An automated system for analysis and interpretation of epileptiform activity in the EEG

Abstract

Electroencephalography is an important clinical tool for the evaluation and treatment of neurophysiological disorders related to epilepsy. However, the analysis and interpretation of the electroencephalogram (EEG) is not an easy task due to the variety of waveforms that are possible. Consequently, EEG analysis is in need of an objective and quantitative methodology. In this paper, an automated system for diagnosing epilepsy is presented. The system combines both the electrocerebral activity related to epilepsy resulting from EEG and other neurophysiological expertise, mainly based on clinical symptoms that occur during the patient's clinical attack, to avoid misdiagnosis. The system consists of two major stages. The first is a feature extractor in which half-waves are detected and artefacts are eliminated. The second and most important stage is a knowledge-based system for recognising and classifying epileptiform events. In particular, the analysis is based on the detected EEG patterns representing epileptiform activity, localization information of discharge focus and clinical symptoms. Once a diagnosis is established, the system also proposes a therapy. The proposed system has been tested using many different clinical cases, and the obtained experimental results are acceptable.