Clinical BCI Challenge-WCCI2020
(Supported by IEEE FSTC TFs)
In this competition we will provide a dataset of EEG (encephalography) brain imaging data for 10 hemiparetic stroke patients having hand functional disability. The dataset consists of two classes which are left and right-hand grasp attempt movements. The participants in the competition will be required to provide accurate and robust decoding of these movements, from the provided brain activity only. The automated decoding of these kinesthetic movements from brain signals is helpful for the development of robot-assisted therapies or interfaces for assistive technologies or rehabilitation. The decoding will be done in two ways: 1) one is within-subject classification where the training data from the same subject will be used to classify that subject’s test data, 2) another challenge will be to perform these decoding across subjects where the training data of the 8 out of 10 subjects will be used to predict the test data of the remaining 2 subjects. These will require the development of novel computational intelligence models of high generalizability to solve this challenge. The dataset will be provided in .mat (MATLAB) format along with documentation describing the file components in a GitHub repository.
The participants can use any computational intelligence technique using the training and test sets provided by the competition. The submissions can be done in any programming language, software, platform or toolbox. The source code and instruction about how to run it, and a description of the data processing pipeline/method must be provided open-source repository (GitHub).
'Submission' means the material submitted by you in the manner and the format specified on the Website via the Submission form on the Competition Website. The submission will include a 4-page IEEE conferences format paper describing the methods. A link to the open-access software repository were the original code has been uploaded (GitHub, Bitbucket or similar). And a Microsoft Excel file with the predictions for the test data. This file must include the following columns: 1) subject ID, 2) trial index, 3) prediction. The winners of the competition will be based on the kappa value of the predictions with respect to the original labels.
Submissions must be received prior to the Competition deadline and adhere to the guidelines for Submissions specified on the Website. The sharing of codes and data privately outside the team is prohibited. The competition host has the right to publicly disseminate any entries or models. The solutions need to be made available under a popular OSI-approved license in order to be eligible for the award/prize. There is no maximum team size. In case of multiple submissions from the same team, the latest submission will be considered for evaluation. The winners of 1) within subject and 2) cross subject challenge will be announced separately. Results from the competition will be announced during the conference and it will be publicized via IEEE channels. For "within subject" prediction use the training file for a particular subject and predict the labels of all the 40 trials given in the "evaluation/testing" file of the same subject. Make table as mentioned above for all the predictions with 3 columns as subject name, trial index, and prediction for all the subjects P01 to P8. Remember that the predictions should be given for all the participants and all the trials. Any partially incomplete submission will be disqualified. Tables must given in "Microsoft Excel file" and separate sheets within the "Microsoft Excel file" should be there for separate subjects predictions. For cross-subject prediction the training data of participant P01 to P08 will have to be used to make a generalised classifier which will be used predict the classess of participant P09 and P10. It is to be noted that the training files for P09 and P10 would not be there. The results of Cross-subject prediction needs to be uploaded in a separate "Microsoft Excel file" file containing separate sheets for subject P09 and subject P10. There will be a submission link in the website, which will redirect to a google form asking for "TeamName", "Name of the team-lead", "Affiliation", "Contact details" and "file upload options" for "within subject" and "cross-subject" results. There would also be a text box to share the link of the github repository where the code must be shared and readme file (*.md) should be there describing the algorithm "how to run" instruction for the code. A PDF file upload option would also be there to upload 4-page IEEE conferences format paper describing the methods. Please note that individuals may only belong to a single team.
An example of submitting the predictions on test data is given as follows:
July 5th, 23:59 (GMT)
Dataset download link
The dataset can be downloaded from the github repository as shared below. The "readme file in the repository contains the detailed description of the dataset along with other information. Please read the "README.md" carefully and thoroughly before attempting the competition.
The submission will be through a google form the link of which will be uploaded soon. The participants need to upload separate "microsoft excel worksheet" (.xlsx file) files for "within subject" and "cross-subject results."