ABSTRACT
Automated detection of epileptic seizures from EEG records is an old problem, and work continues because detectors are driven by technology. However to date little or no standardization exists in what ‘good’ performance or ‘good’ detection means. To knowledge no comprehensive review of seizure detection algorithms where state of the art detection strategies are compared on the same dataset has been published. This chapter explores the difficulties of seizure detection as well as to provide a norm under which current and future seizure detectors can be validated. Whilst there are a few independent studies that evaluate the performance of algorithms, to date no mass testing of seizure detection algorithms on the same dataset exists. There are many reasons why neurophysiologists need seizures to be detected from EEG. Seizure detection is useful for diagnosis: long hours of EEG must be reviewed and classified.
