[FieldTrip] What does minimum-norm estimate tell me + correlation in the source space

Mon Mar 14 09:40:36 CET 2016

Hi Runnan,

Sorry for the confusion. The part of the documentation you are working with is currently being improved, but not finalized yet.
In particular, the part you refer to ‘Source model: co-registration of the source space to the sensor-based coordinate system’ needs a heavy update. I would assume that you have done the tutorial up until there. In particular, if you have done the preprocessing of the anatomical MRI, where you have created 2 transformation matrices (using the same anatomical volume), you’re almost there. So please forget about the orig.mgz, either nomask or not, and please be a bit patient until we (i.e. I) find time to update the documentation.

This is how you get the appropriate coregistration:

(Note: I assume that all the objects are defined in ‘mm’, otherwise it won’t work. To get all things in ‘mm’, use ft_convert_units).

Preliminaries:
-we know how to get from an arbitrary voxel-space to the sensor-based coordinate system: transform_vox2ctf
-we know how to get to the ‘mni’ coordinate system from the same arbitrary voxel-space: transform_vox2spm
-we know that the sourcespace is expressed in the ‘mni’ coordinate system (check the units, is it mm?)

Based on the first two points above, we can easily construct a transformation matrix that goes from ‘mni’ to ‘ctf’:

T = transform_vox2ctf/transform_vox2spm

We can now use T to transform the sourcespace:

sourcespace = ft_transform_geometry(T, sourcespace);

I am pretty sure that this will do the trick.

Best,
Jan-Mathijs

On 13 Mar 2016, at 13:20, Harold Cavendish <harold.cav89 at gmail.com<mailto:harold.cav89 at gmail.com>> wrote:

Dear Runnan,

I'd be happy to help, but sadly, I have no experience with FreeSurfer and skipped that portion of the tutorial entirely (instead, I used files from fieldtrip/template).

However, using just common sense, I believe there's an error in the tutorial because orig-nomask.mgz is mentioned for the first time where you need to load it but it isn't present in any of the outputs or in the Subject01 directory. I believe it's just a renamed copy of orig.mgz which is created as explained in the tutorial.

As far as I understand, you take an anatomical MRI image (e.g. Subject01.mri), use FT to do some preprocessing and get Subject01.mgz (a FreeSurfer-compatible format), then do some segmentation and get Subject01masked.mgz, then do some conversion in FreeSurfer and transform those two into orig.mgz and brainmask.mgz respectively, and to make it even more convoluted, some filled.mgz pops up as well. To be honest, I don't follow… I have zero experience with FreeSurfer as well as MRI/fMRI analysis.

Briefly, try using orig.mgz instead of orig-nomask.mgz whenever you need to use ft_read_mri('orig-nomask.mgz'). If it doesn't do the job, hopefully someone else will come to the rescue.

Best regards,
Harold

On 12 March 2016 at 01:02, <runna90 at 126.com<mailto:runna90 at 126.com>> wrote:
Dear Harold,
I am very sorry that I am not writing to reply your question but to request you something about minimum norm estimate because I am a rather new fieldtripper. I noticed that you have got results applying MNE, thus I think you must be more experienced! I followed the tutorial at http://www.fieldtriptoolbox.org/tutorial/minimumnormestimate and did source reconstruction with  MEG data acquired from ctf 275 system. But I got stuck in here
. Source model: Co-registration of the source space to the sensor-based head coordinate system,
I could not find the file named ‘orig-nomask.mgz’ under the mir folder. Would you tell me in which step should this file be created? And how to create this file?

Best regards,

Runnan

在 2016年3月10日，08:40，Harold Cavendish <harold.cav89 at gmail.com<mailto:harold.cav89 at gmail.com>> 写道：

Dear FieldTrip users,

I have a few questions regarding source localisation and statistics:

1. Suppose I'm doing source reconstruction of ERP using minimum-norm estimate; one condition, multiple subjects. I'm using ft_sourceanalysis() and getting rather meaningful results which I visualise via ft_plot_mesh().

What is the validity of this result? I understand there are many individual dipoles spread over the cortical sheet and we're estimating best fit to the data given the conductance model while minimising overall power, but what does this actually tell us? If I find an area with higher power, what conclusions can I draw?

2. Can I compute a correlation between the result in the source space and a behavioural variable? Perhaps even specify a region of interest? If yes, is this approach considered to be valid?

I want to see if e.g. the power of the reconstructed activity in the source space averaged over a time window correlates with my behavioural variable.

Thank you!

Best regards,
Harold
_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl<mailto:fieldtrip at donders.ru.nl>
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl<mailto:fieldtrip at donders.ru.nl>
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl<mailto:fieldtrip at donders.ru.nl>
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160314/a4582655/attachment-0001.html>