神经科学与认知科学 突破级 暂无讲解视频
发表时间
2026-05-11
arXiv
2605.10337

收录解读

CORTEG investigates whether pretrained scalp-EEG foundation models can be adapted to intracranial ECoG decoding, reducing the need for extensive per-patient calibration.

The method combines a pretrained EEG backbone, electrode-aware spatial adaptation, dual-stream tokenization for low-frequency and high-gamma activity, and leave-one-subject-out fine-tuning.

The reported results show competitive or improved decoding on finger trajectory and audio envelope tasks, especially in low-data calibration settings.

For BCI and NeuroAI, the paper is useful because it treats cross-modality transfer as a deployment problem for patient adaptation, not merely as another representation-learning benchmark.

链接