Current dipole moment densities

system · March 31, 2016, 3:20pm

Hi All,

I'm trying to plot current dipole moment densities (e.g. Murakami and Okada
2015 NeuroImage) in mne_analyze and mne-python, and it all seems rather
confusing.

As far as I can tell based on the manual, MNE-C can compute densities if
the cortical patch statistics are available. But I'm confused about when
the cortical patch statistics are used after they are calculated. What do I
need to do to specify that I want current moment densities rather than
unnormalized moments?

Then when it comes to MNE-python I can't find any references to current
densities. Will I need to implement this myself? I think I can see how to
do it using the 'tri-area' field combined with the 'pinfo' vertices in the
SourceSpaces, and the current dipole moments from the SourceEstimate. I'm a
bit worried that I'm not understanding the relationship between the
vertices/patches in the SourceSpaces and the SourceEstimate (since the stc
doesn't have all of the patches from the src), and that this might affect
how I compute the area. Can I just forget about the patches in the
SourceSpaces that aren't in the SourceEstimate?

Thanks,
Emily
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20160331/268cde0c/attachment.html

system · April 1, 2016, 8:01am

hi Emily,

I'm trying to plot current dipole moment densities (e.g. Murakami and Okada
2015 NeuroImage) in mne_analyze and mne-python, and it all seems rather
confusing.

As far as I can tell based on the manual, MNE-C can compute densities if the
cortical patch statistics are available.

But I'm confused about when the
cortical patch statistics are used after they are calculated. What do I need
to do to specify that I want current moment densities rather than
unnormalized moments?

the forward BEM code makes use of it. It basically yields surface
weighted forward fields.

see

github.com

mne-tools/mne-python/blob/main/mne/forward/forward.py#L451


      
          if tag is None:
              raise ValueError("MRI/head coordinate transformation not found")
          cand = tag.data
          if cand["from"] == coord_mri and cand["to"] == coord_head:
              info["mri_head_t"] = cand
          else:
              raise ValueError("MRI/head coordinate transformation not found")
          
          # Get the MEG device <-> head coordinate transformation
          tag = find_tag(fid, parent_meg, FIFF.FIFF_COORD_TRANS)
          if tag is None:
              raise ValueError("MEG/head coordinate transformation not found")
          cand = tag.data
          if cand["from"] == coord_device and cand["to"] == coord_head:
              info["dev_head_t"] = cand
          elif cand["from"] == coord_ctf_head and cand["to"] == coord_head:
              info["ctf_head_t"] = cand
          else:
              raise ValueError("MEG/head coordinate transformation not found")
          
          bads = _read_bad_channels(fid, parent_meg, ch_names_mapping=ch_names_mapping)

github.com

mne-tools/mne-python/blob/main/mne/minimum_norm/inverse.py#L1294


      
                  tstep=tstep,
                  subject=subject,
                  vector=(pick_ori == "vector"),
                  source_nn=source_nn,
                  src_type=src_type,
              )
              logger.info("[done]")
          
              return stc
          
          
          def _apply_inverse_epochs_gen(
              epochs,
              inverse_operator,
              lambda2,
              method="dSPM",
              label=None,
              nave=1,
              pick_ori=None,
              prepared=False,
              method_params=None,

so when you use MNE or a sparse solver that returns source amplitudes
you should get source densities when the patch info is present.

ok?

ALex

system · April 1, 2016, 7:54pm

Hi Alex,

Are you saying that MNE-python does that?

When I run standard MNE solutions in python (no depth-weighting, fixed
orientation sources) using patch_stats=True (or False) in
mne_read_source_spaces, I get exactly the same forward solution and source
estimates. If the code were normalized by patch area for patch_stats=True,
I would have expected them to be quite different.

Code is attached. I've also included a version of the src file where the
patch statistics were computed using MNE-C (eeganes07-ico-3p-src.fif). It
produces a different forward solution, but the resulting stcs are nearly
identical (I think this comes from using the patch statistics in the
surface normal decision, which doesn't happen if the patches were computed
using mne-python, possibly because 'patch_inds' is never set in the
SourceEstimates). Sorry for the large download size (it's the BEM file).
?
patch_stats.zip
<https://drive.google.com/file/d/0B8PjVL8SDL7IQUhLWjhLWUpiMFU/view?usp=drive_web>
?
Emily

Topic		Replies	Views
Find projected current on sensors Support & Discussions source-localization , visualization , evoked , dipole-fit	14	531	June 14, 2022
source estimate stc files and current source density estimates Mailing List Archive (read-only) list-archive	6	244	April 10, 2014
Strange behavior of mne.preprocessing.current source density Mailing List Archive (read-only) list-archive	1	185	June 3, 2020
'mne.preprocessing' has no attribute 'compute_current_source_density' Mailing List Archive (read-only) list-archive	2	424	April 11, 2020
Source current directions Mailing List Archive (read-only) list-archive	3	159	April 21, 2014

Current dipole moment densities

Related topics