Stephen Jose Hanson, Ph.D. - US grants

EIA-0205178
Paul Kantor
Rutgers University

Novel Modeling of Dynamic Brain Images for Indexing and Retrieval

The proposed research will develop new techniques for representation, organization and retrieval of dynamic brain images. High resolution brain imaging facilities at two sites will conduct targeted experiments to develop image sequences supporting the research. Multiple techniques will be used to provide higher time resolution than the native fMRI format. Images will be archived with web access for the research team, and for other teams elsewhere around the world. The research will develop novel models of episodes of brain activity and effective schemes for indexing and retrieving those episodes.

A practical research goal of the project is to support a new kind of "Query by Example" access to archives of space-time brain images, usable by clinicians and researchers for research and diagnosis. The project also contributes to clinical applications of cognitive science, to the advance of scientific knowledge, and to scientific education, at the undergraduate, graduate and post-graduate levels. The Rutgers-Princeton project is a collaboration among cognitive scientists, computer scientists and engineers, information scientists, applied mathematicians, statisticians, and cognitive neuroscientists, with guidance from an advisory committee of clinicians and radiologists.

The algorithms developed will better approximate judgments that trained humans make when scanning an array of stored images for ones "similar to" a given image, that is, "Query by Example". This project will help to make neuro-imaging analysis tools available to the clinical and industrial investigators who are best positioned to adapt new methods to applied domains.

Principal investigator/Program Director (Last, first, middle): Hanson, Stephen, José RFA-EB-15-006 Project Summary/Abstract Since the earliest days of neuroscience research, core methods have focused on matching specific functions to local brain structure and neural activity. The relationship between brain structure and function has been a key motivation for the development and application of novel methods and discovery. Despite the apparent success of this program in identifying brain areas associated with memory, attention, executive control, action- perception, language, etc.. it is typical for many other areas to be engaged during basic cognitive/perceptual tasks, areas that are often considered ?background,? ?secondary? or often just irrelevant and are consequently ignored. Given the fundamental nature of the connectivity in brain, theories of cognitive neuroscience will very likely involve hypotheses about the influence?sometimes called ?effective connectivity? (Friston et al, 1994, Sporns, 2011)--- that one brain area may have upon another in the course of basic mental processes. Whether we consider language processing, working memory or simple detection tasks, cognitive and perceptual processes are likely to include networks of regions that operate interactively to define, both, a distributed as well as a kind of local computation. It has become increasingly common to posit that networks, circuits, or clusters of brain areas communicate with one another in the implementation of various potential social or social-perceptual functions. Many of these hypothesized networks are thought to be organized around hubs that synchronize other areas but are neither exclusive, nor necessary and sufficient, for a given function. Part of this apparent flexibility of brain networks can be attributed to continued ambiguity about the components or particular function of a given network. For example, many of the brain networks associated with social functioning, include similar function, similar areas, and overlapping networks. As social/affective and cognitive neuroscience continues to evolve it will be more and more critical to disentangle these networks in order to identify the role that individual networks play in various social, perceptual and cognitive function. Unfortunately, the muddle of networks and their functions has increased rather than decreased in recent years. The field of social and cognitive neuroscience has evolved to a point where principled methods for identifying network connectivity, and the tools to do so, could well be trans-formative but certainly are urgent. In this proposal we aim to advance the development of a novel framework based on a model of effective connectivity and Bayesian search called IMaGES (Ramsey et al 2010) using simulation and experimental tests. We also aim to develop novel Cognitive Neuroscience tactics and strategies to specifically test graphical models in the brain and finally we will also develop two new directions including estimation of Recurrent (feedback) network information flow and the Latent structure supporting the complexity and communication within brain networks.