چكيده به لاتين
Abstract:
Epilepsy is the one of the most serious neurological disorders. About 1% of people around the world are suffering from epilepsy. Therefore, investigation on epilepsy has drawn the attention of researchers recently. Detection of a seizure attack is traditionally carried out by viewing brains signals. It is often difficult and time consuming. Thus, providing automatic seizure detection system is important. In this study, we have tried to be provided an automatic seizure detection method using brain signals.
In this research, Electrocorticogram (ECoG) and intracortical signals of the rats are used to seizures detection. ECoG signals were recorded from 40 rats and 32-channels. In this project, Time domain characteristics, frequency domain characteristics and nonlinear characteristics have been used to detect seizures. One of the important issues of automatic diagnosis of seizures is selection appropriate feature and Suitable channels. In this project, method based on mutual information for feature selection and channel selection is provided. The accuracy of the the combination of different channels is calculated and the result is that the combination of 5 channels provides a high accuracy. 10 most significant features among 84 features and 5 important channel between 32 channels using mutual information were selected. The results showed 99/44±0.17 accuracy.
Intracranial signals were recorded from 20 rats during move freely. One of the most important characteristics of the intracranial signals is the manifestation of the spikes. In this study, spikes activity in normal state, ictal and interictal is studied using Spike density function (SDF). The results show that spike rate increased during the seizure.
In this study, seizure detection algorithm based on mutual information is applied to an epileptic EEG dataset provided by Dr. R. Andrzejak. That in compared with previous works is achieved 100% accuracy.
Keywords: Electrocorticographic (ECoG), Intracortical signal, feature selection, channel selection, spike.