Epilepsy is a brain disorder characterized by epileptic seizures that affects over 50 million people worldwide. Epileptic seizures can affect the motor, sensory and autonomic capacity, the state of consciousness, emotional or even the behavior of patients. Electroencephalogram (EEG) is one of the most widely used diagnostic tests for epilepsy investigation. Generally, epileptic seizures identification is done by a trained neurophysiologist. However, automatic techniques for identifying seizures in EEG signals can be faster and less sensitive to failure. Recently, an algorithm has been proposed by Campanharo et al to map a time series in a complex network. Therefore, the objective of this work is the computational implementation of this map, using the concept of bins, to the automatic identification of epileptic seizures in EEG signals.
Here, we used an artifact-free extracranial EEG database that has been provided by Boston Children`s Hospital. In total, it were analyzed eighteen EEG signals of eight epileptic patients under antiepileptic drugs. In order to analyze the topology of the generated networks, topological measures were computed, such as, bipartivity, mean jump length, estrada index and reciprocity. Subsequently, linear discriminant analysis (LDA) technique was applied to discriminate networks associated with seizures from networks not associated with seizures. To evaluate the performance of the proposed methodology, the statistical concepts of sensitivity, specificity and accuracy were used. In conclusion, the proposed methodology performed satisfactory results in identifying epileptic seizures.