High‐frequency longitudinal white matter diffusion‐ and myelin‐based MRI database: Reliability and variability

Abstract Assessing the consistency of quantitative MRI measurements is critical for inclusion in longitudinal studies and clinical trials. Intraclass coefficient correlation and coefficient of variation were used to evaluate the different consistency aspects of diffusion‐ and myelin‐based MRI measures. Multi‐shell diffusion and inhomogeneous magnetization transfer data sets were collected from 20 healthy adults at a high‐frequency of five MRI sessions. The consistency was evaluated across whole bundles and the track‐profile along the bundles. The impact of the fiber populations on the consistency was also evaluated using the number of fiber orientations map. For whole and profile bundles, moderate to high reliability of diffusion and myelin measures were observed. We report higher reliability of measures for multiple fiber populations than single. The overall portrait of the most consistent measurements and bundles drawn from a wide range of MRI techniques presented here will be particularly useful for identifying reliable biomarkers capable of detecting, monitoring and predicting white matter changes in clinical applications and has the potential to inform patient‐specific treatment strategies.


MRI acquisition
There were no upgrades or changes in MR scanner hardware or software during the period of this study. Procedures were standardized to limit session variability. The positioning of the head followed the standard operating procedure and the definition of the field of view (FOV) during the acquisition followed an explicit procedure; both were checked by the first author. Subject movements and evidence of motion artifacts were checked during the acquisition. No subject was excluded during acquisition, and MRI images had to be repeated in one subject during a session due to artifacts.

MT Saturation pulse
10 Hann pulses of 0.9 ms duration with 1.5 ms interval

Frequency offset of +/-7000 Hz
Echoes -Echo spacing 3 -6 ms *The directions are specified in the standard way in coordinates of the patient (LPH).

Inhomogeneous Magnetization Transfer imaging (ihMT)
Acquisition The multi-echo segmented-EPI sequence used for ihMT was using an EPI factor of 5 and allowed the acquisition and reconstruction of 3 echoes (echo spacing = 6.0 ms), for a total number of k-space lines per excitation of 15. However, only data from the first of the three echoes (TE = 3.6 ms) was used for our analysis. It is important to note that the other 2 echoes could be acquired without any scan time penalty, given that the minimum repetition time (TR) was SARlimited (see Figure 1). The root means square B1 (B1rms) are computed using the equations from the Soustelle et al., 2022: -the B1rms over the sequence TR was 2.7 uT (equation 2).

Figure 1 Sequence diagram for one TR
ihMT metrics Here we computed the ihMTdR1sat metric, which is not as widely used nowadays as the ihMTsat metric. However, the ihMT literature is still evolving rapidly and the data acquisition and processing for that project started before most of the literature on ihMTsat was published. The ihMTdR1sat metric as defined in Varma et al., 2015 thus seemed a good choice when looking at metrics that could be extracted from the ihMT data, in addition to the standard ihMTR. The calculation of ihMTdR1sat was performed following the signal derivation introduced in Varma et al., 2015 with one small generalization to allow for a reference image acquired with a different TR (and not only a different flip angle), as regularly performed for MTsat (or ihMTsat) calculations.
where Ssingle is the signal with MT saturation at a single frequency (average of the MT saturation with positive or negative frequency), Sdual is the MT signal with alternating frequency saturation, αMT and αT1 are the flip angles of the MT-weighted and T1-weighted acquisitions, respectively and TRMT and TRT1 are the repetition times of the MT-weighted and T1-weighted acquisitions, respectively. S0 is the signal without saturation, at the same flip angle and TR as the acquisition with MT saturation.
In addition, when calculating both ihMTdR1sat and ihMTsat from the same dataset, the two metrics show a very high correlation (Figure 2).

Figure 2 ihMT metrics correlations.
Correlation between ihMTsat and ihMTdR1sat computed from a single subject and single visit.

Figure 4 Processes and analyses overview.
In A (blue), MTI input and output files using ihMT flow. In B (blue), the DWI process using Tractoflow to generate the tractograms and diffusion measures maps, the bundles' virtual segmentation using RecoBundlesX to obtain whole bundle mask, and bundles resampling using the Tractometry flow which provides bundle profile masks. In C (light blue), diffusion common space generation using ANTs. In D (green), consistency analyses of diffusion and myelin measures from the whole and profile bundle masks. In E (red) processes that take the NuFO map generated in B and the whole and profile bundle mask generated in D to separate them into single and multi compartments masks and perform consistency analyses of diffusion and myelin measures from these masks.

Visual quality criteria
The visual quality assessment procedure was applied to the main steps according to the following criteria:

MRI images
Step Exclusion criteria

T1, MTI, DWI Raw data
Presence of artifacts that cannot be corrected Incorrect field of view Too high noise in the image Broken image Missing part of the brain (T1, MTI) Incorrect distribution of gradient (DWI) High inhomogeneity (MTI)

T1, MTI, DWI Brain extraction
Eyes included in the brain mask Exclusion of a part of the brain Inclusion of a large part of the background in the brain mask

Impact of resampling on bundles volume
To ensure that each section of bundles contains enough voxels to assess consistency measurements, we extracted the volume of each section corresponding to the bundle profile analyses. A minimum threshold of 1000 voxels (dotted line) was used to perform analyses, therefore, sections of bundles with fewer than 1000 voxels were excluded. Only the volume of section 10 of the cingulum is below the threshold (circle and red arrow). This section was therefore excluded from the analyses.