Thank you again for the good discussing on that previous issue. Now I
have a more MNE-specific question.
Suppose I run compute_mne_raw_inverse using a single label (e.g.
cortex-lh.label), and then convert the output of that function to
a .mat file using mne_raw2mat. In the resulting .mat file I see a
[channels] by [time] matrix with data, which makes sense. Judging by
the number of channels, I imagine they map to the sources within that
label, and not the vertices. Then, in the info structure there are
names for all these channels, and if I understand it correctly that
contains the number of the vertex that the source maps to. In that
case, is there a 1-1 relationship between sources and vertices? That
would mean that there's an awful lot of quiet voxels in the results
(e.g. given the ~3000 sources I get in the left hemisphere, using 7mm
spacing, and 160K vertices being indexed). Shouldn't a given source
affect more than one vertex in the brain, if the triangles are small
enough?
I guess my question boils down to the definition and relationship
between vertices and sources. Please let me know if I'm mistaken in my
line of thought.
Thank you again for the good discussing on that previous issue. Now I
have a more MNE-specific question.
Suppose I run compute_mne_raw_inverse using a single label (e.g.
cortex-lh.label), and then convert the output of that function to
a .mat file using mne_raw2mat.
First, the use of mne_raw2mat is quite unnecessary. You can read the
signals directly from the fif files with the Matlab routines. This
saves time and disk space. A good starting point is mne_ex_read_raw.
In the resulting .mat file I see a
[channels] by [time] matrix with data, which makes sense. Judging by
the number of channels, I imagine they map to the sources within that
label, and not the vertices. Then, in the info structure there are
names for all these channels, and if I understand it correctly that
contains the number of the vertex that the source maps to. In that
case, is there a 1-1 relationship between sources and vertices? That
would mean that there's an awful lot of quiet voxels in the results
(e.g. given the ~3000 sources I get in the left hemisphere, using 7mm
spacing, and 160K vertices being indexed). Shouldn't a given source
affect more than one vertex in the brain, if the triangles are small
enough?
I guess my question boils down to the definition and relationship
between vertices and sources. Please let me know if I'm mistaken in my
line of thought.
Anyhow, in the info structure contained in the mat file output by
mne_raw2mat there is a field ch_lognos which contains the vertex
numbers of the sources in the densely triangulated surface. There is
the logno field in the channel descriptors if you load the fif file
directly.
For visualization and morphing MNE uses a "spreading operator" to fill
in values to all vertices. This is discussed in section 8.3 of the MNE
manual.