A novel approach for classification of epileptic seizures using matrix determinant

Objective: An epileptic seizure is recognized as a neurological disorder caused by transient and unexpected disturbance resulting from the excessive synchronous activity of the neurons in the brain. Analysis of epileptic seizures derived from long-term recordings of electroencephalogram (EEG) is cumbersome and time consuming for a neurologist. Therefore, introducing an automated detection system surrogate the neurologist involvement all time and speed up the treatment procedure. This study introduces a matrix determinant of EEG as a significant feature for recognition of epileptic seizures. Initially, artifact-free filtered EEG time series was arranged sequentially to form a square matrix of order, namely 13, 16, 23, and 32 and determinant was estimated. Assumed that the total elements in the square matrix represent a typical segmentation length. The experiment was conducted using EEG database obtained from the University of Bonn and Ramaiah Medical College and Hospital (RMCH). In total, eleven classification problems among non-epileptic group and epileptic EEG were composed to examine temporal dynamics of brain activity in different states of the epileptic activity. Next, the extracted feature was classified using support vector machine (SVM), K-nearest neighbor (K-NN), multi-layer perceptron (MLP) classifiers with 10-fold cross-validation. Experimental results revealed the highest classification accuracy of 99.45% (using University of Bonn) and 97.56% (using RMCH). between normal and epileptic EEG. In addition, other classification problems and matrix orders showed better results using all the classifiers. Further, descriptive analysis, histogram plot in polar coordinates and the bivariate histogram analysis was performed. In conclusion, matrix determinant found to be a potential biomarker for the real-time detection of epileptic seizure with minimal computational complexity.