Epilepsy seizure prediction based on ViT-2DCNN spatio-temporal fusion model

This study presents a machine learning model combining Vision Transformer and 2D convolutional neural networks to predict epileptic seizures using EEG data with spatial and time-frequency information. The model achieved high accuracy (97.95%) on a public dataset by integrating electrode topology and brain complexity patterns to distinguish between preictal and interictal states.