Trial-by-trial inter-areal interactions in visual cortex in the presence or absence of visual stimulation

State-of-the-art computational models of vision focus on fitting trial-averaged spike counts triggered by visual stimuli using overparameterized neural networks. Understanding visual cortex requires predicting the dynamic responses of neurons within and across areas, in single trials, and across different experimental conditions. Here we investigated trial-by-trial inter-areal interactions in the visual cortex by predicting neuronal activity in one area based on activity in another area, including stimulus-driven and non-stimulus-driven activity. We analyzed two datasets: calcium imaging from mouse V1 layers 2/3 and 4, and extracellular neurophysiological recordings from macaque V1 and V4. Neuronal activity could be predicted bidirectionally between L2/3 and L4 in mice, and between V1 and V4 in macaque monkeys, even in the absence of visual stimuli, with distinct contributions for bottom-up and top-down interactions. In mice, we observed a bimodal distribution of neurons, with some neurons primarily driven by visual inputs and others showing predictable activity during spontaneous activity despite lacking consistent visually evoked responses. Predictability depended on the stimulus type, intrinsic neuronal properties, receptive field overlap, and the relative timing of activity across areas. These findings provide initial steps towards computational models that capture the dynamic transformation of neuronal responses throughout visual cortex.

Neuronal activity in one visual cortical area can be predicted from another area in single trials, with or without visual stimulation, in both mice and macaques.