Estimation of the spatial and temporal patterns of intrinsic neural activity has become a popular approach to gaining insight into the functional organization of the brain. One method for measuring these patterns of spontaneous activity, referred to as functional connectivity, is functional Magnetic Resonance Imaging (MRI) measured at rest while other studies developed experimental tasks for learning about connectivity in specific areas of the brain. Functional connectivity maps have been shown to change with age, training, levels of consciousness, and disease status. Under certain assumptions, these functional connectivity maps demonstrate deviations from independence between various areas of the brain, often including spatially incongruous areas. Functional connectivity maps have been utilized to learn about differences of brain activation patterns between disease groups via clustering voxels based on their connectivity patters. We propose a general framework for estimating associations of brain connectivity maps with predictors of interest after controlling for confounders using covariance regression - a statistical modeling approach that allows for using the special structure of covariance outcomes for improved parameter estimation. The statistical significance of the association of the predictors with the outcome maps will be assessed in this model while correcting for the effects of other variables. The estimation of connectivity maps in a covariance regression framework, where the map is the outcome is underdeveloped. Our proposed framework will extend the model and incorporate evidence based realistic assumptions in the context of static functional connectivity analysis where we assume that the connectivity is constant during the scanning session and in dynamic connectivity analyses where time-varying patterns of changes in connectivity during the scanning session is of interest. An important contribution of this proposal is the extension of the model to high dimensional settings as the connectivity maps based on fMRI are often large. The proposed framework will be used to learn about functional organization of the brain during an adaptation learning task in a functional MRI study focusing on visual-motor connectivity changes during the task.
Showing the most recent 10 out of 60 publications