Resting-state functional connectivity correlates of gait and turning performance in multiple sclerosis: a multivariate pattern analysis

Abstract

Multiple sclerosis (MS) often leads to mobility impairments, yet the neural mechanisms underlying these deficits remain poorly understood. This study examined whether resting-state functional connectivity (rs-FC) differs between people with MS (PwMS) and healthy controls in relation to spatiotemporal mobility performance. We hypothesized that group differences within the default mode (DMN), frontoparietal (FPN), somatomotor (SN), and visual (VIS) networks would be associated with gait and turning metrics. Twenty-nine PwMS and 28 matched controls completed a two-minute walk test, 180° walking turns, and 360° in-place turns at natural and fast speeds. fMRI data were analyzed using multivariate pattern analysis (MVPA) and post-hoc seed-to-voxel analyses for gait speed, cadence, double support time, stride length, turn duration, peak velocity, and turn angle. PwMS exhibited slower gait speed, shorter stride length, and impaired 360° turning, but no group differences in cadence, double support, or 180° turn metrics. MVPA revealed rs-FC differences across DMN, FPN, SN, and VIS networks. While rs-FC differences were evident for walking metrics, within-group associations were not significant. In contrast, 360° turn angle showed distinct within-group rs-FC associations, particularly involving VAN and DAN networks. These findings highlight turning as a sensitive task for capturing functional neural differences in MS.

Keywords: Gait performance; Multiple sclerosis (MS); Multivariate pattern analysis (MVPA); Neural network connectivity; Resting-state functional connectivity; Turning kinematics.

Fig. 1

(a) Six seed clusters identified from the MVPA. (b) Post-hoc seed-to-voxel correlation clusters for cadence, gait speed, and stride length displaying differences between groups for seed cluster B. (c) Post-hoc seed-to-voxel correlation clusters for cadence, gait speed, and stride length differences between groups for seed cluster D. Color bar represents statistical significance.

Fig. 2

(a) Three seed clusters identified from the MVPA. (b) Post-hoc seed-to-voxel correlation clusters for 360° turn angle differences between groups for Seed Cluster (A) (c) Post-hoc seed-to-voxel correlation clusters for 360° turn angle differences between groups for Seed Cluster (B) Color bar represents statistical significance.

Fig. 3

(a) Two seed clusters identified from the MVPA. (b) Post-hoc seed-to-voxel correlation clusters for 360° turn angle differences between groups for Seed Cluster A. Color bar represents statistical significance.

Fig. 4

(A) In healthy controls, stronger anticorrelations within the VAN were associated with larger 360° turn angles. In PwMS, stronger positive correlations were linked to larger turn angles. (B) In healthy controls, stronger positive correlations between the DMN, limbic network, and the DAN were associated with larger 360° turn angles. In PwMS, stronger anticorrelations between these networks were linked to larger turn angles.

Fig. 5

(A–B) In healthy controls, stronger positive connectivity between the VIS and the DAN was associated with larger 360° turn angles, while in PwMS, stronger anticorrelations in these regions were linked to larger turn angles. (C) In healthy controls, stronger positive connectivity between the DMN and the DAN was associated with larger 360° turn angles, whereas in PwMS, stronger anticorrelations in these regions were linked to larger turn angles.