Feature-based detection of epileptic seizures using ECoG recordings and a PAM model

This paper develops a detection scheme for identifying seizure onset using single channel electrocorticograms recorded from intractable epilepsy patients. A statistic which measures changes in the peaks of the autocovariance function of a signal is developed based upon a pulse amplitude modulation model. This method is useful both for detecting seizure onset and for localizing the seizure focus within the cortical tissue by comparing data collected from multiple electrodes.