Brant: Foundation Model for Intracranial Neural Signal
Brant is a foundation model in the field of intracranial recordings, which learns powerful representations of intracranial neural signals.
- Model Homepage: https://zju-brainnet.github.io/Brant.github.io/
- Github: https://github.com/yzz673/Brant
News
[2025.12.04] Considering that some users still have questions about the spectrum power calculation described in the paper, we have updated the code with the calculation process we designed for reference. Please refer to
pretrain/pre_utils.[2025.05.17] We have uploaded the model code on HuggingFace. Now anyone can access the finetuning code without sending email to the authors!
[2024.05.08] Due to the request from our collaboration partner, we are temporarily unable to directly share our code publicly. If you require access, please contact [email protected].
Source Code
The source code of Brant is provided in Brant_src.
Brant_src/train1.py,Brant_src/evaluate1.pyare the code for the seizure detection task.Brant_src/train2.py,Brant_src/evaluate2.pyare the code for short- and long-term, frequency-phase forecasting tasks.Brant_src/train3.py,Brant_src/evaluate3.pyare the code for the imputation task.
Pre-trained weights
We also release the pre-trained weights of Brant
You can find the usage of the pre-trained embeddings in
Brant_src/utils.py:get_emb()
Disclaimer
The pre-training data for the Brant model was collected during routine treatment procedure of epilepsy patients from a hospital, and is intended solely for medical or research use. The pre-trained weights of Brant are released only for medical or research purposes and must not be subjected to any form of misuse.