Network Connectivity Modeling of Heterogeneous Brain Data to Examine Ensembles of Activity Across Two Levels of Dimensionality

点击翻译按钮获取中文摘要

Project Summary/Abstract Network methods have emerged as some of the most useful approaches for analyzing functional MRI data. While great advancements have been made in these methods, limitations hamper the progress fMRI researchers can make in better understanding brain processes. In particular, researchers are typically limited to looking at properties within a network, such as how regions relate across time, and cannot simultaneously look at relations between known networks. However, increasingly hypotheses require understanding the brain at two scales: at the resolution of the regions of interest within a known network and at the network level. Further hampering progress is that few network methods available can reliably arrive at network models for individuals. Indeed, increasingly researchers are finding that brain processes vary greatly across individuals, and thus methods are needed that do not assume homogeneity. Brain processes in heterogeneous data can be better studied by using reliable and valid approaches that attend to individual nuances while assessing relations within and between networks. We propose to develop, test, and freely disseminate an algorithm for the network-based analysis of brain processes that attends to these problems. The theory driving our approach is that interactions within and between large neural systems and brain areas – including multiple sensory systems, cognitive functioning, and attentional control - drive behavior and subjective experiences by working in concert with each other. Towards these ends, our software will provide statistical inference frameworks for conducting network connectivity and causal-inference analyses. Importantly, the proposed algorithm uniquely would enable researchers to address data dimensionality by correlating ensembles existing at lower dimension brain activity (i.e., data reduced to network activity) as well at higher dimensionality (i.e., the full functional brain parcellated into regions) within a unified modeling framework. Following the first stage of development and testing, we will validate the algorithms on data from within a highly controlled experimental design.