Unsupervised Classification of Epileptic EEG Signals with Multi Scale K-Means Algorithm

Abstract

Most epileptic EEG classification algorithms are supervised and require large training data sets, which hinders its use in real time applications. This paper proposes an unsupervised multi-scale K-means (MSK-means) algorithm to distinguish epileptic EEG signals from normal EEGs. The random initialization of the K-means algorithm can lead to wrong clusters. Based on the characteristics of EEGs, the MSK-means algorithm initializes the coarse-scale centroid of a cluster with a suitable scale factor. In this paper, the MSK-means algorithm is proved theoretically being superior to the K-means algorithm on efficiency. In addition, three classifiers: the K-means, MSK-means and support vector machine (SVM), are used to discriminate epileptic EEGs from normal EEGs using six features extracted by the sample entropy technique. The experimental results demonstrate that the MSK-means algorithm achieves 7% higher accuracy with 88% less execution time than that of K-means, and 6% higher accuracy with 97% less execution time than that of the SVM.