Retinotopy constrains the topology of neural manifolds in macaque visual cortex

Abstract

Cortical population activity often evolves along low-dimensional manifolds, shaped by external activity, task-demands and the underlying neuronal network structure. The structure of population activity has been studied across many brain areas and species. However, the influence of retinotopic organization in the structure of population activity of the primate visual cortex in vivo remains understudied. Here, we show that retinotopy in primary visual cortex (V1) determines the topological structure in neural population activity in the macaque. We analysed large-scale extracellular recordings from two macaques implanted with 896-channel Utah arrays in V1 performing a receptive field mapping task, with high-contrast moving bars. Isometric Mapping (isomap) applied to the neural population activity revealed a loop-like manifold. Persistent homology on 10-dimensional embeddings confirmed a single dominant one-dimensional loop. To test whether retinotopy alone can account for this structure, we constructed a minimal neural response model and simulated activity with the same stimuli as the macaque neural data. The model reproduced the loop-like topology observed in the brain. Taken together, the experimental and simulated data show that retinotopic maps naturally induce loop-like neural manifolds in response to moving stimuli, directly linking travelling waves and rotational dynamics. Our results provide a basis for further analysis on how neuronal populations encode and process visual information at the population level.