UCLA

BACKGROUND: Clinical and neuroimaging measures incompletely explain behavioral deficits in the acute stroke setting. We hypothesized that electroencephalography (EEG)-based measures of neural function would significantly improve prediction of acute stroke deficits. METHODS: Patients with acute stroke (n=50) seen in the emergency department of a university hospital from 2017 to 2018 underwent standard evaluation followed by a 3-minute recording of EEG at rest using a wireless, 17-electrode, dry-lead system. Artifacts in EEG recordings were removed offline and then spectral power was calculated for each lead pair. A primary EEG metric was DTABR, which is calculated as a ratio of spectral power: [(Delta*Theta)/(Alpha*Beta)]. Bivariate analyses and least absolute shrinkage and selection operator (LASSO) regression identified clinical and neuroimaging measures that best predicted initial National Institutes of Health Stroke Scale (NIHSS) score. Multivariable linear regression was then performed before versus after adding EEG findings to these measures, using initial NIHSS score as the dependent measure. RESULTS: Age, diabetes status, and infarct volume were the best predictors of initial NIHSS score in bivariate analyses, confirmed using LASSO regression. Combined in a multivariate model, these 3 explained initial NIHSS score (adjusted r2=0.47). Adding any of several different EEG measures to this clinical model significantly improved prediction; the greatest amount of additional variance was explained by adding contralesional DTABR (adjusted r2=0.60, P<0.001). CONCLUSIONS: EEG measures of neural function significantly add to clinical and neuroimaging for explaining initial NIHSS score in the acute stroke emergency department setting. A dry-lead EEG system can be rapidly and easily implemented. EEG contains information that may be useful early after stroke.