F154. Machine learning for the analysis of single pulse stimulation in electrocorticography

Abstract

were localized to-hippocampus, amygdala hippocampus, hippocampus-temporal pole, ant. Cingulum, superior temporal gyrus, and frontal operculum. With direct CS, 7 (54%) had typical seizures (aura followed by seizures with impaired awareness), 4 (31%) had atypical seizures (aura and clonic activity), and 2 (15%) had both types. Localization of CS-induced seizures overlapped with spontaneous seizures in 9/13 (70%) patients. Overlapping was 100% with mesial temporal structures. 7 (54%) patients underwent surgery, and 3 had RNS placement. Conclusion: Direct CS induced seizures complemented localization of spontaneous seizures in two-thirds of the patients who underwent stereo EEG investigation. The accuracy of direct CS in the localization of EZ is higher with infrasylvian structures.

Computational modelling and electrical stimulation for epilepsy surgery

“…Machine learning for DRs can be applied in the time-frequency or time domain. Using the same time-frequency images as used for visual examination, two different machine learning methods showed promising results, especially for spikes and riples [164]. Compared to the inter-observer agreement between two humans, the agreement between human and machine was similar for both methods in the spike and ripple bands [165].…”