EEG spectral biomarkers of postoperative delirium in spinal surgery: A high-resolution analysis

Abstract

Background Postoperative delirium is common in older adults following spinal surgery, yet current EEG-based detection methods rely on broad conventional frequency bands that may obscure clinically meaningful oscillatory changes. More granular spectral approaches may reveal candidate frequency-specific features that could improve diagnostic performance. Methods We recorded single-electrode frontal EEG (Fp1) from 47 patients at four perioperative timepoints and compared traditional band-level power with high-resolution 1-Hz spectral analysis (1–45 Hz). Group differences were evaluated using non-parametric statistics with FDR correction, and diagnostic accuracy was assessed using AUROC and threshold-optimized metrics. All diagnostics thresholds were derived and evaluated in the same dataset. Results Seven patients (14.9%) developed delirium. Conventional band analysis detected only reduced theta power during active delirium (|r| = 0.31; AUROC = 0.53). In contrast, 1-Hz analysis revealed a richer pattern: elevated 1-Hz power (|r| = 0.27), reduced 3–7 Hz power (|r| = 0.25–0.27), and increased 14–15 Hz power (|r| = 0.22–0.26). The 23-Hz narrow-band feature achieved the best discrimination (AUROC = 0.74), despite showing no significant effect within the traditional beta band (12–30 Hz). Several frequencies remained altered even after clinical resolution of delirium. Conclusions In this exploratory analysis, high-resolution spectral analysis identified candidate frequency-specific EEG features that distinguished delirium cases more accurately than conventional frequency bands in this cohort. Broadband averaging may not detect frequency-specific oscillatory patterns; adopting finer spectral resolution could enhance the utility of single-channel EEG for routine clinical monitoring.