variability, highlighting the utility of this framework for the development of visual prosthetics.
[1] Gothe et al. (2002). Brain. doi:10.1093/brain/awf045
[2] Rózsa et al. (2026). Nature Communications. doi:10.1038/s41467-026-68578-y
[3] Antolík et al. (2021). Scientific Reports. doi:10.1038/s41598-021-88960-8