Functional-neuroanatomy of high-level visual cortex: a quantitative multimodal approach

Stanford University, Stanford CA

Investigators

Linked publications & trials

Abstract

PROJECT SUMMARY Humans rapidly comprehend the continuous visual input - perceiving who are the people, what are their actions, and where are they. These percepts emerge from computations across ventral, lateral, dorsal streams, respectively. However, it is unknown how the interplay between brain structure, function, and computation supports these visual behaviors. Leveraging advancements from the prior funding period, we propose a unique multimodal and computational approach to answer this question. The research will focus on ventral and lateral streams, as the ventral is the most understood and the lateral is the least understood. Aim 1 will elucidate the interplay between function, white matter connections, and cytoarchitecture using functional, anatomical, and diffusion MRI, as well as cytoarchitectonic data. We will quantitatively measure the relation between functional and structural data, develop computational models linking these metrics, and evaluate their predictivity in left out data. Aim 2 will determine the nature of basic spatiotemporal computations in the visual system and how they affect visual capacity. Aim 2a, will use fMRI and novel spatiotemporal population receptive field (ST-pRF) modeling to determine how basic spatiotemporal computations vary with stimuli and attention. Aim 2b, will use fMRI and behavioral measurements to measure neural and visual capacity, respectively, to multiple stimuli presented sequentially or simultaneously under different attention conditions. Using ST-pRFs, we will determine what spatiotemporal computations predict neural capacity, and then develop a linking model predicting from brain responses visual capacity. Aim 3 will use exciting new topographic deep neural networks (TDANNs) together with massive fMRI datasets of brain responses to visual stimuli to determine how behavioral goals and structural constraints affect the function and spatial organization of the human visual system. This will not only shed important light into the utility of implementational factors, but also generate a single model linking structure, function, and visual behavior. The research will (i) significantly advance understanding of the interplay between visual function, structure, and computations in multiple visual streams filling in longstanding empirical and theoretical gaps, (ii) generate new computational and theoretical understanding how basic spatiotemporal computations and structural constraints affect visual processing, and (iii) generate innovative open source empirical and computational methodologies. This research has important health ramifications for developing noninvasive diagnostics for clinical conditions associated with malfunction of high-level visual cortex including prosopagnosia, action agnosia, and dyslexia, and will provide a new computational testing bed for developing cortical visual prosthetics.