A methodology for the automatic identification and classification of EEG waves based on clinical guidelines

In order to identify abnormal behaviors related to epileptic seizures and other neurological disorders, in this paper, it is described a methodology with a clinical approach to classify events shown in EEG recordings, based on the international standards and current guidelines. According to medical definitions, for this work the EEG signal was classified into suppressions, rhythms, frequencies and abnormal behaviors, obtained from several combinations of the signal's parameters: amplitude, frequency and patient's age. The mean of peak values was used to obtain the amplitude. The Fast Fourier Transform and high-pass filters were used to extract the dominant frequency. A set of 192 segments of one second duration, randomly selected from 7 patients, were evaluated and their behavior was identified and classified into 13 classes. It was obtained an efficiency of 96.35%. This method found abnormal behaviors related to epileptic seizures and other kinds of neurological disorders. This method identifies abnormal events, their timing and cortical distribution, making possible the extraction of seizure's segments for further feature analysis.