Graph Emotion Decoding (GED)

This repository contains the official code for the paper Graph Emotion Decoding from Visually Evoked Neural Responses (MICCAI 2022).

1  Installation

Follow the steps below to prepare the virtual environment.

Create and activate the environment:

conda create -n ged python=3.6
conda activate ged

Install dependencies:

pip install -r requirements.txt

2  Experiments

2.1  Download Datasets

The preprocessed fMRI data for five subjects and emotion scores are available at figshare. We have downloaded the file feature.tar.gz of emotion category scores and unzipped it into the sub-folder data/feature/ in this repository. However, we do not provide the fMRI data since the size of a single file is at least 1GB. Please follow the instructions below to download the required fMRI data. You can refer to the official website of Horikawa et al. for more details about the data.

Download the preprocessed fMRI data from here and unzip the downloaded file Subject?_preprocessed_fmri.tar.gz into the corresponding sub-folder data/fmri/Subject?/preprocessed/. Take "Subject 1" for example:

tar -zxvf Subject1_preprocessed_fmri.tar.gz

If downloaded and unzipped correctly, the files for Subject 1 can be found at data/fmri/Subject1/preprocessed/fmri_Subject1_Session[1-5].h5.

After getting all data for five subjects following the above steps, the folder tree of data should look like:

data/
  ├───feature/
  │     └───category.mat
  │         categcontinuous.mat
  └───fmri/
        ├───Subject1/
        │     └───preprocessed/
        │           └───fmri_Subject1_Session1.h5
        │               fmri_Subject1_Session2.h5
        │               ...
        │               fmri_Subject1_Session5.h5
        ├───Subject2/
        │     └───preprocessed/
        │           └───fmri_Subject2_Session1.h5
        │               fmri_Subject2_Session2.h5
        │               ...
        │               fmri_Subject2_Session7.h5
        ├───Subject3/
        │     └───preprocessed/
        │           └───fmri_Subject3_Session1.h5
        │               fmri_Subject3_Session2.h5
        │               ...
        │               fmri_Subject3_Session6.h5
        ├───Subject4/
        │     └───preprocessed/
        │           └───fmri_Subject4_Session1.h5
        │               fmri_Subject4_Session2.h5
        │               ...
        │               fmri_Subject4_Session5.h5
        └───Subject5/
              └───preprocessed/
                    └───fmri_Subject5_Session1.h5
                        fmri_Subject5_Session2.h5
                        ...
                        fmri_Subject5_Session5.h5

2.2  Run and Reproduce

Execute the following command to run and reproduce the experiments:

python main.py --subject_id <subject> --num_sessions <session> --fold_idx <fold>

where <fold> takes from 0 to 9, and <subject> takes from 1 to 5. Different <subject> corresponds to different <session>, and their relationships are listed as follows:

	sub 1	sub 2	sub 3	sub 4	sub 5
Session	5	7	6	5	5

3  Cite

If you find this code or our GED paper helpful for your research, please cite our paper:

@inproceedings{huang2022graph,
  title     = {Graph Emotion Decoding from Visually Evoked Neural Responses},
  author    = {Huang, Zhongyu and Du, Changde and Wang, Yingheng and He, Huiguang},
  booktitle = {International Conference on Medical Image Computing and Computer Assisted Intervention},
  year      = {2022}
}