^{1}

^{2}

^{3}

^{4}

^{1}

^{5}

^{3}

^{6}

^{7}

^{1}

^{2}

^{1}

^{5}

^{1}

^{2}

^{*}

^{1}

^{2}

^{3}

^{4}

^{5}

^{6}

^{7}

Edited by: Yogesh Rathi, Harvard Medical School, United States

Reviewed by: Kang Ik Kevin Cho, Seoul National University, South Korea; Hans J. Johnson, The University of Iowa, United States

This article was submitted to Brain Imaging Methods, a section of the journal Frontiers in Neuroscience

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

High angular resolution diffusion imaging (HARDI)-based tractography has been increasingly used in longitudinal studies on white matter macro- and micro-structural changes in the language network during language acquisition and in language impairments. However, test-retest reliability measurements are essential to ascertain that the longitudinal variations observed are not related to data processing. The aims of this study were to determine the reproducibility of the reconstruction of major white matter fiber bundles of the language network using anatomically constrained probabilistic tractography with constrained spherical deconvolution based on HARDI data, as well as to assess the test-retest reliability of diffusion measures extracted along them. Eighteen right-handed participants were scanned twice, one week apart. The arcuate, inferior longitudinal, inferior fronto-occipital, and uncinate fasciculi were reconstructed in the left and right hemispheres and the following diffusion measures were extracted along each tract: fractional anisotropy, mean, axial, and radial diffusivity, number of fiber orientations, mean length of streamlines, and volume. All fiber bundles showed good morphological overlap between the two scanning timepoints and the test-retest reliability of all diffusion measures in most fiber bundles was good to excellent. We thus propose a fairly simple, but robust, HARDI-based tractography pipeline reliable for the longitudinal study of white matter language fiber bundles, which increases its potential applicability to research on the neurobiological mechanisms supporting language.

The characterization of the brain and language network and its development, disruption, and changes over time represents one of the central themes of cognitive neuroscience. Diffusion magnetic resonance imaging (dMRI)-based tractography has been proven to be a valuable tool for the

It is increasingly accepted that WM associative fiber bundles play a crucial role in mediating the transfer of information among specialized language brain areas, distributed along two main processing streams, namely the dorsal and ventral streams (

The use of advanced probabilistic fiber tracking based on high angular resolution diffusion imaging (HARDI) has proven to be particularly suitable for the reconstruction of fiber bundles with complex configurations (i.e., crossing, kissing, or fanning fibers), such as language-related fiber bundles (

The combination of micro- and macro-structural measures allows a more comprehensive analysis of WM fiber bundle characteristics. Microstructural properties of bundles reconstructed with tractography are usually inferred from the extraction of different scalar metrics, such as fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD). These measures are sensitive to different fiber properties such as axonal ordering, myelinization, and density (

While there is growing interest in the use of tractography and tractometry in longitudinal studies to investigate language-related fiber bundles’ changes over time (e.g.,

In order to fill this gap, the aim of the present study is to assess test-retest reliability of the reconstruction, as well as the micro- and macro-structural characteristics of the major WM fiber bundles associated with language processing reconstructed using probabilistic HARDI-tractography. To this aim, we have collected dMRI data from a sample of healthy individuals at two time-points, one week apart, and reconstructed major WM fiber bundles supporting language functions within the left and right hemispheres (AF, ILF, IFOF, and UF). We expect that no measurable changes in the micro- and macro-structural characteristics of the tracts under study would be observed in that short time period. The test-retest reliability of the fiber bundles’ morphology was obtained by calculating, for each subject, the spatial overlap between each tract’s reconstruction at the two time-points as proposed in

Eighteen right-handed cognitively unimpaired participants (age:

The diffusion MRI protocol was acquired using a Skyra 3T MRI scanner (Siemens Healthcare, United States) at the radiology department of Hôpital du Sacré-Coeur of Montreal. At each of the two scanning occasions participants underwent the same acquisition sequence. One high resolution 3D T1-weighted (T1w) image (TR = 2200 ms, TE = 2.96 ms, TI = 900 ms, FOV = 250 mm, voxel size = 1 mm × 1 mm × 1 mm, matrix = 256 × 256, 192 slices, flip-angle = 8) was acquired using a Magnetization Prepared Rapid Gradient Echo (MP-RAGE) sequence. A diffusion weighted imaging (DWI) sequence was also acquired (TR = 8051 ms, TE = 86 ms, FOV = 230 mm, voxel size = 2 mm × 2 mm × 2 mm, flip angle = 90°, bandwidth = 1698; EPI factor = 67; 68 slices in transverse orientation) with one image (^{2}) and 64 images with non-collinear diffusion gradients (^{2}) in a posterior-anterior (PA) acquisition, as well as two additional images (^{2}): one in a PA acquisition, namely in the same direction as the diffusion gradients, and the other in an anterior-posterior (AP) acquisition, namely in the opposite direction of the diffusion gradients.

All analysis steps were conducted using the Toolkit for Analysis in Diffusion MRI (TOAD) pipeline^{1}.

Pre-processing steps included denoising, motion/eddy/distortion corrections, upsampling, registration, segmentation and parcellation, and masking. First, DWI was noise-corrected using overcomplete local principal component analysis (PCA) using the Matlab toolbox DWI Denoising Software (^{2}) (

Fiber orientation distribution functions (fODFs) were estimated using CSD. A whole-brain tractogram was computed using MRtrix3’s probabilistic tractography algorithm with ACT^{3} (

Test-retest analyses were carried out in two steps. First, we used the weighted dice similarity coefficient (wDSC) to determine the degree of overlap between the reconstructed fiber bundles at Times 1 and 2 as in

where

To do so, T1-weighted images in diffusion space taken at Time 1 were registered linearly to anatomical images taken at Time 2 (i.e., seven days later) for each subject with Advanced Normalization Tools (ANTs), version ≥ 2.1 (^{4}. Transformation matrices were applied to all Time 1 bundles using TractQuerier’s tract_math tool (

In a second step, we combined two complementary analyses, the intra-class correlation coefficient (ICC) and the Bland-Altman plots to assess the test-retest reliability of each of the measures extracted in each reconstructed fiber bundle. The intra-class correlation coefficient (ICC) (

where MS_{R} is the mean square for rows, MS_{C} is the mean square of columns, MS_{E} is the mean square for error,

We also created Bland and Altman plots which provide a visual assessment of the agreement of the two time-points (test and retest) of each measure in all four fiber bundles bilaterally (

The degree of overlap was good for all four reconstructed fiber bundles (AF, ILF, IFOF, and UF, bilaterally) between Time 1 and Time 2, with wDSC values ranging between 0.71 and 0.87 (values for each fiber bundle are reported in Table

wDSC values, ICC estimates and their 95% confidence intervals for all measures and fiber bundles.

wDSC | FA |
AD |
MD |
RD |
NuFO |
Volume |
MLS |
||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|

ICC | CI | ICC | CI | ICC | CI | ICC | CI | ICC | CI | ICC | CI | ICC | CI | ||

Left | 0.86 | 0.89^{∗∗∗} |
0.61–0.96 | 0.91^{∗∗∗} |
0.78–0.97 | 0.91^{∗∗∗} |
0.79–0.97 | 0.92^{∗∗∗} |
0.81–0.97 | 0.62^{∗∗} |
0.22–0.84 | 0.83^{∗∗∗} |
0.61–0.93 | 0.71^{∗∗∗} |
0.37–0.88 |

Right | 0.83 | 0.85^{∗∗∗} |
0.65–0.94 | 0.71^{∗∗∗} |
0.38–0.88 | 0.86^{∗∗∗} |
0.67–0.95 | 0.89^{∗∗∗} |
0.73–0.96 | 0.68^{∗∗} |
0.33–0.87 | 0.79^{∗∗∗} |
0.52–0.92 | 0.87^{∗∗∗} |
0.68–0.95 |

Left | 0.79 | 0.78^{∗∗∗} |
0.52–0.91 | 0.95^{∗∗∗} |
0.86–0.98 | 0.95^{∗∗∗} |
0.86–0.98 | 0.88^{∗∗∗} |
0.70–0.95 | 0.50^{∗} |
0.05–0.78 | 0.79^{∗∗∗} |
0.53–0.91 | 0.84^{∗∗∗} |
0.62–0.94 |

Right | 0.71 | 0.86^{∗∗∗} |
0.67–0.94 | 0.88^{∗∗∗} |
0.70–0.95 | 0.90^{∗∗∗} |
0.75–0.96 | 0.89^{∗∗∗} |
0.73–0.96 | 0.56^{∗∗} |
0.14–0.81 | 0.75^{∗∗∗} |
0.46–0.90 | 0.89^{∗∗∗} |
0.65–0.96 |

Left | 0.84 | 0.87^{∗∗∗} |
0.70–0.95 | 0.89^{∗∗∗} |
0.72–0.96 | 0.84^{∗∗∗} |
0.62–0.94 | 0.83^{∗∗∗} |
0.60–0.93 | 0.62^{∗∗} |
0.22–0.84 | 0.70^{∗∗∗} |
0.14–0.90 | 0.70^{∗∗} |
0.35–0.88 |

Right | 0.87 | 0.85^{∗∗∗} |
0.65–0.94 | 0.92^{∗∗∗} |
0.79–0.97 | 0.86^{∗∗∗} |
0.66–0.94 | 0.84^{∗∗∗} |
0.62–0.94 | 0.63^{∗∗} |
0.25–0.85 | 0.58^{∗∗} |
0.12–0.83 | 0.69^{∗∗} |
0.33–0.87 |

Left | 0.78 | 0.62^{∗∗} |
0.22–0.84 | 0.85^{∗∗∗} |
0.65–0.94 | 0.83^{∗∗∗} |
0.60–0.93 | 0.76^{∗∗∗} |
0.45–0.90 | 0.61^{∗∗} |
0.21–0.83 | 0.71^{∗∗∗} |
0.38–0.88 | 0.68^{∗∗} |
0.33–0.87 |

Right | 0.83 | 0.74^{∗∗∗} |
0.42–0.90 | 0.82^{∗∗∗} |
0.58–0.93 | 0.80^{∗∗∗} |
0.52–0.92 | 0.77^{∗∗∗} |
0.48–91 | 0.69^{∗∗} |
0.32–0.87 | 0.41^{∗} |
−0.08–0.74 | 0.82^{∗∗∗} |
0.56–0.93 |

^{∗}p < 0.05;

^{∗∗}p < 0.01;

^{∗∗∗}p < 0.001.

Overlapped 3D volume representations of the reconstructed fiber bundles at the two scanning time-points in a representative subject. Please note that these do not reflect the wDSC values. Blue = time 1, red = time 2, purple indicates the overlap. AF, arcuate fasciculus; ILF, inferior longitudinal fasciculus, IFOF, inferior fronto-occipital fasciculus; uncinate fasciculus; L, Left; R, Right.

Table

In Figure

Bland-Altman Plots for the FA metric in all four fiber bundles, bilaterally. The Y axis represents the mean difference between the measurements at the two timepoints and the X axis represents the mean of these measures. The upper and lower dashed lines represent the two limits of agreements at ± 2 standard-deviations of the mean of differences (i.e., the 95% confidence interval). The solid line represents the mean of the differences between the two timepoints. The dots represent the individual subjects. FA, fractional anisotropy; AF, arcuate fasciculus; ILF, inferior longitudinal fasciculus; IFOF, inferior fronto-occipital fasciculus; UF, uncinate fasciculus; T1, time 1; T2, time 2.

The aim of this study was to demonstrate the test-retest reliability of the reconstruction and micro- and macro-structural characteristics of major WM language fiber bundles using probabilistic CSD-tractography based on HARDI data. The dMRI data were obtained on a group of healthy subjects at two timepoints, spanning one week. First, the results demonstrated that all the reconstructed fiber bundles have a good overlap between the two timepoints. Secondly, tract-specific measures usually used in studying microstructural WM characteristics, such as FA, MD, RD, and AD, as well as the macrostructural measure MLS showed good to excellent test-retest reliability in the AF, ILF, IFOF, and UF, bilaterally. Volume, another macrostructural property, showed good to excellent reproducibility for some fiber bundles (AF, ILF, as well as the IFOF, and UF in the left hemisphere) but only fair reproducibility for others (IFOF and UF in the right hemisphere). NuFO showed good test-retest reliability for all fiber bundles, except the ILF which showed only fair test-retest reliability. Our results agree with and, at the same time, critically expand on previous studies that investigated the test-retest reliability of probabilistic CSD-tractography (

Diffusion MRI tractography is presently the only method that allows the reconstruction of WM fiber bundles

Regarding the first question, our data seem to provide an affirmative response. The obtained wDSC values determining the degree of overlap of the bundles reconstructed at the two timepoints using probabilistic CSD-tractography ranged between 0.71 and 0.87. Based on the minimum value (0.70) of Dice found in the two studies which used this metric to assess the test-retest reliability of CSD-based reconstruction of WM tracts (

Our study also confirms the reproducibility of tensor metrics and MLS. Test-retest reliability of tensor metrics (FA, MD, RD, and AD) has been previously studied using DTI-based tractography (

Even though the present results are very promising, particularly for the tensor metrics and MLS, future studies should be designed in order to confirm our findings. First, these results should be reproduced in larger groups. Secondly, the use of CSD allows to resolve multiple fiber orientations at reasonable angles with a properly data-driven response function at lower ^{2}, ^{2}, ^{2}, or ^{2}, and ^{2}, could help to interpret the differences obtained in the present study by considering other available measures, such as intracellular, extracellular, and isotropic volume (

In conclusion, in an era where initiatives to collect dMRI longitudinal data are multiplying and fiber tracking is considered one of the most popular tools to follow changes in the language network over time, the question of test-retest reliability of dMRI tractography is of paramount importance. Our study provides critical evidence indicating the test-retest reliability of probabilistic CSD-tractography. As in previous studies which demonstrated test-retest reliability of TBSS or DTI-tractography (e.g.,

MB drafted the manuscript, contributed to the design of the study, reviewed the literature, and collected, analyzed, and interpreted the data. SB and KM designed the study, contributed to data collection, supervised data analysis and interpretation, and contributed to the drafting of the manuscript. MD contributed to the design of the study and to the development of the tractography pipeline. AD contributed to the design of the study and MRI data acquisition. CB and MC helped with the data analysis. CB, MD, AB, J-CH, and SD-G developed the tractography pipeline. All authors revised the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

We would like to thank the participants who accepted to take part of this study. We also would like to thank Pamela Ross, Audrey Borgus, and Amélie Brisebois for their valuable assistance with the recruitment and testing of participants. We also thank the radiology team at Hôpital du Sacré-Coeur de Montréal (HSCM) for their support with the scanning of participants.

The Supplementary Material for this article can be found online at: