[BUG] `stc_near_sensors` unexpected behavior

[alexrockhill]

Description of the problem

Trying to use stc_near_sensors there is one bug if there are any bad channels, that causes

if picks is None:
    picks = pick_types(evoked.info, ecog=True, seeg=True, dbs=True)
evoked = evoked.copy().pick(picks)
frames = set(evoked.info["chs"][pick]["coord_frame"] for pick in picks)

This to fail with an index error because when evoked gets picked the bad channels aren't included.

There is another bug/API confusion that the default for surface is 'pial' so if you pass a volume source estimate and no argument for surface then you get an index error because it tried to use the pial surface which has the wrong indices for your volume source estimate. That seems unnecessarily confusing, I think it should work reasonably with default arguments for both surface and volume source estimates.

Also while doing a MWE, this link in the issue template is broken https://mne.tools/dev/overview/datasets_index.html.

Steps to reproduce

import mne

misc_path = mne.datasets.misc.data_path()
subjects_dir = misc_path / "seeg"

raw = mne.io.read_raw(misc_path / "seeg" / "sample_seeg_ieeg.fif")
raw.info['bads'] = ['LENT 1']
trans = mne.coreg.estimate_head_mri_t("sample_seeg", subjects_dir)

epochs = mne.Epochs(raw, detrend=1, baseline=None)
epochs = epochs["Response"][0]  # just process one epoch of data for speed
evoked = epochs.average()

surface = subjects_dir / "sample_seeg" / 'bem' / 'inner_skull.surf'
vol_src = mne.setup_volume_source_space(
    subject="sample_seeg", subjects_dir=subjects_dir, surface=surface,
    pos=10)
vol_stc = mne.stc_near_sensors(
    evoked, trans, subject="sample_seeg",
    subjects_dir=subjects_dir, src=vol_src)  # surface=None)

Link to data

No response

Expected results

Working stc

Actual results

Two index errors

File ~/projects/mne-python/mne/source_estimate.py:3878, in <genexpr>(.0)
   3876     picks = pick_types(evoked.info, ecog=True, seeg=True, dbs=True)
   3877 evoked = evoked.copy().pick(picks)
-> 3878 frames = set(evoked.info["chs"][pick]["coord_frame"] for pick in picks)
   3879 if not frames == {FIFF.FIFFV_COORD_HEAD}:
   3880     raise RuntimeError(
   3881         "Channels must be in the head coordinate frame, " f"got {sorted(frames)}"
   3882     )

IndexError: list index out of range

File ~/projects/mne-python/mne/source_estimate.py:3930, in <listcomp>(.0)
   3928             rrs = apply_trans(trans, rrs)
   3929     else:
-> 3930         rrs = np.concatenate([s_rr[s["vertno"]] for s_rr, s in zip(surf_rr, src)])
   3931     keep_all = True
   3932 # ensure it's a usable one

IndexError: index 126967 is out of bounds for axis 0 with size 126967

Additional information

None

[alexrockhill]

cc @adam2392 @larsoner

[alexrockhill]

Also when I use that to do stc.plot_3d it works the first time and then gets killed the second identical time it's called so I think there might be a memory leak but that's totally separate and probably an issue with mne.viz.Brain.

[larsoner]

Yeah once it's picked you shouldn't use picks anymore so safest to del picks and just use frames = set(ch["coord_frame"] for ch in evoked.info["chs"])