Title

Classification of focal and non focal EEG signals using empirical mode decomposition (EMD), phase space reconstruction (PSR) and neural networks

Document Type

Article

Date of Original Version

1-1-2019

Abstract

Electroencephalogram (EEG) signals can be used to identify the human brain in different disease conditions. Nonetheless, it is difficult to detect the subtle and vital differences in EEG simply by visual inspection because of the non-stationary nature of EEG signals. Specifically, in order to find the epileptogenic focus for medical treatment in the case of a partial epilepsy, an intelligent system that can accurately and automatically detect and discriminate focal and non focal groups of EEG signals is required. This will assist clinicians in locating epileptogenic foci before surgery. In this study we propose a novel method for classification between focal and non focal EEG signals based upon empirical mode decomposition (EMD), phase space reconstruction (PSR) and neural networks. First, EEG signals are decomposed into Intrinsic Mode Functions (IMFs) using EMD, and the third and fourth IMFs components are extracted which contain most of the EEG signals’ energy and are considered to be the predominant IMFs. Second, phase space of the two IMFs componets is reconstructed, in which the properties associated with the EEG system dynamics are preserved. Three-dimensional (3D) PSR together with Euclidean distance has been utilized to derive features, which demonstrate significant difference in EEG system dynamics between the focal and non focal groups of EEG signals. Third, neural networks are then used as the classifier with feature vectors as the input to distinguish between focal and non focal EEG signals based on the difference of system dynamics between the two groups. Finally, experiments are carried out on the Bern Barcelona database to assess the effectiveness of the proposed method. By using the 10-fold cross-validation style, the achieved accuracy on the 50 pairs and 3750 pairs of EEG signals is reported to be 96 % and 95.37 % , respectively. Compared with other state-of-the-art methods, the results demonstrate superior performance and the proposed method can serve as a potential candidate for the automatic detection of focal EEG signals in the clinical application.

Publication Title

Artificial Intelligence Review

Share

COinS