Analysis and definition of morphological descriptors for automatic detection of epileptiform events in EEG signals with artificial neural networks

This study proposes to analyze morphological characteristics of electroencephalogram (EEG) signals in order to define a representation of epileptiform events that can distinguish them from other events occurring in the signal. Despite the existence of several studies on parameterization of EEG signals, particularly for automatic detection of paroxysms related to epilepsy, it was necessary to create a new set of parameters that reveal specific morphological characteristics pertaining to these events, since during the automatic detection process they may get mixed up if only conventional descriptors are used. The proposed parameters are fed to artificial neural networks and the individual and collective contribution of each parameter was evaluated by statistical process. The proposed method achieved automatic detection with a 90% success rate and sensitivity and specificity between 90% and 95%.