BCIwiki (bciwiki.com) — A collaboration involving Peking University, China, posted a preprint to bioRxiv on September 28, 2026. It compared visual cortical responses evoked by natural vision and patterned optogenetic stimulation in awake mice, reconstructing natural movie frames from signals in both conditions. The study has not been peer reviewed. It aims to quantitatively evaluate how visual images should be translated into cortical stimulation patterns.
The team co-expressed the calcium indicator GCaMP6s and the red-shifted, ultra-sensitive opsin rsChRmine-oScarlet in the primary visual cortex (V1). They translated dynamic natural movie frames into spatiotemporal patterns of optogenetic stimulation, which uses light-sensitive proteins to let light modulate neural activity. Quantitative comparisons of macro-scale dynamics showed that artificial stimulation evoked cortical states highly comparable to those driven by natural visual stimulation, with aligned representations.
To evaluate the visual information retained in these responses, the team developed STAR, a deep learning model with spatial and temporal attention mechanisms. It reconstructed natural movie frames from V1 signals during both natural viewing and artificial stimulation. The authors interpret the findings as evidence that complex sensory information can be naturally encoded or synthetically introduced into V1 circuits and decoded. They frame the work as an empirical and computational proof of concept for closed-loop biomimetic encoders, cortical visual neuroprostheses, and bidirectional brain-machine interfaces. These mouse experiments concern neural encoding and reconstruction; they do not establish corresponding visual experiences or restored human vision.