Beyond homogeneity: Charting the landscape of heterogeneity in psychiatric electroencephalography

Abstract

Electroencephalography (EEG) has been thoroughly studied for decades in psychiatry research. Yet its integration into clinical practice as a diagnostic/prognostic tool remains unachieved. We hypothesize that a key reason is the underlying patient's heterogeneity, overlooked in psychiatric EEG research relying on a case-control approach. We combine HD-EEG with normative modeling to quantify this heterogeneity using two well-established and extensively investigated EEG characteristics -spectral power and functional connectivity- across a cohort of 1674 patients with attention-deficit/hyperactivity disorder, autism spectrum disorder, learning disorder, or anxiety, and 560 matched controls. Normative models showed that deviations from population norms among patients were highly heterogeneous and frequency-dependent. Deviation spatial overlap across patients did not exceed 40% and 24% for spectral and connectivity, respectively. Considering individual deviations in patients has significantly enhanced comparative analysis, and the identification of patient-specific markers has demonstrated a correlation with clinical assessments, representing a crucial step towards attaining precision psychiatry through EEG.

Resting-State Electroencephalography Alpha Dynamic Connectivity: Quantifying Brain Network State Evolution in Individuals with Psychosis

“…group-level clustering. Combining EEG with normative modeling is highly promising for developing patient-specific approaches 45 .…”

Characterizing the heterogeneity of neurodegenerative diseases through EEG normative modeling

“…Notably, (Lefebvre et al 2018) used normative models to investigate variability among autistic patients as compared to healthy controls. More recently, we charted the trajectory of brain development in a population aged between 5 and 18 years and mapped the heterogeneity of psychiatric diseases as reflected in spectral power density spectra computed at scalp electrodes from resting-state eyes-closed EEG activity and cortical source functional connectivity estimated from that activity (Ebadi et al 2024). In this study, we utilized normative models and EEG data from 14 datasets to achieve two primary objectives.…”