Bug encountered when computing LCMV spatial filter with a Label

💬 Support & Discussions
1

mne.beamformer.make_lcmv works without any label chosen but crashes when a Label instance is chosen with the label= argument.

Please help! Thank you :slight_smile:

OS details:

  • MNE version: 1.3.0
  • operating system: Ubuntu 20.04

Code snippet:

one_label = mne.read_labels_from_annot(subject, regexp='rostralanteriorcingulate-lh', subjects_dir=subjects_dir)[0]
one_label

# LCMV
filters_tmp = make_lcmv(
info=raw.info,
forward=fwd_vol,
data_cov=data_cov,
reg=0.05,
noise_cov=noise_cov,
label=one_label,
pick_ori="max-power",
weight_norm="unit-noise-gain",
rank=None,
)
# You can save the filter for later use with:
# filters.save('filters-lcmv.h5')
stc_LCMV_tmp = apply_lcmv_raw(raw, filters_tmp)
stc_LCMV_tmp.save(save_dir+sub_dir_name+f"{selected_labels[i].name[:-3]}_LCMV",overwrite=True)

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[17], line 5
      2 one_label
      4 # LCMV
----> 5 filters_tmp = make_lcmv(
      6 info=raw.info,
      7 forward=fwd_vol,
      8 data_cov=data_cov,
      9 reg=0.05,
     10 noise_cov=noise_cov,
     11 label=one_label,
     12 pick_ori="max-power",
     13 weight_norm="unit-noise-gain",
     14 rank=None,
     15 )
     16 # You can save the filter for later use with:
     17 # filters.save('filters-lcmv.h5')
     18 stc_LCMV_tmp = apply_lcmv_raw(raw, filters_tmp)

File <decorator-gen-484>:12, in make_lcmv(info, forward, data_cov, reg, noise_cov, label, pick_ori, rank, weight_norm, reduce_rank, depth, inversion, verbose)

File ~/anaconda3/envs/eegenv/lib/python3.9/site-packages/mne/beamformer/_lcmv.py:162, in make_lcmv(info, forward, data_cov, reg, noise_cov, label, pick_ori, rank, weight_norm, reduce_rank, depth, inversion, verbose)
    158 if inversion == 'single':
    159     depth['combine_xyz'] = False
    161 is_free_ori, info, proj, vertno, G, whitener, nn, orient_std = \
--> 162     _prepare_beamformer_input(
    163         info, forward, label, pick_ori, noise_cov=noise_cov, rank=rank,
    164         pca=False, **depth)
    165 ch_names = list(info['ch_names'])
    167 data_cov = pick_channels_cov(data_cov, include=ch_names)

File ~/anaconda3/envs/eegenv/lib/python3.9/site-packages/mne/beamformer/_compute_beamformer.py:54, in _prepare_beamformer_input(info, forward, label, pick_ori, noise_cov, rank, pca, loose, combine_xyz, exp, limit, allow_fixed_depth, limit_depth_chs)
     52 # Restrict forward solution to selected vertices
     53 if label is not None:
---> 54     _, src_sel = label_src_vertno_sel(label, forward['src'])
     55     forward = _restrict_forward_to_src_sel(forward, src_sel)
     57 if loose is None:

TypeError: cannot unpack non-iterable Exception object
2

Could you use the MNE sample data and open an issue on the GitHub with a minimally reproducible example please?

Well probably actually mne.read_labels_from_annot results a list and so you just want the first element of that list one_label[0]. In general, a pro tip is to pay attention to the plurality as in read_labels_from_annot which will tell you that it probably returns multiple things in a list and if you only want one thing, you need to grab the first element from that list.

3

Thanks for your response!

Calling one_label results in the following output
<Label | sample, 'rostralanteriorcingulate-lh', lh : 1864 vertices>.

I believe this is the correct format for a single Label object.

I will go ahead and open the issue on GH.