Assessment of multi-modal, genetically influenced, dynamic brain connectivity in disease states

点击翻译按钮获取中文摘要

Objects of the Proposed Research Program: This research will continue the applicant's work on statistical signal processing applications applied to clinically-relevant data. The long term goal is to develop novel, fundamental statistical signal processing approaches that will have a lasting impact on the fields of clinical brain imaging. Toward this objective, this proposal focuses on tackling two fundamental topics: data fusion: between imaging modalities as well as between imaging and clinical and genetic data; and dynamical aspects of brain connectivity. Modern imaging technologies has re-ignited a centuries-old debate about how brain activity is represented: after long assuming that brain activity associated with specific tasks is localized, there is now a greater appreciation on how diverse brain areas co-activate. Ongoing, dynamic association between spatially disparate brain regions appears critical for normal brain functioning and disruption of these connectivity patterns is a sensitive marker for disease. Accurate assessment of brain connectivity patterns is thus an overarching objective of this proposal. Scientific Approach Most analyses examining brain connectivity have been specific to a technology (e.g. fMRI, EEG), each of which occupies a limited area in the spatiotemporal plane. Combining different brain modalities to provide a comprehensive assessment of brain connectivity at multiple temporal and spatial scales is non-trivial, as each modality measures different biological activity (e.g. electrical activity vs. changes in blood flow), and has different statistical characteristics. We will expand our prior work on fMRI, EEG-EEG and EEG-EMG connectivity, as well as characterization of simultaneously-recorded behavioral data with linear dynamical system models, so that this complementary information can be merged together to provide sensitive and specific markers for disease processes. The role of genetic influences in neurodegenerative diseases such as Parkinson's disease is being increasingly recognized. People with genetic mutations putting them at high risk for developing disease provide a rare opportunity to examine connectivity changes before symptoms emerge. In complementary fashion, classifying brain connectivity patterns within an extended family may suggest which members share a common disease-related phenotype, substantially narrowing the search for gene-related alterations associated with as specific disease. However, both genetic and imaging data suffer from common challenges: the data are inherently high-dimensional with relatively few available samples. In order to investigate how genetic and imaging data may be meaningfully combined, we will extend our work on sparse regression and sparse precision matrices. Most current models of brain connectivity assume stationarity of connectivity patterns during performance of a task. However the brain in inherently non-stationary, and modeling dynamic changes in connectivity patterns is still in its infancy. Starting with our prior work on assessing connectivity in fMRI data sets, we will employ a dynamical framework to explore deterministic alterations in connectivity patterns during task performance. Novelty and Expected Significance of Work: Besides providing a highly competitive environment for interdisciplinary training of HQP, this work will provide a framework for multimodal assessment of brain activity, with widespread potential impact in the assessment of normal brain functioning and in disease states.