Aperiodic Correction of Posterior Theta/Alpha Ratio Reduces Arousal-State Dependency in Routine EEG-Based Dementia Screening

Abstract

Introduction

Posterior theta/alpha ratio (TAR) is among the best-validated qEEG biomarkers for dementia, yet remains largely unused in routine neurology practice — primarily because it is sensitive to the patient's arousal state, an uncontrolled variable in real-world clinical EEG. We evaluated whether aperiodic spectral correction can bridge this gap between research promise and clinical reality.

Methods

Routine EEGs from 73 participants (18 controls, 55 dementia patients) were analyzed across three arousal segments. Five correction strategies were evaluated for TAR, with the optimal method validated in an external dementia dataset (ds004504, n = 88) and a normative cohort (ds005385, n = 587, age 20-79).

Results

Raw TAR was diagnostically significant at Alert (AUC=0.772) and HVT (AUC=0.812) segments but failed in the Drowsy segment (AUC=0.631, p = 0.110) — the state most representative of uncontrolled clinical EEG. FOOOF-based aperiodic correction restored significance (AUC=0.701, p = 0.014), improved intraclass correlation to 0.945, and generalised to the external dataset (AUC=0.830). In the normative cohort, FOOOF-corrected TAR was completely age-independent (r = 0.000), supporting a universal threshold of ≥0.45 (specificity 92.3%).

Conclusions

A single step of aperiodic correction transforms TAR from an arousal-sensitive research metric into a robust, age-independent clinical tool. Combined with cognitive assessment, FOOOF-corrected TAR offers practical, objective dementia screening within routine EEG practice.