Integrated Automatic Detection, Classification and Imaging of High Frequency Oscillations With Stereoelectroencephalography.

摘要: Objective During presurgical evaluation for focal epilepsy patients, the evidence supporting use of high frequency oscillations (HFOs) delineating epileptogenic zone (EZ) increased over past decade. This study aims to develop and validate an integrated automatic detection, classification imaging pipeline HFOs with stereoelectroencephalography (SEEG) narrow gap between quantitative analysis clinical application. Methods The proposed includes stages channel inclusion, candidate detection labeling four trained convolutional neural network (CNN) classifiers sorting based on occurrence rate imaging. We first evaluated initial detector using open simulated dataset. After that, we validated our full algorithm in a 20-patient cohort against three assumptions previous studies. Classified results were compared seizure onset (SOZ) channels their concordance. receiver operating characteristic (ROC) curve corresponding area under (AUC) calculated representing prediction ability labeled outputs SOZ. Results demonstrated satisfactory performance CNN converged quickly during training, accuracies validation dataset above 95%. localization value was significantly improved by classification. AUC values 20 testing patients after HFO classification, indicating EZ identification. Conclusion Our can provide robust revealing at individual level, which may ultimately push forward transitioning into meaningful part surgical planning.