White Matter Tissue Quantification at Low b-Values Within Constrained Spherical Deconvolution Framework.

Alessandro Calamuneri,Michele Gaeta,Alberto Cacciola,Angelo Quartarone,Silvia Marino,Demetrio Milardi,Gaetana Chillemi,Alessandro Arrigo,Enricomaria Mormina

doi:10.3389/fneur.2018.00716

Abstract

In the last decades, a number of Diffusion Weighted Imaging (DWI) based techniques have been developed to study non-invasively human brain tissues, especially white matter (WM). In this context, Constrained Spherical Deconvolution (CSD) is recognized as being able to accurately characterize water molecules displacement, as they emerge from the observation of MR diffusion weighted (MR-DW) images. CSD is suggested to be applied on MR-DW datasets consisting of b-values around 3,000 s/mm2 and at least 45 unique diffusion weighting directions. Below such technical requirements, Diffusion Tensor Imaging (DT) remains the most widely accepted model. Unlike CSD, DTI is unable to resolve complex fiber geometries within the brain, thus affecting related tissues quantification. In addition, thanks to CSD, an index called Apparent Fiber Density (AFD) can be measured to estimate intra-axonal volume fraction within WM. In standard clinical settings, diffusion based acquisitions are well below such technical requirements. Therefore, in this study we wanted to extensively compare CSD and DTI model outcomes on really low demanding MR-DW datasets, i.e., consisting of a single shell (b-value = 1,000 s/mm2) and only 30 unique diffusion encoding directions. To this end, we performed deterministic and probabilistic tractographic reconstruction of two major WM pathways, namely the Corticospinal Tract and the Arcuate Fasciculus. We estimated and analyzed tensor based features as well as, for the first time, AFD interpretability in our data. By performing multivariate statistics and tract-based ROI analysis, we demonstrate that WM quantification is affected by both the diffusion model and threshold applied to noisy tractographic maps. Consistently with existing literature, we showed that CSD outperforms DTI even in our scenario. Most importantly, for the first time we address the problem of accuracy and interpretation of AFD in a low-demanding DW setup, and show that it is still a biological meaningful measure for the analysis of intra-axonal volume even in clinical settings.

Highlights

In the last decades diffusion MRI allowed to study non-invasively white matter (WM) by analyzing water molecules diffusion process in vivo [1]
When investigating Arcuate Fasciculus (AF) results, differences between probabilistic Constrained Spherical Deconvolution (CSD) based reconstruction and other methods tended to decrease as density increased; all Overlap Fraction (OF) resulted below 60% (Figure 1E)
Estimated fraction was below both the theoretical intra-cellular volume fraction [26] (80%) as well as the intra-axonal volume fraction reported in Raffelt et al [25], where an average value of 75% was found based on b = 3,000 s/mm2

Summary

Introduction

In the last decades diffusion MRI allowed to study non-invasively white matter (WM) by analyzing water molecules diffusion process in vivo [1]. Diffusion Tensor imaging (DTI) was the first model historically developed to describe tridimensional water anisotropic motion in the brain [2]. Taking into account tensor model, tissues quantification has been characterized over the years by means of several scalar indices, such as fractional anisotropy (FA) and mean diffusivity (MD) [4]. These values have been shown to correlate with WM integrity in normal vs pathological conditions [5, 6]. A number of techniques have been developed to overcome such limitations, like multi-tensor models [15,16,17], Q-ball imaging [18], Constrained Spherical Deconvolution (CSD) [19,20,21] and Diffusion Spectrum Imaging [22]

Objectives

Methods

Results

Discussion

Conclusion