EEG Motor Imagery Analysis with PyTorch and MNE Python

This repository provides a comprehensive pipeline for analyzing motor imagery EEG data using MNE-Python and PyTorch. It includes preprocessing scripts, CNN model training notebooks, and visualization tools, facilitating research and development in EEG-based brain-computer interfaces (BCIs).

Features

• Data Preprocessing: Utilizes MNE-Python for EEG data loading, filtering, and epoching.
• Deep Learning Models: Implements Convolutional Neural Networks (CNNs) for classifying motor imagery tasks.
• Visualization: Provides notebooks for visualizing EEG signals and model performance.
• Dataset Integration: Supports multiple publicly available EEG datasets via the MOABB framework.

Installation

1. Clone the repository:

   git clone https://github.com/berdakh/eeg-pytorch.git
   cd eeg-pytorch

2. Create a virtual environment (optional but recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install required packages:

   pip install mne==0.18.0
   conda install pytorch torchvision cudatoolkit=10.1 -c pytorch
   pip install braindecode

Note: Ensure that MNE-Python and PyTorch are installed, as they are central to the functionalities provided.

Usage

The repository contains multiple scripts and notebooks, each serving specific purposes. Here's a brief overview:

• Data Preprocessing:

  • nu_smrutils.py: Contains utility functions for EEG data preprocessing using MNE-Python.

• Model Training:

  • train_CNN_pooled.ipynb: Jupyter Notebook for training CNN models on pooled EEG data across subjects.
  • train_CNN_subspe.ipynb: Jupyter Notebook for training CNN models on subject-specific EEG data.

• Visualization:

  • visualization.ipynb: Jupyter Notebook for visualizing EEG signals and model performance metrics.

To execute a notebook:

1. Launch Jupyter Notebook:

   jupyter notebook

2. Open the desired .ipynb file and follow the cells sequentially.

Datasets

The repository integrates several publicly available EEG datasets through the Mother of All BCI Benchmarks (MOABB) framework:

• BNCI2014001: EEG data from 9 subjects performing four different motor imagery tasks (left hand, right hand, feet, tongue).
• BNCI2014004: EEG data from 9 subjects performing left and right hand motor imagery tasks.
• Weibo2014: EEG data from 10 subjects performing simple and compound limb motor imagery tasks.
• PhysionetMI: EEG data from 109 subjects performing various motor execution and imagery tasks.

Note: Please refer to the MOABB website for detailed information and access to these datasets.

Contribution

Contributions are welcome! If you have suggestions, bug reports, or enhancements, please open an issue or submit a pull request.

License

This project is open-source and available under the MIT License.

Acknowledgments

This repository was developed by Berdakh Abibullaev, focusing on EEG motor imagery analysis using MNE-Python and PyTorch.

---

For detailed explanations and methodologies, refer to the Jupyter Notebooks included in the repository.

If you need further assistance or have specific questions about any script or functionality, feel free to ask!