EEG-SSFormer

Citation

If you use EEG-ChiMamba, please cite

   @article {Neves2026ChiMamba,
	author = {Neves, Christopher and Steele, Christopher J and Xiao, Yiming},
	title = {EEG-ChiMamba: Towards a Robust Mamba-Based Architecture for Dementia Detection from Resting State Electroencephalography},
	year = {2026},
	doi = {10.64898/2026.03.23.713697},
	journal = {bioRxiv}
    }

Dataset

The dataset must be downloaded from: https://github.com/ipis-mjkim/caueeg-dataset

Environment Setup

Requirements: Ubuntu 20.04 (project uses WSL) and CUDA 11.8.

1. Create a new Conda environment

conda create -n ceednetmamba python=3.10 -y 

2. Activate the Conda environment

conda activate ceednetmamba

3. Install PyTorch

conda install pytorch==2.0.1 torchvision==0.15.2 pytorch-cuda==11.8 -c pytorch -c nvidia

4. Install Causal-conv 1D and Mamba

pip install causal-conv1d==1.4.0
pip install mamba-ssm==2.2.2 --no-cache-dir

5. Install the caueeg-ceednet requirements (using their provided requirements.txt file)

pip install -r requirements.txt

6. If you numpy is at version 2, it must be downgraded

pip install --upgrade numpy==1.26.4

7. Clone the Dimentia Mamba repository

git clone https://github.com/HealthX-Lab/EEG-ChiMamba.git

8. cd into the repo

cd EEG-ChiMamba

9. Extract the dataset using 7z into the data/caueeg/caueeg-dataset subdirectory

7z x caueeg-dataset.zip

10. Convert the edf files to feather format

python ./src/external/caueeg/datasets/convert_file_format.py

How to run

To run the full train/test pipeline, run the train_test.sh script in the scripts/ directory.

Generate figures

The figures are generated using the occlusions_final.ipynb notebook in the src/notebooks/ directory. Simply run the cells in the notebook to generate the figures in the report.

Acknowledgements

This codebase builds upon the caueeg-ceednet and SiMBA repositories. We thank the authors for making their code publicly available, and we kindly request that you consider citing their works if relevant.