Home
Remove artifacts from Event-Related Potentials (ERPs) with Adaptive Common Spatio-Temporal Pattern Filter (ACSTP) from continuous EEG recordings.
-Regression-based trials averaging (reduce the effect of overlapping ERP)
-Automatic Subspace reduction (find the best subspace to reject the
artifacts while keeping the maximum of useful information)
-Automatic Jitter Correction at the trial level (ERP peaks are sharper)
-Automatic Trial Weighting (Amplitude correction and rejection of high-energy
trials often related to artefacts)
-less trials required to have a clean ERP estimation
-removal of artifacts such as EMG, EOG, blink, ... that could have distorded the
real shape of the ERP with the standard Arithmetic Ensemble Average (AEA).
-estimation of linear spatio-temporal filter maximizing the signal noise
ratio of the evoked potential(s) according to some user-defined masks
(see below).
-a well-chosen temporal mask for the evoked potential of interest
-a well-chosen spatial mask for the evoked potential of interest
-while the code was heavely tested by the authors, it is still an ongoing
work and bugs will happen. Please contact us if you have troubles to make
it works for your data. ([email protected])
-we recommand to downsample your signal to remove most of the
non-informative frequencies (e.g. Fs=128Hz should be fine for most ERP
studies).
-temporal filter cut-off frequencies should be considered with care as it can distord the
waves. Unfortunately, this method heavely relies on empirical covariance
matrices and low-frequencies implies often ill-conditionned
covariances matrices on short epochs. The ACSTP has shown succesful
results on zero-phase distorsion band-pass filtered data between [0.5-30] Hz.
-Matlab 2011b or above
-download the last snapshot
-put in any folder
-run the installer.m (tested on windows)
-a detailed Matlab tutorial is available : script_ACSTP_test.m with a pre-processed toy dataset EEG_ex.mat.
-please open in issue on GitHub with the detailed Matlab error.
If you use the code, please cite (the methods are described in details inside the paper): Marco Congedo, Louis Korczowski, Arnaud Delorme, Fernando Lopes da Silva. Spatio-temporal common pattern: A companion method for ERP analysis in the time domain. Journal of Neuroscience Methods, Elsevier, 2016, 267, pp.74-88. <10.1016/j.jneumeth.2016.04.008>.
Download the pdf https://hal.archives-ouvertes.fr/hal-01343026/document