Automated EEG-based prediction of delayed cerebral ischemia after subarachnoid hemorrhage. (November 2022)

• Record Type: Journal Article
• Authors: Zheng, Wei-Long; Kim, Jennifer A.; Elmer, Jonathan; Zafar, Sahar F.; Ghanta, Manohar; Moura Junior, Valdery; Patel, Aman; Rosenthal, Eric; Brandon Westover, M.
• Is Part Of: Clinical neurophysiology. Volume 143(2022)
• Page Start: 97
• Page End: 106
• Publication Date: 2022-11
• Subjects: EEG -- Subarachnoid hemorrhage -- Delayed cerebral ischemia -- Machine learning -- Epileptiform discharges -- Biomarkers; Neurophysiology -- Periodicals; Electroencephalography -- Periodicals; Electromyography -- Periodicals; Neurology -- Periodicals; 612.8
• DOI: 10.1016/j.clinph.2022.08.023
• Languages: English
• ISSNs: 1388-2457

Abstract

Highlights: A number of EEG features have been shown to be predictive of delayed cerebral ischemia (DCI), but a continuous assessment of multidimensional EEG features is lacking. Combining spectral and epileptiform discharge feature information, using automated calculations, allows for dynamic prediction of DCI. This dynamic multi-feature assessment increases the feasibility of implementing interventions in response to our EEG derived DCI risk probability. Abstract: Objective: Delayed cerebral ischemia (DCI) is a leading complication of aneurysmal subarachnoid hemorrhage (SAH) and electroencephalography (EEG) is increasingly used to evaluate DCI risk. Our goal is to develop an automated DCI prediction algorithm integrating multiple EEG features over time. Methods: We assess 113 moderate to severe grade SAH patients to develop a machine learning model that predicts DCI risk using multiple EEG features. Results: Multiple EEG features discriminate between DCI and non-DCI patients when aligned either to SAH time or to DCI onset. DCI and non-DCI patients have significant differences in alpha-delta ratio (0.08 vs 0.05, p < 0.05) and percent alpha variability (0.06 vs 0.04, p < 0.05), Shannon entropy (p < 0.05) and epileptiform discharge burden (205 vs 91 discharges per hour, p < 0.05) based on whole brain and vascular territory averaging. Our model improves predictions by emphasizing the most informative features at a given time with an area under the receiver-operator curve of 0.73, by day 5 after SAH and good calibration between 48–72 hours (calibration error 0.13). Conclusions: Our proposed model obtains good performance in DCI prediction. Significance: We leverage machine learning to enable rapid, automated, multi-featured EEG assessment and has the potential to increase the utility of EEG for DCI prediction.