Aarhus University Seal / Aarhus Universitets segl

Mapping axonal density and average diameter using non-monotonic time-dependent gradient-echo MRI

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaperJournal articleResearchpeer-review

Abstract White Matter (WM) microstructures, such as axonal density and average diameter, are crucial to the normal function of the Central Nervous System (CNS) as they are closely related with axonal conduction velocities. Conversely, disruptions of these microstructural features may result in severe neurological deficits, suggesting that their noninvasive mapping could be an important step towards diagnosing and following pathophysiology. Whereas diffusion based MRI methods have been proposed to map these features, they typically entail the application of power gradients, which are rarely available in the clinic, or extremely long acquisition schemes to extract information from parameter-intensive models. In this study, we suggest that simple and time-efficient multi-gradient-echo (MGE) MRI can be used to extract the axon density from susceptibility-driven non-monotonic decay in the time-dependent signal. We show, both theoretically and with simulations, that a non-monotonic signal decay will occur for multi-compartmental microstructures – such as axons and extra-axonal spaces, which we here used in a simple model for the microstructure – and that, for axons parallel to the main magnetic field, the axonal density can be extracted. We then experimentally demonstrate that maps derived from MGE acquired at 16.4 T in ex-vivo spinal cords, where the different tracts characterized by different microstructures are clearly contrasted in parametric maps extracted by fitting the MGE decay to the model. When the quantitative results are compared against ground-truth histology, they seem to reflect the axonal fraction (though with a bias, as evident from Bland-Altman analysis). As well, the extra-axonal fraction can be estimated. The results suggest that our model is oversimplified, yet at the same time evidencing a potential and usefulness of the approach to map underlying microstructures using a simple and time-efficient MRI sequence. We further show that a simple general-linear-model can predict the average axonal diameters from the four model parameters, and map these average axonal diameters in the spinal cords. While clearly further modelling and theoretical developments are necessary, we conclude that salient WM microstructural features can be extracted from these simple, SNR-efficient multi-gradient echo MRI, and that this paves the way towards easier estimation of WM microstructure in vivo.
Original languageEnglish
JournalJournal of Magnetic Resonance
Pages (from-to)117
Number of pages14
Publication statusPublished - 2017

See relations at Aarhus University Citationformats

ID: 110310632