A pilot study on protocol consistency and graph metric reproducibility in microstructure-weighted connectomes

TL;DR

This study assesses the reproducibility of microstructure-weighted brain connectomes using diffusion parameters and identifies reliable metrics for tracking brain connectivity changes.

Contribution

The study introduces a systematic evaluation of reproducibility in microstructure-weighted connectomes using a four-shell acquisition protocol.

Findings

Fractional anisotropy, mean diffusivity, and intra-neurite volume fractions showed high reproducibility with low variability.

Graph metrics from FA-, MD-, and INVF-weighted connectomes were consistent, except for modularity.

ECVF-weighted connectomes had poor reproducibility, with high variability and low correlation.

Abstract

Microstructure-weighted connectomes incorporate diffusion parameters into structural networks, offering a rich characterisation of brain connectivity. While these biologically-informed connectomes have shown sensitivity to pathology-related alterations (for example in multiple sclerosis), their reproducibility remains largely unexplored. In this study, we evaluated the consistency of connectomes weighted with tensor and Bingham-NODDI parameters, employing a four-shell acquisition protocol to ensure accurate fibre reconstruction. Phantom and in vivo (N=4) data were acquired to assess temporal, inter-site and inter-protocol reproducibility of weighting parameters and inter-site stability of graph metrics. High reproducibility was observed for fractional anisotropy (FA), mean diffusivity (MD), and intra-neurite (INVF) and intra-cellular (ICVF) volume fractions, with coefficients of variation (CVs) below 5% and negligible Bland-Altman biases. Orientation dispersion index and \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\beta$$\end{document} concentration parameter showed CVs above 5% and were excluded from connectome construction. Graph metrics extracted from FA-, MD- and INVF-weighted connectomes exhibited good consistency, except for modularity. Extra-cellular volume fraction (ECVF)-weighted connectomes showed poor reproducibility (CV>5%, intraclass correlation coefficient <0.5). These preliminary findings demonstrate the reliability of microstructure-weighted connectomes, identifying the weighting strategies and graph metrics with the highest reproducibility. This supports the use of network metrics derived from weighted connectomes as potential biomarkers of altered brain connectivity in neurological disorders.

Figures