Fuzzification of epileptic data: an application for prediction and identification of partial seizure

Objectives: Electroencephalography (EEG) plays an important role, especially EEG based health diagnosis of brain disorder. However, these approaches fall short when attempting to design an automated system to detect and predict partial seizure for epileptic patients. The situation becomes even more difficult when the detection system is being designed for a ubiquitous application in which the patient is not confined to the hospital and the device is attached to him/her externally while the person is involved in daily chores. This paper presents a classification technique by using Fuzzy Logic System to identify and predict the partial seizure from epileptic data. The presented work covers the initial findings related to some of the brain conditions in different scenarios so that the detection system can produce warning signals for epileptic seizure.
Method: Due to the compact nature of ubiquitous systems, the detection and classification techniques have to be extremely simple work in real-time. The paper presents one such technique which is based on fuzzy classifications of the EEG data using certain statistical features from the signal. This will help in developing a more generalizable solution as a low cost wearable EEG monitoring of partial seizure in the future. A significant advantage of this system is that all the filtering and pre-processing is done by the main sensory unit, the Emotiv EEG headset. By using fuzzy logic technique, the membership functions are calculated in each epoch; where the maximum degree of membership is scored and then classified. Hence in this proposed system, we introduce fuzzification of epileptic EEG data using fuzzy logic interface to classify partial seizure EEG signals from the normal EEG.
Results: By using this type of fuzzy logic classifier, we were able to get the 93% accurate classification for the partial seizure. The algorithm was implemented in the ubiquitous manner. The microcontroller and computer environment could perform all the processing including filtering, fuzzification and classification based on the look-up tables.
Conclusion: In this paper, an innovative strategy is presented to perform computationally low cost classification for the EEG signals. This technique can be used in real-time classification from the EEG signal as it is measured. Hence the system can be used as classifier as well as a predictor for certain epileptic disorder conditions and can be enhanced to clinical applications in the future.