Frontal cortical contributions to ventral stream visual processing

Project Summary Vision must transform patterns of light hitting the retina into an understanding of objects, their spatial relationships, and the surrounding scene context. This process of high-level information extraction is thought to occur in the ventral visual stream, a cortical pathway whose disruption is associated with face and object agnosias and visual memory loss. Both experimental neuroscience and theoretical work suggest that the traditional, purely feedforward model of the ventral stream fails to account for high-level vision in real-world scenarios. I hypothesize that visual cognitive information sent from the frontal cortex (FC) back to the ventral stream is critical for the remarkable robustness of primate vision to complex or ambiguous sensory inputs. The work proposed here examines interactions between the inferotemporal cortex (IT), the final station of the ventral stream, and two neighboring visually-responsive regions of FC: the frontal eye field (FEF), which controls eye movements and plays a role in visual spatial selection, and ventrolateral prefrontal cortex (PFC), which encodes abstract semantic object categories. This research plan leverages advantages of the marmoset, a small monkey with advanced visual behaviors and a lissencephalic brain allowing for application of experimental techniques that are intractable in the macaque. Aim 1 will use high-density electrophysiology probes and naturalistic stimuli to systematically characterize coding properties of visual FC. Aim 2 will use widefield calcium imaging, electrophysiological recordings, and electrical microsimulation to produce both a coarse map and a detailed cellular- and population-level account of how FC signals influence IT encoding. Finally, Aim 3 will use deep neural networks to model the influence of FEF and vlPFC on IT as a means for generating new hypotheses about biological vision. This project will expand understanding of the computational and cortical circuit bases of high-level vision, providing foundational knowledge of visual brain regions that are disrupted in neurological disorders such as frontotemporal dementia and autism. The training plan laid out in this proposal helps me acquire new experimental skills in optical imaging and multi- area recordings, theoretical skills in deep learning models, and scientific fluency in high-level vision. My sponsor, Dr. Elias Issa, will provide expertise in computational modeling and experimental work with the marmoset. My training will benefit from my co-sponsor, Dr. Michael Shadlen, who has decades of experience studying high-level associative cortex in the primate, and my collaborator on optical imaging, Dr. Aniruddha Das, an expert in mesoscale functional mapping. This work will be conducted at Columbia University’s Zuckerman Institute, a world-class neuroscience research and training institution whose faculty specialize in approaches ranging from molecular to systems to theoretical.