Research · Area of focus
Visual Cortical Processing
How do neural circuits enable visual recognition, tracking, and scene segmentation? We use large-scale recordings and targeted optical perturbations in rodents — and higher species — to dissect how cortical microcircuits transform sensory input into perception.
Selected papers
Five papers that define our work on visual cortical processing.
-
The logic of recurrent circuits in the primary visual cortex
Ensemble-specific two-photon optogenetics in the mouse visual cortex isolated the influence of recurrent cortical activity from that of external visual input. The spatial arrangement and feature preference of a stimulated ensemble jointly determined its net effect: photoactivation suppressed cells beyond 30 µm but uniformly activated closer, similarly tuned cells, while among non-similarly tuned cells compact co-tuned ensembles drove suppression and diffuse co-tuned ensembles drove activation. Computational modeling suggested that highly local recurrent excitatory connectivity together with selective convergence onto inhibitory neurons explains this logic.
-
Feature-tuned synaptic inputs to somatostatin interneurons drive context-dependent processing
Single-cell and population-level physiological recordings were combined with perturbation methods to map a context-dependent cortical computation onto the synaptic microarchitecture of mouse primary visual cortex. The work demonstrated a precise like-to-like pattern of synaptic connectivity from cortical pyramidal cells onto somatostatin (SST) inhibitory interneurons. This connectivity rule explains the visual encoding properties of SST cells and the context-driven figure/ground modulation they support, which may underlie the earliest stages of scene segmentation.
-
Cortical gamma band synchronization through somatostatin interneurons
Most circuit models hold that soma-targeting parvalbumin interneurons are the essential subtype for gamma rhythms, but cell-type-specific optogenetic manipulations in behaving animals showed that dendrite-targeting somatostatin (SOM) interneurons are critical for a visually induced, context-dependent gamma rhythm in visual cortex. A computational model independently predicted this dependence, and further experiments showed SOM neurons are required for long-distance coherence across the visual cortex. By operating through dendritic rather than only somatic inhibition, SOM-mediated oscillations establish an alternative mechanism for synchronizing distributed cortical networks.
-
A neural circuit for spatial summation in visual cortex
In contrast to pyramidal cells, somatostatin-expressing inhibitory neurons (SOMs) in the superficial layers of mouse visual cortex increase their responses when the receptive-field surround is stimulated. This difference results from the preferential excitation of SOMs by horizontal cortical axons. Perturbing SOM activity showed that these neurons contribute to surround suppression in pyramidal cells, establishing a cortical circuit for surround suppression and attributing a specific function to a genetically defined inhibitory neuron type.
-
Lateral competition for cortical space by layer-specific horizontal circuits
Horizontal projections interconnect neighboring cortical domains and allow sensory features to be processed in a context-dependent manner, but how they coordinate activity across domains was poorly understood. Selectively activating horizontal projection neurons in mouse somatosensory cortex revealed that these projections suppress superficial layers while simultaneously activating the deeper cortical output layers. This layer-specific modulation arises not from a spatial separation of excitation and inhibition but from a layer-specific ratio between the two opposing conductances, a mechanism through which cortical domains compete for cortical space.