Quantifying human gray matter microstructure using NEXI and 300 mT/m gradients

TL;DR

This study successfully applied the NEXI biophysical diffusion model to in vivo human cortex using advanced MRI technology, enabling detailed microstructure quantification and highlighting the importance of noise correction for accurate estimates.

Contribution

First translation of NEXI to human cortex in vivo using high-gradient MRI, demonstrating its potential for microstructure analysis in neurodevelopment and neurodegeneration.

Findings

NEXI with Rician noise correction provides reliable microstructure estimates.

The model detects individual differences in cortical microstructure.

High-gradient MRI enables broad b-value and diffusion time range acquisition.

Abstract

Biophysical models of diffusion tailored to quantify gray matter microstructure are gathering increasing interest. The two-compartment Neurite EXchange Imaging ($NEXI$) model has been proposed recently to account for neurites, extra-cellular space and exchange across the cell membrane. $NEXI$ parameter estimation requires multi-shell multi-diffusion time data and has so far only been implemented experimentally on animal data collected on a preclinical MRI set-up. In this work, the first ever translation of $NEXI$ to the human cortex in vivo was achieved using a 3T Connectom MRI system with 300 mT/m gradients, that enables the acquisition of a broad range of b-values (0 - 7.5 ms/$\mu m^{2}$) with a window of diffusion times (20 - 49 ms) suitable for the expected characteristic exchange times (10 - 50 ms). Microstructure estimates of four model variants: $NEXI$, $NEXI_{dot}$ (its extension with the addition of a dot compartment) and their respective versions that correct for the Rician noise floor ($NEXI_{RM}$ and $NEXI_{dot,RM}$) that particularly impacts high b-value signal, were compared. The reliability of estimates in each model variant was evaluated in synthetic and human in vivo data. In the latter, the intra-subject (scan-rescan) vs between-subjects variability of microstructure estimates were compared in the cortex. The better performance of $NEXI_{RM}$ highlights the importance of correcting for Rician bias in the $NEXI$ model to obtain accurate estimates of microstructure parameters in the human cortex, and the sensitivity of the $NEXI$ framework to individual differences in cortical microstructure. This groundbreaking application of $NEXI$ in humans marks a pivotal moment, unlocking new avenues for studying neurodevelopment, ageing, and various neurodegenerative disorders.