Skip to main content

Neural Mechanisms of Visual Perception and Attention

Final Report Summary - NEURO-VPA (Neural Mechanisms of Visual Perception and Attention)

Project NEURO-VPA (CIG13-GA-2013-618101) investigates the neural mechanisms of visual attention in the human brain. Attention is a neural control mechanism that optimizes behavior by altering the processing of sensory information. Thus understanding the effects of attention on sensory representations in a critical problem in systems neuroscience. This proposal investigates if and how high-level attention directed to visual categories changes tuning for a broad range of semantic visual features across the human brain. Towards this goal, high-resolution neural responses are recorded in the neocortex using functional MRI, and neural responses are then analyzed via a powerful computational modeling framework.

The project consists of two specific objectives regarding cortical representations and neural mechanisms of visual perception and visual attention. Objective 1 is to assess semantic representations of natural visual scenes during passive viewing and to reveal how category-based attention to daily-life objects and actions affects these representations. Objective 2 is to determine the computational principles that govern the attentional modulations in semantic representations during natural visual search.

Work Performed
Experimental work and subsequent modeling analyses have been performed to address both of the main objectives. Together with his collaborators, the PI has developed novel experimental paradigms that combined natural movie stimuli with passive-viewing and natural visual search tasks. By devising advanced statistical techniques to increase sensitivity in computational modeling of single voxel responses, he measured selectivity for thousands of low-level structural features, object and action categories, and co-occurrence statistics of objects and actions that appear in natural movies. Using these techniques, the PI examined attentional changes in voxel-wise tuning, in patterns of functional connectivity between visual and higher brain regions, and attentional biases in tuning modulation during category-based visual search. A project website has also been launched

The project has revealed neuroscientific insights in several important lines. The results indicate that classical definitions of scene-selective functional regions-of-interest (ROIs) in the human brain overestimate the homogeneity of semantic representations; both semantic representations and functional connectivity across ventral-temporal and frontal cortices are significantly more task-dependent than previously assumed; attentional modulations exhibit nonlinear behavior for brain regions that are strongly selective for a specific category during passive viewing; and semantic representations across neocortex during divided attention to multiple semantic categories can be linearly predicted from isolated attention to single categories. These results have led to 2 journal publications, 6 conference publications and 3 M.S. theses.

Final Results and Impact
The project results have enabled a comprehensive assessment of the effects of high-level visual search tasks on semantic representations of natural visual stimuli. The conducted experiments have collectively advanced our understanding of the neural mechanisms of natural visual search and the computational principles that govern attentional modulations of visual representations.

Therefore, the results of the project have been of critical importance for understanding human vision, and for explaining how the brain uses limited resources to create flexible representations of the world. The project results have future implications for developing appropriate treatments and interventions for attention-related disorders that impede learning and behavior.