mne-tools
Bug with plot_topomap and coordinate transformations following #11081
Description of the problem
Following #11081, topomaps created with digitization points provided as Info don't look the same.
It has to do with dev_head_t (I think). Now that info is not None, sphere value is replaced from None to 'auto' here https://github.com/mne-tools/mne-python/blob/4310c92b85fa7dcc77ffe322501037e5ad82372d/mne/utils/check.py#L848-L857
which means that we now enter this statement https://github.com/mne-tools/mne-python/blob/4310c92b85fa7dcc77ffe322501037e5ad82372d/mne/utils/check.py#L859-L869
Steps to reproduce
from mne.datasets import testing
from mne.io import read_raw_fif
from mne.viz import plot_topomap
directory = testing.data_path() / "MEG" / "sample"
fname = directory / "sample_audvis_trunc_raw.fif"
raw = read_raw_fif(fname, preload=False)
raw = raw.crop(0, 10).pick("eeg").load_data().apply_proj()
plot_topomap(raw.get_data()[:, 101], raw.info)
Expected results
On mne-stable 1.1.1, a nice looking topographic map:
Actual results
On mne-main 1.2, a weird looking topographic map:
It's the MNE sample dataset "MEG" from which I picked only the EEG sensors. In the example above, I used the testing dataset instead of sample, but it's the same. e.g.
from mne.datasets import sample
directory = sample.data_path() / "MEG" / "sample"
fname = directory / "sample_audvis_filt-0-40_raw.fif"
With the filtered file above, the info representation is:
<Info | 15 non-empty values
bads: 1 items (EEG 053)
ch_names: EEG 001, EEG 002, EEG 003, EEG 004, EEG 005, EEG 006, EEG 007, ...
chs: 60 EEG
custom_ref_applied: False
dev_head_t: MEG device -> head transform
dig: 146 items (3 Cardinal, 4 HPI, 61 EEG, 78 Extra)
file_id: 4 items (dict)
highpass: 0.1 Hz
hpi_meas: 1 item (list)
hpi_results: 1 item (list)
lowpass: 40.0 Hz
meas_date: 2002-12-03 19:01:10 UTC
meas_id: 4 items (dict)
nchan: 60
projs: Average EEG reference: on
sfreq: 150.2 Hz
>
@mscheltienne I am lost. I see a PR in pycrostates.
what is the bug to fix?
Sorry, if that was not clear. Let's start over ;)
A while ago, I fixed a bug in mne-python, sphere='eeglab' was not working. PR: https://github.com/mne-tools/mne-python/pull/11081 But that PR had a side-effect that we did not notice at the time, namely, it messed up how the plot looks (c.f. the 2 screenshots and the code snippet with the MNE sample dataset in the first post).
I noticed this visualization issue because the CI testing pycrostates against mne-main started failing. As far as this CI is concerned, there are 2 issues: one in pycrostates (an oversight that I'll fix) and one in mne-python, the visualization issue I described here. Thus the x-ref since pycrostates fix depends on this one as well.
I did not dig up enough yet to figure out why the topoplot is messed up now that https://github.com/mne-tools/mne-python/pull/11081 has been merged.
Closing, as this is not a bug but an unforeseen bug fix by #11081.
Previously (MNE 1.1), if sphere=None was provided; the default sphere size 0.095 was used.
After (MNE 1.2), the sphere is set to 'auto' and is fit to the electrode location. As the sample dataset does not have template but true EEG sensor location, the auto fit yields a different head-circle position, which is the correct behavior and was ignored.
