Exploring inter-subject variability in the emotional domain to characterize typical and atypical neural responses in longitudinal settings: insights from Eye Movement Desensitization and Reprocessing rehabilitative treatment

Background The emotional domain is often impaired in several neurological condition and consequently targeted by rehabilitation interventions. When investigating the neural underpinnings of this domain by functional magnetic resonance imaging (fMRI) a high inter-subject variability (ISV) in neural activity is measured. Rather than reflecting mere noise, such variability may play a critical role in revealing neuroplasticity mechanisms associated with therapeutic interventions, particularly in clinical populations. Objective The aim of this study was to compare a conventional fMRI data analysis approach based on the General Linear Model (GLM), which relies on group-averaged effects, with an innovative method based on threshold-weighted overlap maps (OMth-w) that explicitly exploits ISV to detect neuroplasticity-related changes in fMRI signals before and after a rehabilitation intervention. Methods Thirty healthy controls (HC) and thirteen people with multiple sclerosis (pwMS) undergoing Eye Movement Desensitization and Reprocessing (EMDR) treatment for depression were enrolled. Participants performed an emotional visual stimulation task during fMRI, comprising two conditions with positive and negative emotions. fMRI data were acquired at two time points (pre- and post-EMDR for pwMS). Data analysis compared two approaches: (1) conventional second-level general linear model (GLM) analyses using paired t-tests; (2) normalized individual threshold-weighted overlap maps (OMth-w) to quantify ISV. Results Standard GLM analyses did not reveal significant longitudinal changes in either group. In contrast, the OMth-w approach showed increased neural variability over time in HC and reduced variability in pwMS following EMDR treatment, concomitant with a significant reduction in depressive symptoms (p < 0.001). Conclusions These results underscore the relevance of variability-based analytical approaches for identifying neuroplasticity-related changes that may remain undetected using traditional mean-based inference methods.