Research on Epilepsy Classification Model Based on Variational Mode Quadratic Decomposition

Epilepsy, a widespread neurological disorder, creates substantial physical and psychological challenges for patients. Accurate seizure prediction allows for prompt symptom intervention and treatment guidance. To achieve this goal, we introduce a classification model for epilepsy using variational mode quadratic decomposition. It first processes EEG signals from epilepsy patients with variational mode decomposition. Then, a second variational mode decomposition filters and breaks down the residual signal further. After decomposition, signals are reconstructed into continuous wavelet transform feature images. A combination of convolutional neural network and temporal convolutional network then classifies epileptic seizure periods, including ictal, preictal, interictal, and non-ictal phases. The extensive experimental results show the model reaches 85% accuracy and 88.8% precision in classifying epileptic seizure, pre-ictal, inter-ictal, and non-ictal periods, evidencing the proposed method's effectiveness.