CORTEG: Foundation Models Enable Cross-Modality Representation Transfer from Scalp to Intracranial Brain Recordings
CORTEG adapts large, pretrained EEG foundation models to ECoG for cross-patient learning in brain-computer interfaces. It combines a pretrained EEG backbone with electrode-aware spatial adapters and a dual-stream tokenizer to process both low-frequency and high-gamma signals. This allows the model to achieve state-of-the-art performance on regression tasks while enabling rapid, data-efficient calibration for new patients.