EEG microstate features under visual feedback gain conditions exhibit high sensitivity in identifying early Parkinson's disease patients.

backgroundEarly Parkinson's disease (PD) presents with subtle symptoms and lacks specific diagnostic methods. Clinical diagnosis primarily relies on subjective assessment, with confirmation often occurring at mid-to-late stages. Therefore, identifying objective and quantifiable biomarkers to assist in early PD diagnosis and intervention is of significant clinical value.

methodThis study recruited 20 early PD patients and 18 healthy controls who performed a visually guided motor-cognitive task (VMC) under three visual feedback gain conditions (low, medium, and high), while electroencephalography (EEG) and grip strength data were recorded simultaneously. Subsequently, EEG microstate analysis was used to extract temporal dynamic parameters (Duration, Occurrence, Coverage, and Transition probability), while grip strength parameters (RMSE, CV) were computed. Finally, machine learning models were developed using EEG microstate parameters, grip strength parameters, and their multimodal fusion features to systematically evaluate classification performance in distinguishing early PD patients from healthy individuals.

resultIn the resting-state, PD patients exhibited a significant increase in the Occurrence and Coverage of microstate A, while the Coverage of microstate B significantly decreased. During the VMC task, PD patients exhibited a microstate pattern significantly different from that of HC. PD patients exhibited significantly lower temporal parameters for microstate B but higher values for microstate A and C under medium/high gain conditions, which was exactly the opposite of the performance at low gain. Moreover, PD patients exhibited poorer grip performance across all gain conditions. Finally, compared with the grip strength parameter model and the multidimensional parameter model, the microstate parameter model based on the VMC task demonstrated higher sensitivity in identifying early PD patients. At low, medium, and high visual feedback gains, the highest classification accuracy reached 100% in all cases. Notably, under high-gain conditions, all classification algorithms achieved 100% accuracy.

conclusionThis study, through EEG microstate analysis, revealed abnormal brain network dynamics in early PD patients under resting-state and various visual feedback gain conditions. The EEG microstate parameter model based on the VMC task exhibited high sensitivity in identifying early PD patients, particularly under high gain conditions. These findings suggest that EEG microstate parameters during the VMC task hold potential as objective biomarkers for early PD diagnosis and have significant clinical value.