Towards higher sensitivity and stability of axon diameter estimation with diffusion-weighted MRI

Abstract

Diffusion-weighted MRI is an important tool for in vivo and non-invasive axon morphometry. The ActiveAx technique utilises an optimised acquisition protocol to infer orientationally invariant indices of axon diameter and density by fitting a model of white matter to the acquired data. In this study, we investigated the factors that influence the sensitivity to small-diameter axons, namely the gradient strength of the acquisition protocol and the model fitting routine. Diffusion-weighted ex. vivo images of the mouse brain were acquired using 16.4-T MRI with high (Gmax of 300 mT/m) and ultra-high (Gmax of 1350 mT/m) gradient strength acquisitions. The estimated axon diameter indices of the mid-sagittal corpus callosum were validated using electron microscopy. In addition, a dictionary-based fitting routine was employed and evaluated. Axon diameter indices were closer to electron microscopy measures when higher gradient strengths were employed. Despite the improvement, estimated axon diameter indices (a lower bound of~1.8μm) remained higher than the measurements obtained using electron microscopy (~1.2μm). We further observed that limitations of pulsed gradient spin echo (PGSE) acquisition sequences and axonal dispersion could also influence the sensitivity with which axon diameter indices could be estimated. Our results highlight the influence of acquisition protocol, tissue model and model fitting, in addition to gradient strength, on advanced microstructural diffusion-weighted imaging techniques. © 2016 The Authors. NMR in Biomedicine published by John Wiley & Sons Ltd. ActiveAx axon diameter indices were assessed in diffusion MRI data obtained using ultra-high gradient strengths and fitted using a dictionary-based model (ActiveAx-D). The estimated values of the axon diameter of the mouse brain corpus callosum were compared with measurements obtained from electron microscopy. ActiveAx-D axon diameter indices were closer to the electron microscopy data and were more stable (lower standard deviation) in comparison with the conventional framework.