Clinical correlates of R1 relaxometry and magnetic susceptibility changes in multiple sclerosis: a multi-parameter quantitative MRI study of brain iron and myelin

Abstract

Objectives: The clinical impact of brain microstructural abnormalities in multiple sclerosis (MS) remains elusive. We aimed to characterize the topography of longitudinal relaxation rate (R1) and quantitative susceptibility (χ) changes, as indices of iron and myelin, together with brain atrophy, and to clarify their contribution to cognitive and motor disability in MS.

Methods: In this cross-sectional study, voxel-based morphometry, and voxel-based quantification analyses of R1 and χ maps were conducted in gray matter (GM) and white matter (WM) of 117 MS patients and 53 healthy controls. Voxel-wise between-group differences were assessed with nonparametric permutation tests, while correlations between MRI metrics and clinical variables (global disability, cognitive and motor performance) were assessed both globally and voxel-wise within clusters emerging from the between-group comparisons.

Results: MS patients showed widespread R1 decrease associated with more limited modifications of χ, with atrophy mainly involving deep GM, posterior and infratentorial regions (p < 0.02). While R1 and χ showed a parallel reduction in several WM tracts (p < 0.001), reduced GM R1 values (p < 0.001) were associated with decreased thalamic χ (p < 0.001) and small clusters of increased χ in the caudate nucleus and prefrontal cortex (p < 0.02). In addition to the atrophy, χ values in the cingulum and corona radiata correlated with global disability and motor performance, while focal demyelination correlated with cognitive performance (p < 0.04).

Conclusions: We confirmed the presence of widespread R1 changes, involving both GM and WM, and atrophy in MS, with less extensive modifications of tissue χ. While atrophy and χ changes are related to global and motor disability, R1 changes are meaningful correlates of cognition.

Key points: • Compared to healthy controls, multiple sclerosis patients showed R1 and χ changes suggestive of iron increase within the basal ganglia and reduced iron and myelin content within (subnuclei of) the thalamus. • Thalamic volume and χ changes significantly predicted clinical disability, as well as pulvinar R1 and χ changes, independently from atrophy. • Atrophy-independent R1 and χ changes, suggestive of thalamic iron and myelin depletion, may represent a sensitive marker of subclinical inflammation.

Keywords: Atrophy; Magnetic resonance imaging; Multiple sclerosis; Quantitative susceptibility; Relaxometry.

Fig. 1

Flowchart showing inclusion and exclusion criteria

Fig. 2

Workflow illustrating the main image processing and analysis steps. Initially, quantitative maps were computed and mapped onto the corresponding T1-weighted volumes and demyelinating lesions were automatically segmented on FLAIR images. For voxel-based analyses, T1-weighted volumes were segmented into different tissue classes and normalized to a 1mm isotropic template in MNI space, with the estimated spatial transformations also applied to quantitative maps. Before entering voxel-wise statistical analyses, normalized gray matter and white matter probability maps were modulated and smoothed, while normalized R1 and χ maps were smoothed via a tissue-weighted smoothing procedure to account for the partial volume contribution of tissue density in each voxel. Using lesion and tissue class masks, global brain volumes and median values of R1 and χ in normal-appearing tissues were also obtained

Fig. 3

Results of the between-group voxel-wise comparisons. A lesion probability map (LPM), obtained by summing all the binary lesion masks and dividing by the number of patients to give a lesion probability at each voxel, is presented (with a 5% probability threshold, upper left panel), along with clusters of significant between-group difference in terms of volume (upper right panel), R1 and χ (lower panels) values for both the MS > HC (red-yellow) and MS < HC (blue-light blue) contrasts, all superimposed on axial sections of the average T1-weighted volume in the MNI space. For volume, R1, and χ maps, pooled results of the GM and WM analyses are shown. Images are in radiological orientation

Fig. 4

Effect size maps of between-group differences. Effect size (Cohen’s d) maps of between-group differences in terms of volume, R1, and χ values (from left to right) are presented, superimposed on axial sections of the average T1-weighted volume in the MNI space. Positive effect size values refer to the MS < HC contrast. For volume, R1, and χ maps, pooled results of the GM and WM analyses are shown. Images are in radiological orientation

Fig. 5

Results of the voxel-wise correlations with clinical variables. Clusters of significant association between MRI metrics and EDSS, SDMT, and motor (from top to bottom) scores are presented, superimposed on sagittal, coronal, and axial (from left to right) sections of the average T1-weighted volume in the MNI space. Images are in radiological orientation