AbstractThe objective of this study was to develop an adaptive framework for seizure detection in real-time that is practical to use in the Epilepsy Monitoring Unit (EMU) as a warning signal, and whose output helps characterize epileptiform activity. Our framework uses a one-class Support Vector Machine (SVM) that is being trained dynamically according to past activity in all available channels. This is done to evaluate the novelty of the current instance according to previous activity. Our algorithm was tested on intracranial EEG from human epilepsy patients that are admitted to the EMU for presurgical evaluation. In this study, we compared multiple configurations for using a one-class SVM to assess if there is significance over specific neural features or electrode locations. Our results show our algorithm is capable of running in real-time and achieving a high performance for early seizure-onset detection with a low false-positive rate and robustness to different types of seizure-onset patterns as well as to the number of channels used. This algorithm offers a solution to warning systems in the EMU as well as a tool for seizure characterization during post-hoc analysis of intracranial EEG data for surgical resection of the epileptogenic network.HighlightsThis study proposes a dynamic training algorithm that efficiently detects sudden novel changes in intracranial electroencephalographic activity, creating a reliable seizure onset detection algorithm that does not need prior training.The algorithm described has the capability to be implemented in real-time, independently of the number of channels that are being analyzed.The presented detector shows high performance and reliability to be easily implemented in the Epilepsy Monitoring Unit to quickly alert clinical staff of seizure events.
An algorithm for seizure onset detection using intracranial EEG
