Website: View Project Website
Grantee: Rutgers - The State University of New Jersey, Newark, NJ, USA Project Summary: The goal of this group was to assess whether new developments in areas of data mining, graph theory, and causal inference might provide better tools for the analysis of brain imaging, particularly functional Magnetic Resonance Imaging (fMRI) data. The diversity of disciplinary backgrounds participating in this effort is symbolic of the breadth and depth of the data processing problems that confront further development and refinement of brain imaging technologies. The best way to understand these problems is to contrast what it means to understand brain activity versus brain interactivity. The brain images we see in the popular press and even in most scientific articles are representations of brain activity. They show areas of the brain that are active when the participant in the experiment is performing an experimental task. These are static images and the activated areas are identified by choosing an activation threshold with areas in the brain above that threshold appearing as patterns of spots in the brain image. Neuroscientists know, however, that the brain is a highly connected organ, where interactions among brain areas occur simultaneously or sequentially, when study participants are performing the experimental task. Thus, to understand the brain we must understand the interactivity, both spatial and temporal, among brain areas. Ideally, we would like to understand the causal relations that occur in activated brain areas: Does activation in area A cause activity in area B, or is B the cause of A? The brain relies on networks to execute tasks, but what is the causal path through the network? The collaborative plans to develop improved simulations of fMRI data, further test and validate their current algorithms, compare their methods with those of dynamic causal modeling, and improve variable selection for developing causal models of brain circuits. In addition, this project will move outside basic research to assess how these tools could be used to identify neural disruptions associated with neuropsychiatric disorders. Finally, they will also produce a handbook for the analysis of brain activity. Over the next three years, this collaborative plans to develop improved simulations of fMRI data, further test and validate their current algorithms, compare their methods with those of dynamic causal modeling, and improve variable selection for developing causal models of brain circuits. In addition, this project will move outside basic research to assess how these tools could be used to identify neural disruptions associated with neuropsychiatric disorders. Finally, they will also produce a handbook for the analysis of brain activity.
Project Lead: Stephen J. Hanson, Ph.D.Clark Glymour (Carnegie Mellon University) and Russell Poldrack (University of California, Los Angeles)
Grant Title: Accessing Brain Interactivity II
Program Area: Understanding Human Cognition
Grant Type: Collaborative Activity Award
Year Awarded: 2009
Duration: 3 years