Improvement in White Matter Tract Reconstruction with Constrained Spherical Deconvolution and Track Density Mapping in Low Angular Resolution Data: A Pediatric Study and Literature Review.

Benedetta Toselli,Andrea Canessa,Mariasavina Severino,Marco Massimo Fato,Domenico Tortora,Andrea Rossi,Giovanni Morana,Gabriele Arnulfo

doi:10.3389/fped.2017.00182

Abstract

Diffusion-weighted magnetic resonance imaging (DW-MRI) allows noninvasive investigation of brain structure in vivo. Diffusion tensor imaging (DTI) is a frequently used application of DW-MRI that assumes a single main diffusion direction per voxel, and is therefore not well suited for reconstructing crossing fiber tracts. Among the solutions developed to overcome this problem, constrained spherical deconvolution with probabilistic tractography (CSD-PT) has provided superior quality results in clinical settings on adult subjects; however, it requires particular acquisition parameters and long sequences, which may limit clinical usage in the pediatric age group. The aim of this work was to compare the results of DTI with those of track density imaging (TDI) maps and CSD-PT on data from neonates and children, acquired with low angular resolution and low b-value diffusion sequences commonly used in pediatric clinical MRI examinations. We analyzed DW-MRI studies of 50 children (eight neonates aged 3-28 days, 20 infants aged 1-8 months, and 22 children aged 2-17 years) acquired on a 1.5 T Philips scanner using 34 gradient directions and a b-value of 1,000 s/mm2. Other sequence parameters included 60 axial slices; acquisition matrix, 128 × 128; average scan time, 5:34 min; voxel size, 1.75 mm × 1.75 mm × 2 mm; one b = 0 image. For each subject, we computed principal eigenvector (EV) maps and directionally encoded color TDI maps (DEC-TDI maps) from whole-brain tractograms obtained with CSD-PT; the cerebellar-thalamic, corticopontocerebellar, and corticospinal tracts were reconstructed using both CSD-PT and DTI. Results were compared by two neuroradiologists using a 5-point qualitative score. The DEC-TDI maps obtained presented higher anatomical detail than EV maps, as assessed by visual inspection. In all subjects, white matter (WM) tracts were successfully reconstructed using both tractography methodologies. The mean qualitative scores of all tracts obtained with CSD-PT were significantly higher than those obtained with DTI (p-value < 0.05 for all comparisons). CSD-PT can be successfully applied to DW-MRI studies acquired at 1.5 T with acquisition parameters adapted for pediatric subjects, thus providing TDI maps with greater anatomical detail. This methodology yields satisfactory results for clinical purposes in the pediatric age group.

Highlights

Diffusion-weighted magnetic resonance imaging (DW-MRI) allows noninvasive investigation of brain structure in vivo
The first method to be developed for analyzing DWI sequences was Diffusion Tensor Imaging (DTI) [3]: this method estimates a diffusion tensor for each voxel, from which the main diffusion direction is computed as the major eigenvector of the tensor
The EV maps allowed to visualize the main fiber tracts in older children, while they presented a more blurred appearance for unmyelinated neonates; the distinguishable level of detail was lower in the EV maps than in stDEC-track density imaging (TDI) for all subjects

Summary

Introduction

Diffusion-weighted magnetic resonance imaging (DW-MRI) allows noninvasive investigation of brain structure in vivo. Diffusion-weighted magnetic resonance imaging (DW-MRI) has become the method of choice to noninvasively analyze brain structure and pathology in the clinical setting, providing in vivo quantitative and qualitative information about white matter (WM) microstructure and fiber tract pathways (see for example Schaefer et al [1] for a review of the possible applications, or Moritani et al [2] for a more complete discussion). This analysis is performed by mapping the motion of water molecules in the brain; while in the cerebrospinal fluid and gray matter, water motion does not significantly prevail in any direction (i.e., isotropic diffusion); in WM, water molecules move preferentially along the direction of axons and fiber tracts (i.e., anisotropic diffusion). The pixel intensity of TDI maps reflects the number of probabilistic streamlines traversing the voxel and the color reflects diffusion streamline orientations similar to those of conventional DTI, allowing better WM anatomical localization and characterization [9]

Objectives

Methods

Results

Conclusion