Introduction

This is the repository of my undergraduate research project. The goal of this project is to find the relationship between EEG signals and different mental states of human. Currently, this project uses a GNN model to classify EEG signals into 2 states: rest and motor. The architecture of this model originates from the correlation of brain signals within the same brain areas. So this project also implements a visualization tool to show the connectivity matrix, the parameters of the GNN layer in the model.

Environment

The environment of this project is managed by conda. To get the environment ready, here's the steps:

1. Install conda if you haven't.
2. Add conda to your system path.
3. Run the following command from project root: conda env create -f environment.yml -n {env_name}. Replace {env_name} with the name you want to give to the environment. This environment takes 8.02GB on my Windows 11 machine, so make sure you have enough space.
4. To execute scripts in this project, run conda activate {env_name} first to activate the environment.

Dataset

Currently, this research project uses WAY-EEG-GAL only.

To get the data ready, here's the steps:

1. Download the dataset from figshare. This dataset includes 12 sub-datasets, one for each participant. The sizes of those sub-datasets vary from 700MB to 1000MB, here's the basic information and download page of them. You don't need to downloa all of them, but if I recommend downloading at least 2 of them, so that you can test the code on different participants.

   participant	size	download page
   1	799.88MB	P1.zip
   2	905.53MB	P2.zip
   3	709.43MB	P3.zip
   4	764.96MB	P4.zip
   5	803.37MB	P5.zip
   6	822.52MB	P6.zip
   7	988.70MB	P7.zip
   8	745.41MB	P8.zip
   9	793.71MB	P9.zip
   10	812.92MB	P10.zip
   11	847.04MB	P11.zip
   12	866.80MB	P12.zip

2. Move the downloaded zip files into data/WAY-EEG-GAL folder.
3. Run the following command from project root: python script/dump_all.py data/WAY-EEG-GAL. If you want to preserve the zip files for whatever reason, you can add --no-delete option to the command. Note that this script relies on the original file names, so if you want it to work, don't rename the zip files.
4. After the command succeeds, the data is now ready to be used. The structure of the data folder should looks like following:

   data
   └── WAY-EEG-GAL
       ├── sub-01
       │   ├── raw
       │   │   ├── HS_P1_S1.mat
       │   │   ├── ...
       │   │   ├── HS_P1_S9.mat
       │   │   ├── WS_P1_S1.mat
       │   │   ├── ...
       │   │   └── WS_P1_S9.mat
       │   ├── series-01
       │   │   ├── metadata.json
       │   │   └── samples.npz
       │   ├── series-02
       │   │   ...
       │   └── series-09
       ├── ...
       └── sub-12
   

   The raw folder contains the original mat files extracted from the zip files. These MatLab files are not further used in this project, so you may delete them at your will. You'll save approximately 75% of the disk space by doing so.