Newborn EEG seizure detection using signal structural complexity

A method for the automatic detection of seizure in newborns is presented. The proposed method is derived from the ability to detect changes in signal structure as the newborn EEG changes from the background state to the seizure state. Matching Pursuit decomposition technique, with an overcomplete time-frequency dictionary, is shown to be an adequate technique for detecting changes in signal structure. Changes are detected by using a new signal measure referred to as structural complexity, which is directly related to the dictionary being used for decomposition. The structural complexity measured is then incorporated in the proposed automatic newborn seizure detection algorithm.