ICA.plot_sources returns straight lines

💬 Support & Discussions
1
  • MNE version:0.24.0
  • operating system: Windows 10
    Dear mne experts, can anyone help me with this problem?
    My raw data seems alright and I applied ICA methods to erase the eye movement.

raw
raw2146×1037 441 KB

This is my raw data, it seems good here.
After filter and re-reference, I applied the ICA method, code here:

ica = mne.preprocessing.ICA(n_components=0.99999, method='picard', max_iter=1200)
        ica.fit(raw_filter)
        raw_filter.load_data()
        ica.plot_components()
        ica.plot_sources(raw_filter, show_scrollbars=True)
        plt.show(block=True)

after ICA, the plot_components method seems normal, the result is:

components
components975×967 128 KB

But the plot_sources method returns straight lines which are really strange.

sources
sources2124×1021 117 KB

I don’t know what went wrong. Could anybody help me with this plzzzzzz?

2

I would think you have a huge artifact somewhere in your data. You should mark it as BAD with
an annotation as ICA is not good for transient artifacts

ALex

2 Likes
3

I’m not sure it’s the problem, but asking ICA to explain 99.999% of the variance is asking a lot.
I would change this to an integer one less than the number of channels you have.
If you’ve interpolated any of those channels, you’ll want to use one less than the number of channels before interpolation.

4

Thank you a lot ALex!
This is exactly the problem in my case. I forgot to stop the recording after the experiment is done and the data of subjects taking off their hats are recorded. After I deleted that part the ICA performed normally

1 Like