Machine learning to detect intraoperative ischemia from electroencephalography in carotid endarterectomy surgery

The article addresses the use of machine learning for automated detection of cerebral ischemia during carotid endarterectomy (CEA), a high-risk surgical procedure. The research team tested five different machine learning models, including random forest, XGBoost, logistic regression, support vector classifier, and naive Bayes, to analyze quantitative electroencephalographic features. The random forest model achieved the highest sensitivity of 0.79-0.83, while XGBoost achieved the highest specificity of 0.93-0.96. Feature importance analysis revealed that alpha-band activity and hemispheric asymmetry are the most important indicators of ischemia. The results suggest that machine learning assistance could support neurophysiologists and enhance patient safety during high-risk surgery.