Symplectic geometry decomposition-based features for automatic epileptic seizure detection

Abstract

In this paper, a novel method employing symplectic geometry decomposition-based features is proposed for automatic seizure detection. This study explores the performance of the suggested method in signal representation, and it also investigates the discrimination ability of the proposal. To obtain the best tradeoff between classification results and computational complexity, a selection experiment for optimizing the embedding dimension d is introduced. Subsequently, simplified eigenvalues obtained from the symplectic geometry decomposition are adopted as vectors fed into a support vector machine (SVM) to verify the effectiveness of the proposed method. In the experimental processes, a comparison is made between the decomposition capacity of symplectic geometry under various intensities of artificial noise. Further, this study analyzes the efficiency and transferability of the proposed model for multi-class tasks with great clinical significance using different electroencephalogram (EEG) datasets, including the Bonn and (Children's Hospital Boston-Massachusetts Institute of Technology) CHB-MIT datasets. In all classification tasks for the Bonn dataset, the accuracies were greater than 99.17%. The average classification accuracy (ACC) and Matthews correlation coefficient (MCC) of 99.620% and 0.918 were respectively achieved using the CHB-MIT dataset. In comparison to the state-of-art methods, the superior competence of the proposed methodology has high accuracy and low complexity as shown in the experimental results. Furthermore, the transferable ability is verified. The proposed approach is beneficial as an assistant diagnostic tool for clinicians.

Keywords: Automatic seizure detection; EEG; SVM; Symplectic geometry decomposition.