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

Messaritaki, E.; Dimitriadis, Stavros I.; Jones, Derek K.

doi:10.1101/553743

Cited by 14 publications

(22 citation statements)

References 56 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We present detailed results for the seven schemes that a) resulted in reproducibility (for the graphs and their various attributes) 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. Specifically, we excluded all schemes that resulted in mean graph similarity over participants of under 0.75 (schemes 10, 14, 15, 16, 17, 18 and 20 in Table 2; also see Messaritaki et al. (2019) for more details on the graph similarity for the FA- and MD-schemes).…”

Section: Resultsmentioning

confidence: 99%

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

2019

Self Cite

View full text Add to dashboard Cite

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 s/mm 2 ). 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.

show abstract

Section: Resultsmentioning

confidence: 99%

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

2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…Currently, there is no consensus regarding which weighting factor in the construction of the graphs is the most representative measure of structural connectivity. Other definitions of edge weight, such as fractional anisotropy, mean diffusivity, level of myelination, might also be used in further work . Another important limitation of the present study is the relatively small sample size.…”

Section: Discussionmentioning

confidence: 99%

Network diffusion modeling predicts neurodegeneration in traumatic brain injury

Poudel

Verhelst

et al. 2020

Ann Clin Transl Neurol

Self Cite

View full text Add to dashboard Cite

Objective Traumatic brain injury (TBI) is a heterogeneous disease with multiple neurological deficits that evolve over time. It is also associated with an increased incidence of neurodegenerative diseases. Accordingly, clinicians need better tools to predict a patient’s long‐term prognosis. Methods Diffusion‐weighted and anatomical MRI data were collected from 17 adolescents (mean age = 15y8mo) with moderate‐to‐severe TBI and 19 healthy controls. Using a network diffusion model (NDM), we examined the effect of progressive deafferentation and gray matter thinning in young TBI patients. Moreover, using a novel automated inference method, we identified several injury epicenters in order to determine the neural degenerative patterns in each TBI patient. Results We were able to identify the subject‐specific patterns of degeneration in each patient. In particular, the hippocampus, temporal cortices, and striatum were frequently found to be the epicenters of degeneration across the TBI patients. Orthogonal transformation of the predicted degeneration, using principal component analysis, identified distinct spatial components in the temporal–hippocampal network and the cortico‐striatal network, confirming the vulnerability of these networks to injury. The NDM model, best predictive of the degeneration, was significantly correlated with time since injury, indicating that NDM can potentially capture the pathological progression in the chronic phase of TBI. Interpretation These findings suggest that network spread may help explain patterns of distant gray matter thinning, which would be consistent with Wallerian degeneration of the white matter connections (i.e., “diaschisis”) from diffuse axonal injuries and multifocal contusive injuries, and the neurodegenerative patterns of abnormal protein aggregation and transmission, which are hallmarks of brain changes in TBI. NDM approaches could provide highly subject‐specific biomarkers relevant for disease monitoring and personalized therapies in TBI.

show abstract

“…In the second approach, we constructed SBNs with edges weighted by the NS and FA and applying a threshold to remove edges with the lowest weights. Rather than imposing an arbitrary threshold, an absolute threshold was determined by imposing the constraint that the graphs exhibit the same sparsity as the OMST graphs that exhibited the highest reproducibility (Messaritaki et al, 2019b). Once the topology of each of those graphs was specified, the weights of the edges were either kept as they were or re-weighted with one of the remaining two metrics.…”

Section: Graph Construction Schemesmentioning

confidence: 99%

“…We weighted each edge using the seven most reproducible (out of 21) different graphconstruction schemes as they were explored in our previous analysis on the same dataset (Messaritaki et al, 2019b). Those seven graph construction schemes (see Section 2.3.4) were based on different combinations of the following metrics: fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), number of streamlines (NS), streamline density (SLD), percentage of streamlines (PS), tract volume (TV), tract length (TL) and Euclidean distance between the nodes (ED) (see Table 1 ).…”

Section: Edge Weightsmentioning

confidence: 99%

“…In this work, we constructed structural brain networks from a set of repeat diffusion MRI scan data from the Human Connectome project using the b=2000 s/mm 2 data and the seven most reproducible graph-construction schemes as derived from our previous study on the same dataset (Messaritaki et al, 2019b). We then estimated the reproducibility of nodal Pi and zi and also the detection of provincial and connector hubs, based on both modular network metrics.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The Impact of Graph Construction Scheme and Community Detection Algorithm on the Repeatability of Community and Hub Identification in Structural Brain Networks

Dimitriadis

Messaritaki

Jones

2020

Preprint

Self Cite

View full text Add to dashboard Cite

The human brain is a complex network of volumes of tissue (nodes) that are interconnected by white matter tracts (edges). It can be represented as a graph to allow us to use graph theory to gain insight into normal human development and brain disorders. Most graph theoretical metrics measure either whole-network (global) or node-specific (local) properties of the network. However, a critical question in network neuroscience is how nodes cluster together to form communities, each of which possibly plays a specific functional role. Community partition analysis allows us to investigate the mesoscale organization of the brain. Various algorithms have been proposed in the literature, that allow the identification of such communities, with each algorithm resulting in different communities for the network. Those communities also depend on the method used to weigh the edges of the graphs representing the brain networks. In this work, we use the test-retest data from the Human Connectome Project to compare 32 such community detection algorithms, each paired with 7 graph construction schemes, and assess the reproducibility of the resulting community partitions.The reproducibility of community partition depended heavily on both the graph construction scheme and the community detection algorithm. Hard community detection algorithms, via which each node is assigned to only one community, outperformed soft ones, via which each node can be a part of more than one community. The best reproducibility was observed for the graph construction scheme that combines 9 white matter tract metrics paired with the greedy stability optimization algorithm, with either discrete or continuous Markovian chain. This graph-construction scheme / community detection algorithm pair also gave the highest similarity between representative group community affiliation and individual community affiliation. Connector hubs were better reproduced than provincial hubs.

show abstract

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

Cited by 14 publications

References 56 publications

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

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

Network diffusion modeling predicts neurodegeneration in traumatic brain injury

The Impact of Graph Construction Scheme and Community Detection Algorithm on the Repeatability of Community and Hub Identification in Structural Brain Networks

Contact Info

Product

Resources

About