Macromolecular proton fraction (MPF) is a key biophysical parameter determining the magnetization transfer (MT) effect within the two-pool model¹ and defined as a relative amount of macromolecular protons involved into cross-relaxation with free water protons. A fast method allowing whole-brain MPF mapping based on a single MT-weighted image was recently developed². This technique demonstrated a promise as a robust clinically-targeted quantitative myelin imaging approach in multiple sclerosis³ and mild traumatic brain injury⁴ studies. A new modification of the single-point method¹ based on a synthetic reference image for data normalization has been recently reported as the fastest possible MPF mapping technique⁵. This technique requires only three source images (MT-, T1-, and proton density (PD)-weighted) to reconstruct an MPF map and enables whole-brain high-resolution MPF mapping with clinically feasible acquisition time. Due to high contrast between white matter (WM) and gray matter (GM), high-resolution three-dimensional (3D) MPF maps potentially can provide not only the means for quantitative assessment of myelination but also source images for neuroanatomical volume measurements. The objective of this study was to characterize scan-rescan reproducibility of both MPF and volume measurements for WM and GM based on automated segmentation of high-resolution whole-brain MPF maps.

Participants: Eight healthy volunteers (mean age ± standard deviation (SD): 44.6 ± 12.2 years, age range: 29-66 years, 4 females/4 males) underwent two repeated scans with the interval between 6 and 12 months.

MRI Protocol: MRI data acquisition was performed on a 3.0 Tesla Philips Achieva scanner with an 8-channel phased-array head coil. 3D PD- and T1-weighted spoiled gradient-echo images were acquired with TR=21 ms and flip angle (FA)= 4° and 25°, respectively. 3D MT-weighted images were acquired with TR=28 ms and FA=10°. Off-resonance saturation was achieved by applying the single-lobe sinc pulse with Gaussian apodization, offset frequency 4 kHz, effective saturation FA=560°, and duration 12 ms. All images were acquired with non-selective excitation, FOV = 240x240x180 mm3, and isotropic voxel size of 1.25x1.25x1.25 mm³. 3D dual-echo B0 maps (TR/TE1/TE2=20/2.3/3.3 ms, FA=10°)⁶ and AFI B1 maps (TR1/TR2=40/160 ms, TE=2.3 ms, FA = 60°)⁷ were obtained in the same geometry with voxel sizes of 2.5x2.5x2.5 mm³ and 2.5x2.75x5.0 mm³, respectively. Parallel imaging (SENSE) was used for all scans in two phase encoding directions with acceleration factors 1.5 (anterior-posterior) and 1.2 (left-right). Acquisition time was 19 min for the entire protocol.

Image reconstruction and analysis: MPF maps were reconstructed using the recently described algorithm⁵ comprising the following steps: reconstruction of R1 and PD maps from two variable FA images with B1 correction, computation of a synthetic reference image from R1, PD, and B1 maps, and reconstruction of an MPF map from an MT-weighted image normalized to the synthetic reference image according to the single-point method². Before reconstruction, non-brain tissues were removed by applying a brain mask created from the PD-weighted image using the brain extraction tool (BET)⁸ in FSL software. Cerebrospinal fluid (CSF) was segmented out from R1 maps based on a threshold value of R1 = 0.33 s^-1. MPF maps were segmented into WM and GM. using an automated segmentation tool (FAST)⁹ in FSL software with the Markov random field weighting parameter 0.25. To account for potentially incomplete CSF segmentation and exclude voxels containing partial volume of CSF (PVCSF), the third mixed tissue class was also prescribed. Tissue classes were defined by specifying initial tissue-type priors.

Statistical analysis: Scan-rescan agreement between global WM and GM MPF and volume measurements was assessed using Bland-Altman plots. To estimate variability between measurements, within-subject coefficients of variation (CoV) were calculated.

An example high-resolution whole-brain 3D MPF map (Figure 1) demonstrates sharp WM-GM contrast and clear definition of anatomical details. Figure 2 illustrates a scheme of brain segmentation with color-coded tissue masks. Bland-Altman plots characterizing repeatability of brain tissue MPF and volume measurements are shown in Figures 3 and 4. No significant bias was detected in any measurement. MPF measurements were characterized by remarkably low variability with CoV of 1.5% and 1.1% for WM and GM, respectively. CoV for volume measurements were 2.1% for WM and 3.2% for GM.This study demonstrates that whole-brain MPF measurements in WM and GM are highly reproducible, and their variability is substantially lower than that for a variety of other quantitative MRI parameters¹⁰. Fast high-resolution 3D MPF mapping provides a potential for simultaneous characterization of myelination and volumetric changes in brain tissues, which enables an “all-in-one” solution for a variety of neuroscience studies where quantitative assessment of both demyelination and atrophy is of interest.