Improving the Detection of Myelin Integrity in Multiple Sclerosis Using Selective Inversion Recovery for MRI With Quantitative Magnetization Transfer

Abstract

Background: Selective inversion recovery quantitative magnetization transfer (SIR-qMT)-derived macromolecular to free water pool size ratio (PSR) and diffusion tensor imaging (DTI)-derived radial diffusivity (RD) are potential metrics for assessing myelin integrity in multiple sclerosis (MS). However, establishing their accuracy in identifying tissue injury is essential for clinical translation.

Purpose: To compare the accuracy and Cohen's effect size (ES) of PSR and RD in detecting and quantifying tissue injury in early MS.

Study type: Cross-sectional prospective study.

Subjects: Fourty-three subjects with newly diagnosed MS (mean age 38 ± 11 years, 70% females) and 18 age- and sex-matched healthy controls (HCs; age 38 ± 12 years, 62.5% females).

Field strength/sequence: 3-T MRI using T₁-weighted (T₁-w) turbo spin echo, T₂-w fluid-attenuated inversion recovery (FLAIR), DTI, and SIR-qMT sequences.

Assessment: T₂-lesions were identified as hyperintense on T₂-w-FLAIR, and chronic black holes (cBHs) by simultaneous T₂-w-FLAIR hyperintensity and T₁-w hypointensity. Regions of interest (ROIs) in normal-appearing white matter (NAWM) were classified as proximal (p) or distant (d) to lesions, while normal white matter (NWM) was identified in HCs. PSR and RD values of T₂-lesions and cBHs were compared to their matched p/dNAWM and NWM in HCs. Comparisons were also made between T₂-lesions and cBHs.

Statistical tests: Receiver operating characteristic curves evaluated metric accuracy, and paired t tests compared ES values of PSR and RD, with significance set at P < 0.050.

Results: We identified 823 T₂-lesions, 392 cBHs, 426 p-, and 213 d-NAWM ROIs in patients, and 162 NWM ROIs in HCs. PSR differed significantly in all comparisons, while RD was differed in all except cBHs vs. T₂-lesions (P = 0.051). PSR had significantly higher accuracy in differentiating T₂-lesions from p/dNAWM and NWM, with a larger ES when comparing T₂-lesions to p/dNAWM and NWM and cBHs to pNAWM and NWM.

Data conclusion: PSR offers superior accuracy and ES over RD in detecting tissue injury in MS.

Level of evidence: 1 TECHNICAL EFFICACY: Stage 2.

Keywords: diffusion tensor imaging; multiple sclerosis; myelin; selective inversion recovery quantitative magnetization transfer imaging.

FIGURE 1

Anatomical sequences and parametric maps. T₂‐weighted (T₂‐w) fluid‐attenuated inversion recovery (a) and T₁‐w turbo spin echo (b) sequences were used for tissue class identification. Co‐registered macromolecular‐to‐free water pool size ratio (c), relative color‐scaled heat map in (e), and radial diffusivity (d), relative color‐scaled heat map in (f) parametric maps were derived from selective inversion recovery quantitative magnetization transfer and diffusion tensor imaging, respectively, and used to derive quantitative measurements. The white arrow represents T₂‐lesions while the black arrows represent chronic black holes.

FIGURE 2

Examples of the comparators' pairs. Differences in the macromolecular‐to‐free water pool size ratio (PSR) and radial diffusivity (RD) were computed between the following pairs: T₂‐lesions (white arrow) and corresponding proximal (white boxes anterior and posterior to the lesions in a1)/distant (white box contralateral to the lesions in a1) normal appearing white matter (NAWM) and normal white matter (NWM) in healthy controls (white box in a2); chronic black holes (cBHs) (black arrow in b1) and corresponding proximal (black boxes posterior and anterior to the cBH in b1)/distant NAWM (black box contralateral to the cBH in b1) and NWM (black box in b2); cBHs (black arrow in b1) and contralateral T₂‐lesion (black arrow in a1)

FIGURE 3

Differences in macromolecular‐to‐free water pool size ratio (PSR) between different types of tissues. In (a) we present the PSR values measured in chronic black holes (cBHs) and anatomically matched T₂‐lesions, proximal (p) and distant (d) normal‐appearing white matter (NAWM), and normal white matter (NWM) of healthy controls. In (b) we present PSR values measured in a different set of T₂‐lesions and anatomically matched pNAMW and dNAWM, and NWM. T₂‐lesions, pNAWM, dNAWM, and NWM in (a) indicate the regions of interest matched to cBHs. pNAWM ROIs and dNAWM, and NWM in (b) indicate the ROIs matched to a different subset of T₂‐lesions. In the violin plot, the height represents the values of the metric, the width indicates the number of lesions, and the horizontal black line represents the PSR median value.

FIGURE 4

Differences in radial diffusivity (RD) between different types of tissues. In (a) we present the RD values measured in chronic black holes (cBHs) and anatomically matched T₂‐lesions, proximal (p) and distant (d) normal‐appearing white matter (NAWM), and normal white matter (NWM) of healthy controls. In (b) we present RD values measured in T₂‐lesions and anatomically matched pNAWM, dNAWM and NWM. T₂‐lesions, pNAWM, dNAWM and NWM in (a) indicate the regions of interest matched to cBHs. pNAWM and dNAWM, and NWM in (b) indicate the ROIs matched to a different subset of T₂‐lesions. In the violin plot, the height represents the values of the metric, the width indicates the number of lesions, and the horizontal black line represents the RD median value.

FIGURE 5

Receiver operating characteristic (ROC) curves. The following pairs are presented: T₂‐lesions and normal white matter (NWM) in healthy controls (HCs, a); chronic black holes (cBHs) and NWM in HCs (b); T₂‐lesions and proximal normal appearing white matter (pNAWM) (c); T₂‐lesions and distant (d) NAWM (d); cBHs and pNAWM (e); cBHs and dNAWM (f). The inverted RD value (1/RD) is presented to facilitate the visualization and the comparisons between PSR and RD. The blue curve indicates PSR, the grey curve indicates 1/RD and the green line is the reference line (50%).

FIGURE 6

Macromolecular to free‐pool size ratio (PSR) and radial diffusivity (RD) effect size (ES). We report the ES of pairs for which PSR performed significantly better than RD. The following pairs are presented: T₂‐lesions and normal white matter (NWM) of healthy controls (a); T₂‐lesions and distant normal‐appearing white matter (dNAWM, b); T₂‐lesions and proximal (p) NAWM (c); chronic black holes (cBHs) and NWM (d); cBHs and pNAWM (e). The blue boxes represent PSR values while the grey boxes represent RD values. In the violin plot, the height represents the values of the metric, the width indicates the number of lesions, and the horizontal black line represents the metric median value.