Obstacle Avoidance in Healthy Adults and People With Multiple Sclerosis: Preliminary fNIRS Study

Abstract

This study examined how gait adaptation during predictable and non-predictable obstacle avoidance affects the sensorimotor network in both healthy controls (HC) and persons with multiple sclerosis (pwMS). We utilized fNIRS measurements of HbO2 and HHb to estimate cortical activations and connectivity networks, which were then analyzed using power spectral density (PSD) and partial directed coherence (PDC). The findings revealed distinct patterns of cortical activation and connectivity for each task condition in both groups. Healthy individuals displayed lower cortical activations in the bilateral motor cortex (MC) during non-predictable obstacle avoidance, indicating efficient neural processing. On the other hand, pwMS exhibited lower cortical activations across most brain areas during non-predictable tasks, suggesting potential limitations in neural resource allocation. When tasks were combined, pwMS demonstrated higher cortical activation across all recorded brain areas compared to HC, indicating a compensatory mechanism to maintain gait stability. Functional connectivity analysis revealed that pwMS recruited higher bilateral somatosensory association cortex (SAC) than HC, whereas healthy individuals engaged more bilateral premotor cortices (PMC). These findings suggest alterations in sensorimotor integration and motor planning in pwMS. Four machine learning models (KNN, SVM, DT, and DA) achieved high classification accuracies (92-99%) in differentiating between task conditions within each group. These results highlight the potential of integrating fNIRS-based cortical activation and connectivity measures with machine learning as biomarkers for MS-related impairments in cognitive-motor interaction. Such biomarkers could aid in predicting future mobility decline, fall risk, and disease progression.

Keywords

• Multiple sclerosis
• fNIRS
• PSD
• connectivity network
• machine learning