Epileptic seizure detection from ECoG signals acquired with experimental epilepsy

In recent years, with the development of computer technologies and automatic diagnosis methods, the required time for diagnosis and the number of incorrect diagnoses have been decreased. These developments will also be helpful in diagnosis of epilepsy caused by short term brain function disorder and finding the source of this in a short time. In this study, automatic classification of the ECoG signals into normal, interictal and ictal periods which are occurred during the epileptiform activity was performed for the diagnosis of epilepsy. For the feature extraction, discrete Fourier transform and Hjorth descriptors, for the classification k-nearest neighbor (k-NN) and multilayer artificial neural networks (MLANN) were used. The results of the experiments show that the proposed feature extraction method has a superior performance with the use of MLANN classification. The obtained recognition rates were 97.45% and 99.75% for the selection of training-test sets from different and same channels, respectively.