Reproducibility and intercorrelation of graph theoretical measures in structural brain connectivity networks

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Network Science

Background:

• Diffusion-weighted magnetic resonance imaging (dMRI) non-invasively probes brain microstructure.
• Advances enable identification of complex white matter fiber configurations, improving structural connectivity analysis via tractography.
• Whole-brain structural connectivity networks (connectomes) are reconstructed by gray matter parcellation and tractography.

Purpose of the Study:

• To investigate the reproducibility and intercorrelation of network properties, connectivity weights, and tractography parameters in structural brain networks.
• To provide guidelines for reproducible investigations of structural brain networks.

Main Methods:

• Reconstruction of structural brain connectivity networks using constrained spherical deconvolution (CSD) based probabilistic streamlines tractography.
• Acquisition of dMRI data from 19 subjects (b=2800 s/mm², 75 gradient orientations).
• Computation of intrasubject variability using residual bootstrapping and replication using a Human Connectome Project test-retest dataset.

Main Results:

• Graph theoretical metrics showed excellent reproducibility, except for betweenness centrality.
• Approximately one million streamlines are necessary for excellent reproducibility, with higher densities further improving it.
• Reproducibility slightly decreases with higher CSD order but improves in binary networks with high threshold values; network properties and connectivity weights are highly intercorrelated.

Conclusions:

• Established parameters for reproducible structural brain network analysis.
• Findings offer practical guidance for researchers using tractography and graph theory to study brain connectivity.
• Highlights the importance of streamline density, thresholding, and CSD order for reliable connectome reconstruction.