Robust expert system design for automated detection of epileptic seizures using SVM classifier

Abstract

The classification of normal and ailing brain activities through visual inspection proves to be very challenging even for any experienced neurologist. The case is even worse for detection of heterogeneous anomalies like epileptic seizures. Authors have presented robust expert system design for classification of epileptic seizures automatically with an improvement over the existing systems. The developed scheme illustrates selection methodology for feeding energy, entropy and standard deviation feature sets to the support vector classifier. The results display maximum classification rate of 99.53 % with sensitivity and specificity rates above 98.8 %. These results were validated over 10 folds of sub-divisions using rotation estimation technique with minimum computation time noted to be 0.0131 s. Therefore, the expert system developed during this study holds promising grounds for automated clinical diagnosis in real time.