Embedded Fuzzy Classifier for Detection and Classification of Preseizure State Using Real EEG Data

• Main Authors: Qidwai, Uvais, Malik, Aamir Saeed, Shakir, Mohamed
• Other Authors: Goh, James
• Format: Book Section
• Published: Springer International Publishing 2013
• Subjects: Q Science (General); R Medicine (General); T Technology (General)
• Online Access: http://eprints.utp.edu.my/10974/1/Embedded%20Fuzzy%20Classifier%20for%20Detection%20and%20Classification%20of%20Preseizure%20State.pdf; http://dx.doi.org/10.1007/978-3-319-02913-9_105; http://eprints.utp.edu.my/10974/
• Institution: Universiti Teknologi Petronas

A Classification technique using Fuzzy Logic Inference System to identify and predict the partial seizure from the epileptic EEG data along with preliminary brain conditions in different scenarios is presented in this paper. This detection system can produce warning signals for epileptic seizures. Electroencephalography (EEG) plays an important role, especially EEG based health diagnosis of brain disorder. However, the common clinical methods are insufficient when it comes to design an automated module to detect and predict partial seizure for epileptic patients. If the detection system is to be designed for ubiquitous applications, the system becomes even more complex if the patient is not confined to clinical environment when the device is monitoring continuously while the patient is involved in daily activities. Therefore, the work presented here includes embedded hardware system that works with classification algorithm on real EEG signals, in a ubiquitous setting. The performance of the system is shown under various conditions of daily activities. In order to make all this in a ubiquitous form factor, the algorithm for classification and detection of the pre-seizure conditions should be tremendously simple for processing the signal in a low cost ubiquitous microcontroller. This has been achieved in this work through the use of Fuzzy Classifiers based on the lookup table to empower system simplicity. The algorithm also utilizes certain statistical features from the EEG signal that are used as features to the classifier logic. While the clinical testing of the device is still awaited, various scenarios have been implemented using a custom-built hardware simulator based on empirical modeling of the real EEG signals. This shown various performance modes of the system and confirms the detection of pre-seizure state for a number of parameters related to the patients such as age, gender, etc... By using this type of fuzzy logic classifier, we were able to get over 90% accurate classifications for the partial seizure.