Application of EEG and EMG Signals in Multi-Classification of Epilepsy

The classification of epileptic seizures is crucial for the treatment of epilepsy, and currently, this task primarily relies on clinical doctors, which is complex and time-consuming. Our study aims to perform an eight-category classification of epileptic seizures using multi-domain features from electroencephalogram (EEG) and electromyogram (EMG) signals, combined with deep learning algorithms. In this research, EEG signals are decomposed via the Variational Mode Decomposition (VMD) method and features are extracted in the time domain, frequency domain, and from information theory, while EMG signals are processed to extract relevant features directly. Following feature fusion, the Resnet model is employed for classification. Under the evaluation criteria based on seizure events and across subjects, the accuracy of the eight-category classification of seizures is 75%, with a weighted F1 score of 0.75, where the accuracy and recall for Tonic-Clonic Seizures are 100%. Moreover, in this task, we also compared the classification effects using only EEG signals and using both EEG and EMG signals, as well as the performance differences between machine learning models and deep learning models. The results indicate that the classification performance is optimized when both EEG and EMG signals are used in conjunction with a neural network model.