EEG microstate-EMG associations during freezing of gait in Parkinson’s disease

Background Freezing of gait (FOG) is one of the most disabling motor symptoms in Parkinson’s disease (PD), yet its neural mechanisms remain incompletely understood. Objective This study investigated whether electroencephalogram (EEG) microstate dynamics are altered during FOG, whether these alterations are associated with changes in muscle activity, and whether electromyography (EMG)-derived features can predict FOG onset. Methods Walking EEG and EMG recordings from PD patients with FOG were analyzed. EEG microstate metrics and EMG root mean square (RMS) amplitudes were compared across baseline, preFOG, FOG, and postFOG phases. Linear mixed-effects modeling was used to assess EEG-EMG coupling, and RMS-derived features were used to train random under-sampling boosting classifiers for FOG prediction using leave-one-subject-out cross-validation. Results Among the three identified microstates, a hemispheric-gradient microstate exhibited significant modulation across gait phases. Its coverage was significantly reduced during FOG compared with preFOG. EMG RMS demonstrated a similar temporal pattern, reaching its lowest level during FOG and showing significant reductions relative to both baseline walking and preFOG phases. Linear mixed-effects modeling revealed a significant negative association between hemispheric-gradient microstate coverage and EMG RMS. Furthermore, RMS-derived features enabled prediction of FOG onset, achieving areas under the receiver operating characteristic curve of 0.62, 0.63, 0.65, 0.68, 0.77, and 0.90 at prediction horizons of 5, 4, 3, 2, 1, and 0.5 s, respectively. Predicted FOG risk increased progressively as FOG onset approached. Conclusions A hemispheric-gradient EEG microstate showed FOG-related modulation and an association with EMG activity, suggesting a potential link between cortical state dynamics and motor output. RMS-derived EMG features showed potential for short-horizon FOG prediction in offline analyses. Clinical trial number Not applicable.