adjacency matrix for laplace filtered data

💬 Support & Discussions
, ,
1

Hello MNE Team,

I am running CSD on epoched, cleaned data.

Afterwards I want to compare the activity between two conditions using a cluster based permutation test. Unfortunately the input (ch_type) for setting up the adjacency matrix is only allowed to be EEG or MEG channels but not CSD.

epochs_a = mne.preprocessing.compute_current_source_density(epochs_a)

ch_matrix, ch_list = mne.channels.find_ch_adjacency(epochs_a.info, ch_type='eeg')

Is there a reason why it is not recommended to run a CBP test on laplace filtered data? Otherwise I would suggest to allow for laplace filtered data as input to ch_type.

Best,

Carina

2

Hello @CarinaFo, I’m not very familiar with CSD, but it seems that currently, running compute_current_source_density() not only removes the existing EEG channels and replaces them with CSD channels, but it also drops the EEG montage.

This means that there’s no information about the location of those (virtual) CSD channels available, and therefore no adjacency matrix can be calculated.

What I believe you could do to work around this is to compute the adjacency matrix on the data with the EEG channels, and then use this matrix in the cluster-based permutation test with the CSD data.

However like I said, I’m not an expert in CSD and I haven’t tried what I just proposed, so please take this advice with a big grain of salt. But it might be worth a shot!

Best wishes,
Richard

1 Like