[Mne_analysis] morphing source spaces VS morphing stcs

[Denis-Alexander Engemann]

Dear list,

as this has been coming up repeatedly over the last weeks and months and I
was looking for a more performant way of dealing with morphing in
connectivity studies,
I set up a few comparisons between different available options that I
wanted to share with you.

All plots are based on this small script:

https://gist.github.com/dengemann/3b45637bbe3d89650255

Make sure your MNE-C tools are appropriately setup, as for morphing source
spaces I'm using the command line tools.

One general remark, while morphing a subject's source space seems to lead
to topological errors (this needs to be better investigated), however
morphing fsaverage to the subjects seems to work flawlessly.

To summarize, here are thee main observations: 1) in a free-orientation
setting source space morphing VS stc morphing give similar results. 2)
These differences are more enhanced for inverse solutions with surface
constraints, which is probably related to differences in folding, etc. 3)
However, these differences seem minimal when comparing activities in
labels. The differences seem to be constant across different subsampling
schemes (ICO values).

Here are a few slides with figures produced from that script:
https://drive.google.com/file/d/0B_62rpZloQ5bVHVGdExKT1puRlE/view?usp=sharing

These comparisons can clearly be improved but it's a start. Please share
your thoughts and feel free to carry on with this.
I personally have the impression that it should be ok to morph fsaverage
source spaces to a given subject, the speed-accuracy trade-off seems good.

Cheers,
Denis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20151008/968a6920/attachment.html