DECODING INFORMATION PROPAGATION FROM V1 TO V4 DURING ELECTRICAL MICROSTIMULATION FOR VISUAL PROSTHESIS OPTIMIZATION

Electrical microstimulation of visual cortex can evoke phosphenes in blind individuals, but optimal stimulation parameters remain unknown. We hypothesize that effective stimulation should produce not only strong responses in V1 but distinctive activity patterns in higher visual areas (V4). We developed a decoding-based framework to assess information propagation during microstimulation in awake macaques, using downstream neural activity as a proxy for percept generation.
We recorded simultaneously from V1 (32-channel ultra-flexible electrodes by ReVision, KU Leuven) and V4 (64-channel Utah array) during four paradigms: single-pulse stimulation with monopolar and bipolar electrode configurations, pulse-trains at varying intensities, visual stimulation with tDCS and combined protocols. Artifact removal procedures were compared in all datasets using the single-pulse dataset as baseline in order to thoroughly characterize the artifact.
Configuration strongly influenced V1-to-V4 propagation: bipolar stimulation produced net excitatory V1 responses with short latencies that propagated to V4 with longer delays, while monopolar configurations showed different temporal profiles. These inter-area delays were also compared to the literature for validation.
Our decoding paradigm addresses two main questions: can V4 activity patterns distinguish stimulation parameters applied to V1, and, how does incorporating V1 activity improve discrimination? Successful decoding indicates effective information transfer through cortical circuits. This framework provides quantitative metrics for optimizing stimulation strategies in cortical visual prosthesis, using information content in downstream perceptual areas as the target criterion rather than local and remote response magnitude.