The History of Seizure Prediction

摘要

Seizure-prediction methods, which are based on EEG recordings only, could bean integral part of future therapeutical devices in epilepsy. The current focus ofresearch is on a seizure-control system which can suppress an upcoming seizure(closed-loop intervention), or at least warn the patient prior to an expected event.However, this goal seems to be very far off but important progress has beenmade in neurology, biology, network modeling, microsystems engineering, andtime series analysis. The state of the art in seizure prediction is that, to date, aclinically applicable seizure-prediction algorithm does not exist, but several studieshave shown that one can in fact predict seizures significantly better than a randompredictor. Much too optimistic results obtained in the last two decades have beenshown not to be specific for seizure precursors, resulting in frequent false alarmsof the methods. A patient will certainly not accept a seizure-control system whichproduces too many false alarms, because it will hamper the patient instead ofhelping him. This review chapter presents the history, the current state of theart, and also an outlook on future developments in seizure prediction; a researchfield which has gained much attention in the last thirty years. We discuss thepossible existence of a preictal phase, review several-prediction methods, elaborateon statistical evaluation, address open problems and caveats, and discuss the designof seizure-prediction studies, referring to the chapters of this book contributed byspecialists in the field.