Dimensionless, Scale Invariant, Edge Weight Metric for the Study of Complex Structural Networks.

Luis M Colon-Perez,Catherine Price,Shelby Goicochea,Mansi Parekh,William Triplett,Thomas H Mareci,Paul R Carney,Caitlin Spindler,Eric Montie,Heye Zhang

doi:10.1371/journal.pone.0131493

Luis M Colon-Perez, Catherine Price + Show 8 more

Open Access

https://doi.org/10.1371/journal.pone.0131493

Copy DOI

Abstract

High spatial and angular resolution diffusion weighted imaging (DWI) with network analysis provides a unique framework for the study of brain structure in vivo. DWI-derived brain connectivity patterns are best characterized with graph theory using an edge weight to quantify the strength of white matter connections between gray matter nodes. Here a dimensionless, scale-invariant edge weight is introduced to measure node connectivity. This edge weight metric provides reasonable and consistent values over any size scale (e.g. rodents to humans) used to quantify the strength of connection. Firstly, simulations were used to assess the effects of tractography seed point density and random errors in the estimated fiber orientations; with sufficient signal-to-noise ratio (SNR), edge weight estimates improve as the seed density increases. Secondly to evaluate the application of the edge weight in the human brain, ten repeated measures of DWI in the same healthy human subject were analyzed. Mean edge weight values within the cingulum and corpus callosum were consistent and showed low variability. Thirdly, using excised rat brains to study the effects of spatial resolution, the weight of edges connecting major structures in the temporal lobe were used to characterize connectivity in this local network. The results indicate that with adequate resolution and SNR, connections between network nodes are characterized well by this edge weight metric. Therefore this new dimensionless, scale-invariant edge weight metric provides a robust measure of network connectivity that can be applied in any size regime.

Highlights

A more complete understanding of brain function requires detailed information about which brain regions are structurally connected and how the strength of these connections relates to function [1, 2]
Using diffusion-weighted magnetic resonance imaging (DWI) tractography techniques [12] and network analysis [13], graph theory models of brain white matter (WM) connectivity networks [14] in vivo can be created in which gray matter (GM) anatomical regions correspond to network nodes and WM fibers correspond to network edges connecting nodes
For the successful creation of a network representation of the brain, an edge weight metric is needed that provides a stable, dimensionless value, that is independent of length scale, image resolution, and the tractography seeding used to estimate WM fibers

Summary

Introduction

A more complete understanding of brain function requires detailed information about which brain regions are structurally connected and how the strength of these connections relates to function [1, 2]. Using DWI tractography techniques [12] and network analysis [13], graph theory models of brain WM connectivity networks [14] in vivo can be created in which GM anatomical regions correspond to network nodes and WM fibers correspond to network edges connecting nodes. An ideal edge weight metric should quantify the strength of connection in a logical manner that can be applied to a brain of any size (e.g. rodents to humans) and be independent of image resolution. A scheme for calculating edge weights is needed that provides a reliable measure of connectivity strength, which converges to a stable value as the seed density increase, and is independent of length scale, image resolution, and the tractography scheme used to estimate WM fibers

Methods

Results

Discussion

Conclusion