Functional imaging has been one of the most important measurement tools in the study of brain function. While earlier fMRI-based findings were limited to localization of brain regions strongly modulated by the experimental protocol, the images also contain a wealth of information about the spatial nature of the networks involved in perception and cognition. This project aims to construct novel representations of spatial patterns of activation, to develop methods for extracting such representations from fMRI images and to build visualization tools for interpretation of the detected patterns. Computational tools for modeling spatial patterns of activation will significantly increase the utility of fMRI for investigating the brain function and its organization.
The present approach is based on the empirical observation that the spatial patterns of activation are inherently hierarchical. Similar to anatomical hierarchies that represent the structure of the brain as a tree of increasingly simple systems, the functional description of the brain should also be of a hierarchical nature. In this work, functional region hierarchies are constructed to represent functional organization of the brain. In addition to enabling robust and accurate models of functional organization, this new representation reduces the gap between the theories of functional organization of the brain and the functional models extracted from fMRI data. Moreover, a hierarchical structure can be naturally interpreted by a human observer, providing an intuitive and meaningful representation that can be easily integrated with neuroanatomy and other available sources of information.
The methodology developed in this project will improve our understanding of spatial connectivity patterns implicitly captured by fMRI. The analysis and visualization tools will enable new ways to investigate the function of the brain. The experimental findings in the collaborative studies will directly produce new hypotheses about functional organization of the relevant systems. The tightly integrated educational and outreach components of the project will directly affect a large number of MIT students and other groups outside MIT by exposing them to open problems and research methods in biomedical imaging and analysis.
The goal of this project was to develop novel methods for analysis of fMRI data to provide insight into the functional organization of the brain. Our approach is to identify and characterize spatial patterns of activation as observed through fMRI imaging of the brain. As part of this project we developed and demonstrated computational analysis methods and deployed them in application to studies of normal brain function and of effects of pathology on the functional organization of the brain. Early in the project, we demonstrated that the brain can be robustly subdivided into functionally coherent stable functional areas. Our later research produced methods for automatic detection of brain networks characterized by a stable pattern of co-activation. These methods further enabled a study of the spatial variability of the brain networks in the normal brain and of functional reorganization due to severe pathologies such as tumors. In cases where spatial variability of the functional regions can be accounted for by spatial normalization methods based on anatomical alignment, we developed a novel approach to characterize a graph of co-activations in the brain that represents a normal population and a graph of abnormal patterns of co-activation due to disease. We demonstrated this approach in a study of schizophrenia where we identified a small set of regions that jointly explain observed differences in fMRI data between the clinical group and the normal controls. And finally, we demonstrated how to perform spatial normalization of subjects based on their functional activation patterns. This work enables better fMRI studies of normal brain function in subjects engaged in complex tasks such as language processing or generation. Research activities in this project trained students and postdocs in research methods in computational analysis, neuroscience, and clinical research. Educational activities associated with the project trained a large number of students in statistical modeling and inference methods. Outreach efforts educated the general public about functional organization of the brain and fMRI scanning for mapping out the human brain.
