Classification of healthy and epileptic seizure EEG signals based on different visibility graph algorithms and EEG time series

Recently, the idea of processing time series by transforming them onto graphs has been used in many studies. One of the simple methods proposed to convert a time series onto a graph is the visibility graph (VG). The current study investigates the ability of different VG algorithms for epileptic seizure detection. In the algorithm, single-channel Electroencephalogram (EEG) signals are transformed onto five different VG graphs, and then 13 features are generated from obtained graphs. After that, efficient features are extracted using the Sequential forward feature selection (SFFS) algorithm and classified by Random Forest (RF) into two or three classes. The experimental results show that VG algorithms are fast and easy on the performance of classification. In addition, it has shown that the proposed method not only is able to discriminate two classes with 100% accuracy, but also recognizes three classes with high accuracy, sensitivity, and specificity of 97.98%, 96.19%, and 99.12%, respectively. The comparison of this study with other methods shows the effectiveness of the proposed method for seizure detection.

Automated summary

Recently, the concept of transforming time series into graphs has been widely used in various studies. This particular study focuses on the visibility graph (VG) algorithms for epileptic seizure detection. Single-channel EEG signals are transformed into five different VG graphs, and 13 features are extracted. The results show that the proposed VG algorithms are efficient for classification, with high accuracy for two or three classes, making it effective for seizure detection.