Unraveling the MRI-Based Microstructural Signatures Behind Primary Progressive and Relapsing-Remitting Multiple Sclerosis Phenotypes

Abstract

Background: The mechanisms driving primary progressive and relapsing-remitting multiple sclerosis (PPMS/RRMS) phenotypes are unknown. Magnetic resonance imaging (MRI) studies support the involvement of gray matter (GM) in the degeneration, highlighting its damage as an early feature of both phenotypes. However, the role of GM microstructure is unclear, calling for new methods for its decryption.

Purpose: To investigate the morphometric and microstructural GM differences between PPMS and RRMS to characterize GM tissue degeneration using MRI.

Study type: Prospective cross-sectional study.

Subjects: Forty-five PPMS (26 females) and 45 RRMS (32 females) patients.

Field strength/sequence: 3T scanner. Three-dimensional (3D) fast field echo T1-weighted (T1-w), 3D turbo spin echo (TSE) T2-w, 3D TSE fluid-attenuated inversion recovery, and spin echo-echo planar imaging diffusion MRI (dMRI).

Assessment: T1-w and dMRI data were employed for providing information about morphometric and microstructural features, respectively. For dMRI, both diffusion tensor imaging and 3D simple harmonics oscillator based reconstruction and estimation models were used for feature extraction from a predefined set of regions. A support vector machine (SVM) was used to perform patients' classification relying on all these measures.

Statistical tests: Differences between MS phenotypes were investigated using the analysis of covariance and statistical tests (P < 0.05 was considered statistically significant).

Results: All the dMRI indices showed significant microstructural alterations between the considered MS phenotypes, for example, the mode and the median of the return to the plane probability in the hippocampus. Conversely, thalamic volume was the only morphometric feature significantly different between the two MS groups. Ten of the 12 features retained by the selection process as discriminative across the two MS groups regarded the hippocampus. The SVM classifier using these selected features reached an accuracy of 70% and a precision of 69%.

Data conclusion: We provided evidence in support of the ability of dMRI to discriminate between PPMS and RRMS, as well as highlight the central role of the hippocampus.

Level of evidence: 2 TECHNICAL EFFICACY STAGE: 3.

Keywords: 3D-SHORE; EDSS; SVM; diffusion MRI; gray matter; morphometry.