Towards computer supported search for semiological features in epilepsy seizure classification

Abstract

Seizure semiology has always been an important part of seizure classification. Value of the most common ictal signs for localization and lateralization of a seizure focus, as well as their sensitivity and specificity for certain focal epilepsies, is well known. All over it, there still remain many signs and poorly described patient behaviours during a seizure whose relation to a seizure focus have yet to be specified and confirmed. Some new signs have been introduced recently but all of them have been based on data provided from just a few dozens of patients. This is no surprise since checking for presence of a specific ictal sign in a patient requires lengthy manual review of video records documenting his/her seizures. We suggest a novel approach toward identification/verification of new ictal signs based on computer supported systematic review of unique extensive dataset of Na Homolce Hospital containing approximately 1.000 seizures (representing data of 400 patients with up to 5 seizures annotated). This requires transforming the original set of patient records into a database consisting of annotated ictal video-EEG recordings in a structured form suitable for statistical analysis as well as for analysis of sequence patterns. This contribution describes our SW tool ASTEP designed and developed for this purpose and demonstrates some properties of ASTEP database, namely advantages of the used seizure description as a sequence of considered ictal signs complemented by detailed information on timing, duration, repetition and mode of appearance of these signs. Finally, some preliminary results are reported.