Predicting Gait Speed Using Brain Functional Connectivity Maps During Walking

Understanding the complex relationship between brain connectivity and gait dynamics is critical for elucidating the role of neural control in locomotion towards the early detection of gait abnormalities and personalized treatment of movement-related dysfunction. This study quantitatively investigates the association between brain functional connectivity and gait speed by leveraging EEG-derived Partial Directed Coherence (PDC) maps and AI-based classification models. By capturing and analyzing gait-related neural interactions, this work offers a novel approach for predicting gait speed and understanding the vital role of brain connectivity in locomotion. Gait data were collected from 8 healthy participants walking at three predefined speeds (0.5 m/s, 0.75 m/s, and 1 m/s) on a treadmill. The preprocessed EEG signals were converted into brain functional connectivity maps and used as inputs for a convolutional neural network (CNN). A Leave-One-Subject-Out cross-validation strategy was applied to ensure robust and subject-independent performance evaluation. The model achieved an average classification accuracy of 60.87%, with higher precision (0.76) and F1 scores (0.64) observed at faster gait speeds, indicating the capability of neural networks in reflecting motor control. This work demonstrates the potential for integrating brain functional connectivity and AI models in developing personalized gait diagnostics and rehabilitation tools for clinical applications and beyond.