[Mne_analysis] mne_analyze, mne_compute_raw_inverse questions

[Gustavo Sudre]

I'm interest in the answers for these questions as well, but I'd like  
to jump in this discussion to ask a somewhat related question. I think  
I'm just having problems with the terminology, but it'd be good to  
clarify it.

When I run mne_compute_raw_inverse for a single label, I get the  
activation for all the sources within the label (correct?). In the  
screen output of this program, I see that my forward solution has  
about 13000 sources, and if my label is the entire left hemisphere I  
get the estimates for about 6500 sources, which makes sense. Using  
mne_analyze, when selecting a label there are several ways to combine  
the activations in the vertices, as Eliezer pointer out. However,  
based on how the brain was tessellated, it seems to me that there are  
a lot more vertices than sources, right? Is there a mapping between  
sources and vertices? If I run the same command with labeldir, do I  
get the average of the sources in each label, or vertices? Also, is  
there a way to get one of the other four options, instead of the  
average?

Thanks in advance,

Gus

On Jun 10, 2009, at 5:32 PM, Eliezer Kanal wrote:

> Hello folks -
>
> A colleague and I noticed some seeming inconsistencies in how these  
> two programs compute waveforms on a per-label basis, and I want to  
> see if we can clear this up.
>
> When using mne_compute_raw_inverse without the -labeldir switch, the  
> program provides output in terms of "sources per label", and  
> provides a waveform for each of those sources. When using the - 
> labeldir switch, the program provides a single waveform per label.  
> That waveform is NOT the average of all sources in the first method,  
> as can be seen by simply exporting with and without the switch, and  
> averaging the results of the second and comparing it to the first.
>
> When using mne_analyze, waveforms are obtained using the timecourse  
> manager, which seems to compute the values on a per-voxel basis,  
> with multiple options as to how the voxels can be combined  
> (averaging, maximum, etc.). A cursory examination shows that the  
> timecourse created using this model is different again from either  
> of the above two methods.
>
> I was wondering if one of the developers could explain the  
> difference between these three methods of acquiring an ROI  
> timecourse using a label. Thanks in advance -
>
> Sincerely,
> Eliezer Kanal
> _______________________________________________
> Mne_analysis mailing list
> Mne_analysis at nmr.mgh.harvard.edu
> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis