drive.google.com/file/d/0B93wwoGtlGkmRWJWTlV2cG16SGM/view?usp=sharing&resourcekey=0-JJ728b-0px8QTUtiFrBfKA We recently showed that Classification Accuracy (CA), the typical metric used to study BCI user training, may hide some learning effects or hide the user inability to self-modulate a given EEG pattern.
Indeed, CA is an appropriate measure to study how well the BCI can decode the users' mental commands, but on its own, it is not enough to study how well users' are producing the EEG patterns used to drive the BCI and how well they are learning to do so. To address these issues, we proposed new performance metrics to specifically measure how distinct and stable the EEG patterns produced by the user are. Our new metrics could reveal learning effects that CA missed, as well as identify when a mental task performed by a user was no different than rest EEG. In order to ease the adoption of such metrics by the BCI community, we provide their Matlab codes for free and open-source on this page.
Note that to use them you will need the Matlab Riemannian geometry toolbox by Alexandre Barachant, available for free and open-source there: https://github.com/alexandrebarachant/covariancetoolbox
Please cite the corresponding paper if you use this code in your publications.
Code
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Spatial metrics, for BCI with two classes only:
Paper:
F. Lotte, C. Jeunet, “Online classification accuracy is a poor metric to study mental-imagery based BCI user learning: An experimental demonstration and new metrics”, 7th international BCI conference, 2017 – pdf
Code: download here
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Spatial and spatio-spectral metrics, for BCI with any number of classes
Paper:
F. Lotte, C. Jeunet, “Defining and Quantifying Users’ Mental Imagery-based BCI skills: a first step", Journal of Neural Engineering, 2018 - pdf
Code: download here
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Note: these are Matlab codes. Alexandre Barachant maintains a Python toolbox about Riemannian geometry (PyRiemmann - http://pythonhosted.org/pyriemann/) , so if you are a Python user, you should be able to implement these metrics in Python fairly rapidly using that toolbox.