I am working with ecog data. My goal is to run ICA on the epoched signals and then get the reconstructed time series signal from the selected ICA components. To visualize the ICA components to exclude
ica.plot_components() is typically used but meg value must be one of [‘grad’, ‘mag’, ‘planar1’, ‘planar2’] or bool, not ecog
I am wondering if there is a way to plot the components in 2d/3d space for epoched ecog data with a given montage ?
Also would it better to use sklearn like in this example given I am running ICA on epochs that have evoked data ? Though this tutorial seems to only show ICA transformed plots on evoked data but not really the ICA components Analysis of evoked response using ICA and PCA reduction techniques — MNE 1.4.2 documentation
It looks like ch_types only allows meg or eeg channels, if they are not picked by default, you could change the channel type of your data object to eeg in order to use it for this function (on a copy of your data). Otherwise, I think you’d need a PR.
Up to you on using the unsupervised filter, it might make the evoked signal cleaner but you might not need it and it could just add complexity to your methods.