Data-Driven Peak Alpha Identification Reveals Distinct EEG Spatial Signature Phenotypes in Chronic Pain

Chronic pain affects over 20% of US adults, making it a widely prevalent disorder characterized by a highly heterogeneous patient experience and lack of objective biomarkers for characterization. Recent studies have suggested that cortical oscillations captured by electroencephalography (EEG), specifically in the alpha frequency band, may encode information relevant to chronic pain. In this study, we investigate two recent data-driven approaches for extracting the peak alpha frequency (PAF) from EEG data. We use these PAF values, specifically their spatial variation across the whole cortex, to perform unsupervised stratification of 93 chronic pain patients and show that the resulting clusters are potentially clinically meaningful, each with a distinct PAF spatial signature. We also show that both approaches define patient outcome phenotypes that link these spatial signatures to patient-reported measures of pain interference. This work brings us one step closer to objective, data-driven metrics for chronic pain using EEG to enable personalized disease management.