Improving the Reliability of Network Metrics in Structural Brain Networks by Integrating Different Network Weighting Strategies into a Single Graph

Abstract

Structural brain networks estimated from diffusion MRI (dMRI) via tractography have been widely studied in healthy controls and patients with neurological and psychiatric diseases. However, few studies have addressed the reliability of derived network metrics both node-specific and network-wide. Different network weighting strategies (NWS) can be adopted to weight the strength of connection between two nodes yielding structural brain networks that are almost fully-weighted. Here, we scanned five healthy participants five times each, using a diffusion-weighted MRI protocol and computed edges between 90 regions of interest (ROI) from the Automated Anatomical Labeling (AAL) template. The edges were weighted according to nine different methods. We propose a linear combination of these nine NWS into a single graph using an appropriate diffusion distance metric. We refer to the resulting weighted graph as an Integrated Weighted Structural Brain Network (ISWBN). Additionally, we consider a topological filtering scheme that maximizes the information flow in the brain network under the constraint of the overall cost of the surviving connections. We compared each of the nine NWS and the ISWBN based on the improvement of: (a) intra-class correlation coefficient (ICC) of well-known network metrics, both node-wise and per network level; and (b) the recognition accuracy of each subject compared to the remainder of the cohort, as an attempt to access the uniqueness of the structural brain network for each subject, after first applying our proposed topological filtering scheme. Based on a threshold where the network level ICC should be >0.90, our findings revealed that six out of nine NWS lead to unreliable results at the network level, while all nine NWS were unreliable at the node level. In comparison, our proposed ISWBN performed as well as the best performing individual NWS at the network level, and the ICC was higher compared to all individual NWS at the node level. Importantly, both network and node-wise ICCs of network metrics derived from the topologically filtered ISBWN (ISWBNTF), were further improved compared to the non-filtered ISWBN. Finally, in the recognition accuracy tests, we assigned each single ISWBNTF to the correct subject. We also applied our methodology to a second dataset of diffusion-weighted MRI in healthy controls and individuals with psychotic experience. Following a binary classification scheme, the classification performance based on ISWBNTF outperformed the nine different weighting strategies and the ISWBN. Overall, these findings suggest that the proposed methodology results in improved characterization of genuine between-subject differences in connectivity leading to the possibility of network-based structural phenotyping.

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

“…The OMST algorithm is used to select the edges (NS+FS OMST). This is the same scheme that is described by Dimitriadis et al (2017b), using only two metrics to weigh the edges.…”

“…We used the algorithm described by Dimitriadis et al (2017b,c) to generate an integrated network for the data from each diffusion-weighting and each scan of each participant. Following Dimitriadis et al (2017b), for each participant at each time point, a two-step process was followed: 1) We used the diffusion-distance (Hammond et al, 2013) between individual CMs to maximise information provided by each metric and create a linear-combination (or integrated) graph, and 2) we used an Orthogonal Minimal Spanning Tree (OMST) algorithm (Dimitriadis et al, 2017a,c) to selectively remove edges of the resulting graph, so as to maximize the difference (Global Efficiency -Cost), while maintaining the connectivity of the nodes. The benefit of the method lies in the fact that both the topology that results from selectively removing edges and the assignment of the edge weights are performed in a data-driven manner, and no arbitrary threshold needs to be imposed.…”

“…In addition to using the algorithm of Dimitriadis et al (2017b), we also constructed networks using the NS, FA or MD as edge weights, due to their prevalent use in the literature (for example Honey et al (2009) andCollin et al (2016)), employing the OMST algorithm to select the edges in a data-driven manner.…”

“…Additionally, high reproducibility and low within-participant variability results in higher power for the studies and therefore a lower number of required participants, which in turn results in reduction of the cost and resources required for the study. Several papers have investigated different aspects of the the reproducibility of structural brain networks and their graph theoretical metrics (Dennis et al, 2012;Owen et al, 2013;Buchanan et al, 2014;Smith et al, 2015;Owen et al, 2013;Zhong et al, 2015;Dimitriadis et al, 2017b;Yuan et al, 2018), each using a specific scheme for constructing the graphs, predominantly using data acquired with low diffusion weightings (b-values of up to 1500 s/mm 2 ). The work of Roine et al (2018) also investigated a higher diffusion weighting of b = 3000 s/mm 2 .…”

The effect of network thresholding and weighting on structural brain networks in the UK Biobank

“…NODDI estimates neurite density (intra-cellular volume fraction; ICVF), extra-cellular water diffusion (isotropic volume fraction; ISOVF) and tract complexity/fanning (orientation dispersion; OD), biomarkers which can also be used as network weightings. Previous research has compared some conventional weightings (Buchanan et al, 2014;Dimitriadis et al, 2017;Qi et al, 2015), but it is not yet clear how thresholding affects differentlyweighted networks and their relationships with external variables, such as age.…”