Connectivity of the Brain from Magnetic Resonance Imaging

Citation

Neumann, Dirk (2010) Connectivity of the Brain from Magnetic Resonance Imaging. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/1A5T-S275. https://resolver.caltech.edu/CaltechTHESIS:04282010-153942989

Abstract

How do different parts of the brain work? The naive and somewhat ill-posed question nonetheless admits of a serious answer. Different regions of the brain carry out their function principally through two components: the pattern of inputs and outputs that connect a region with the rest of the brain, and the computational transformations implemented by neurons within the region itself. Here we focus on the former problem and study the connectivity of the primate brain, with an emphasis on neocortex.

We develop a novel set of algorithms for modeling anatomical connectivity based on diffusion-weighted magnetic resonance (MR) imaging. The approach is novel in several respects: it utilizes a new way of deriving a globally optimal solution from local message passing; it can be applied to the whole-brain level in a computationally tractable fashion; and it can flexibly incorporate much other information, such as constraints about the geometry of white-matter tracts and high-resolution anatomical MR images. The algorithm is first described as a hierarchical Bayesian model, and then applied to the diffusion MRI data obtained from two perfusion-fixed brains of macaque monkeys.

Based on the connectivity output provided from applying our novel algorithm to high-angular resolution MR data, we next derive several new insights about the connectivity of the macaque brain. We compare our results against those from published tracer studies, and we derive the relative weights of connections known from such prior studies. We also demonstrate the ability of the algorithm to generate entirely novel connectivity data, both at the level of specific anatomical regions that are queries, and also at the whole-brain level. The latter permits new insights into whole-brain connectivity and its architecture.

In addition to this focus on the structural connectivity of the macaque brain, we also analyze an extant set of public data of BOLD-fMRI from the macaque brain. This data set yields information regarding the functional connectivity of the macaque brain that we put together with our new connectivity results in order to relate structural and functional connectivity, with several new discoveries about their relationship.

In the final chapter, we apply these methods to MR data we collected from the live human brain. We provide an overview of structural and functional connectivity results obtained from this data set, and we apply the investigation to the brains of rare patients with agenesis of the corpus callosum, who lack the normal connection between the left and right hemispheres. We close by illustrating the power of the approach to ask questions that integrate functional questions with connectivity information on which function must ultimately be based: using connectivity profiles in order to segment cortical regions based on their pattern of inputs and outputs, with the aim of then querying these segmented regions using fMRI in cognitive activation studies. The description of our algorithm, the demonstration of its reliability, validity, and application to yield new data, together with the extensive software libraries on which the work is based, will provide cognitive neuroscientists with an array of new tools to investigate brain function in both health and disease.

Thesis Files