Dual-Cross Tri-Level Routing Transformer Based Metric Learning Network for Epileptic Seizure Prediction Using Single-Channel iEEG

This paper presents DC-TRT-MLNet, a deep learning model designed to predict epileptic seizures from single-channel intracranial EEG recordings for closed-loop brain stimulation therapies. The approach uses graph attention networks and transformer architectures to capture temporal and spectral dependencies while employing metric learning to distinguish pre-seizure states from normal brain activity.