Optimization of graph construction can significantly increase the power of structural brain network studies

Eirini Messaritaki,Stavros I Dimitriadis,Derek K Jones

doi:10.1016/j.neuroimage.2019.05.052

Eirini Messaritaki, Stavros I Dimitriadis + Show 1 more

Open Access

https://doi.org/10.1016/j.neuroimage.2019.05.052

Copy DOI

Journal: NeuroImage	Publication Date: Jun 6, 2019
Citations: 37	License type: cc-by

Affiliation: Cardiff University, Australian Catholic University

Abstract

Structural brain networks derived from diffusion magnetic resonance imaging data have been used extensively to describe the human brain, and graph theory has allowed quantification of their network properties. Schemes used to construct the graphs that represent the structural brain networks differ in the metrics they use as edge weights and the algorithms they use to define the network topologies. In this work, twenty graph construction schemes were considered. The schemes use the number of streamlines, the fractional anisotropy, the mean diffusivity or other attributes of the tracts to define the edge weights, and either an absolute threshold or a data-driven algorithm to define the graph topology. The test-retest data of the Human Connectome Project were used to compare the reproducibility of the graphs and their various attributes (edges, topologies, graph theoretical metrics) derived through those schemes, for diffusion images acquired with three different diffusion weightings. The impact of the scheme on the statistical power of the study and on the number of participants required to detect a difference between populations or an effect of an intervention was also calculated.The reproducibility of the graphs and their attributes depended heavily on the graph construction scheme. Graph reproducibility was higher for schemes that used thresholding to define the graph topology, while data-driven schemes performed better at topology reproducibility (mean similarities of 0.962 and 0.984 respectively, for graphs derived from diffusion images with b=2000 s/mm2). Additionally, schemes that used thresholding resulted in better reproducibility for local graph theoretical metrics (intra-class correlation coefficients (ICC) of the order of 0.8), compared to data-driven schemes. Thresholded and data-driven schemes resulted in high (0.86 or higher) ICCs only for schemes that use exclusively the number of streamlines to construct the graphs. Crucially, the number of participants required to detect a difference between populations or an effect of an intervention could change by a factor of two or more depending on the scheme used, affecting the power of studies to reveal the effects of interest.

Highlights

We present detailed results for the seven schemes that a) resulted in reproducibility that is high enough for the graphs to be useful for comparative and longitudinal studies and b) are different enough from each other to convey a different picture of the structural connectome
Schemes 11, 12 and 13 resulted in graphs that are very similar to the 9-m orthogonal minimal spanning tree (OMST) scheme, and we decided to discuss the latter in detail due to the fact that it exhibits higher reproducibility for the graph theoretical metrics
We presented and compared several alternative graph construction schemes for structural brain networks

Summary

Introduction

The view that the human brain is a network of cortical and subcortical areas connected via white matter tracts emerged several decades ago and has been very useful in characterizing brain structure and function in both healthy participants and diseased populations (Hagmann et al, 2008; Griffa et al (2013); Caeyenberghs and Leemans (2014); Fischer et al (2014); van den Heuvel and Fornito (2014); Yuan et al (2014); Baker et al (2015); Collin et al (2016); Drakesmith et al (2015); Aerts et al (2016); Nelson et al (2017); Vidaurre et al (2018); Imms et al.(2019), to name just a few of the studies). If that is the case, such graphs are bound to be reproducible, within experimental error, when generated from data collected at different times, in the absence of any true changes in the structural connectome This means that graphs that are not reproducible in the absence of age-related changes or intervention-induced plasticity are not reliable in representing the structural organization of the human brain. This is important because various graph construction schemes have been presented in the literature that use different techniques to perform tractography, define graph topology and assign edge-weights, all producing graphs that are quite different from each other, with different levels of robustness and reproducibility. The work of Roine et al (2018) investigated a higher diffusion weighting of b 1⁄4 3000 s/mm

Methods

Results

Discussion

Conclusion