Identification of Parkinson's Disease Subtypes from Resting-State Electroencephalography

Abstract

Background: Parkinson's disease (PD) patients present with a heterogeneous clinical phenotype, including motor, cognitive, sleep, and affective disruptions. However, this heterogeneity is often either ignored or assessed using only clinical assessments. Objectives: We aimed to identify different PD sub-phenotypes in a longitudinal follow-up analysis and their electrophysiological profile based on resting-state electroencephalography (RS-EEG) and to assess their clinical significance over the course of the disease. Methods: Using electrophysiological features obtained from RS-EEG recordings and data-driven methods (similarity network fusion and source-space spectral analysis), we have performed a clustering analysis to identify disease sub-phenotypes and we examined whether their different patterns of disruption are predictive of disease outcome. Results: We showed that PD patients (n = 44) can be sub-grouped into three phenotypes with distinct electrophysiological profiles. These clusters are characterized by different levels of disruptions in the somatomotor network (Δ and β band), the frontotemporal network (α2 band) and the default mode network (α1 band), which consistently correlate with clinical profiles and disease courses. These clusters are classified into either moderate (only-motor) or mild-to-severe (diffuse) disease. We showed that EEG features can predict cognitive evolution of PD patients from baseline, when the cognitive clinical scores were overlapped. Conclusions: The identification of novel PD subtypes based on electrical brain activity signatures may provide a more accurate prognosis in individual patients in clinical practice and help to stratify subgroups in clinical trials. Innovative profiling in PD can also support new therapeutic strategies that are brain-based and designed to modulate brain activity disruption.