Spatiotemporal Dynamics of Mild Cognitive Impairment in Parkinson’s Disease: A Pilot TMS-EEG Study

Abstract

Early identification of mild cognitive impairment in Parkinson’s disease (PD-MCI) is crucial for delaying dementia progression, yet the mechanisms underlying cortical excitability and time-varying network dysconnectivity remain elusive. This study utilized transcranial magnetic stimulation combined with electroencephalography (TMS-EEG) to characterize time-varying directed brain network alterations targeting the right posterior parietal cortex (PPC) in PD-MCI. 20 PD patients were categorized into PD-MCI and cognitively normal (PD-NC) groups using the Montreal Cognitive Assessment (MoCA). Adaptive directed transfer function (ADTF) was applied to assess whole-brain directed time-varying functional connectivity following right PPC stimulation. Machine learning models were then employed to classify PD-MCI using spatiotemporal network features. Compared to PD-NC, the PD-MCI group exhibited significantly reduced cross-regional directed connectivity across the full 1–45 Hz frequency band. Notably, this decoupling was most pronounced in the γ band, with widespread disruptions across multiple time windows. Abnormal connectivity strengths were significantly positively correlated with MoCA cognitive scores. A KNN classification model was constructed based on 5 key spatiotemporal features, and using 5 repetitions of 5-fold cross-validation, ultimately achieving an optimal classification accuracy of 93.33%. These findings reveal that PPC-targeted, full-frequency time-varying network decoupling is a core neuropathological mechanism in PD-MCI. The identified spatiotemporal features hold promise as objective biomarkers for the early identification of PD-MCI, providing a foundation for early diagnosis and targeted neuromodulation.