The variability of MR axon radii estimates in the human white matter.

Jelle Veraart,Nikolaus Weiskopf,Erika P Raven,Luke J Edwards,Derek K Jones

doi:10.1002/hbm.25359

Abstract

The noninvasive quantification of axonal morphology is an exciting avenue for gaining understanding of the function and structure of the central nervous system. Accurate non‐invasive mapping of micron‐sized axon radii using commonly applied neuroimaging techniques, that is, diffusion‐weighted MRI, has been bolstered by recent hardware developments, specifically MR gradient design. Here the whole brain characterization of the effective MR axon radius is presented and the inter‐ and intra‐scanner test–retest repeatability and reproducibility are evaluated to promote the further development of the effective MR axon radius as a neuroimaging biomarker. A coefficient‐of‐variability of approximately 10% in the voxelwise estimation of the effective MR radius is observed in the test–retest analysis, but it is shown that the performance can be improved fourfold using a customized along‐tract analysis.

Highlights

The white matter of the central nervous system is an intricately organized system of neural pathways that link together anatomical areas to create functional circuits
In an animal model of Angelman syndrome, a rare genetic disorder linked to autism, widespread reductions in white matter volumes were linked to reduced numbers of axons with large radii (Judson et al, 2017)
Similar observations were made in children with autism spectrum disorder (ASD), where electron microscopy identified a lower percentage of large-radii axons in the corpus callosum compared to age-matched typical developing children (Wegiel et al, 2018)

Summary

| INTRODUCTION

The white matter of the central nervous system is an intricately organized system of neural pathways that link together anatomical areas to create functional circuits. Since the spherical mean of the signal is formally a rotationally invariant feature (Mirzaalian et al, 2016), the curvature of the underlying fiber within the segment is assumed not to have a significant impact on the signal averaging and on the estimation This presented strategy is suited to biophysical models that are derived from rotationally-invariant signal features (Novikov, Fieremans, Jespersen, & Kiselev, 2019; Raven et al, 2020) and that might suffer from poor robustness to noise or partial volume effects. The metric changes widely along and across the various tracts

| RESULTS

| DISCUSSION

Findings

| CONCLUSION