I don't think there is a way to define uniquely the sign of the first
eigenvariate of a group of dipoles with different orientations. There
is some meaning to the sign of one dipole but not when you summarize a
group. So you can flip the sign if it suits you.
Vladimir
On Tue, Jan 31, 2012 at 4:17 PM, Anette Giani
<[log in to unmask]> wrote:
> Hi Vladimir,
>
> Thanks for having a look at the data. I will try using a fine mesh as well
> as COH or IID and let you know.
> One last question, however: Do you have any idea why the field intensities
> seem to be flipped/inverted depending on which radius you chose? (see
> dropbox folder: SourceExtractionVladimir.pptx, slide 4 and 5).
>
> Best,
> Anette
>
>
>
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] On
> Behalf Of Vladimir Litvak
> Sent: Tuesday, January 31, 2012 5:03 PM
> To: [log in to unmask]
> Subject: Re: [SPM] M/EEG: problems with sourcewave extraction
>
> Hi Anette,
>
> The problem with using radius of 0 was a bug. It's fixed in the attached
> version. Regarding the problem with sensitivity to the exact location it
> looks like it indeed has to do with MSP being very sparse solution so some
> vertices have very high values while others have low.
> So a difference of a mm in your location can result in some vertex either
> being included in the sphere or not and can make a big difference if that
> vertex has high activation.
>
> I can suggest 3 things you can try:
>
> 1) Use IID or COH instead of MSP. I tried IID and the results for -40
> -18 7 and -41 -18 8 look very similar.
> 2) Try using MSP with the fine mesh. I haven't tried it but it might help as
> there will be more vertices to average over. In 10mm radius there are just
> 2-3 vertices with the 'normal' mesh. They probably have different
> orientations which might explain why some have very strong activation and
> some very weak.
> 3) Increase the radius even more until you see that the sensitivity to
> precise location goes away.
>
> Let me know if it helps.
>
> Best,
>
> Vladimir
>
> On Mon, Jan 30, 2012 at 7:38 AM, Anette Giani
> <[log in to unmask]> wrote:
>> Hi Vladimir,
>>
>>
>>
>> Sorry for that. Apparently I stopped the upload process to early. All
>> files should be available now.
>>
>>
>>
>> Best,
>>
>> Anette
>>
>>
>>
>> From: Vladimir Litvak [mailto:[log in to unmask]]
>> Sent: Friday, January 27, 2012 12:50 AM
>> To: Anette Giani
>> Subject: Re: [SPM] M/EEG: problems with sourcewave extraction
>>
>>
>>
>> Hi Anette,
>>
>>
>>
>> I also need the mat file from the mcMra... dataset. You only sent the
> .dat.
>>
>>
>>
>> Vladimir
>>
>> On Wed, Jan 18, 2012 at 7:47 PM, Vladimir Litvak
>> <[log in to unmask]>
>> wrote:
>>
>> Dear Anette,
>>
>>
>>
>> The source data is not really helpful. I need the data you are
>> extracting sources from. You can send it via yousendit or dropbox.
>>
>>
>>
>> Vladimir
>>
>>
>>
>> On Wed, Jan 18, 2012 at 10:57 AM, Anette Giani
>> <[log in to unmask]> wrote:
>>
>> Thanks for the effort!
>>
>> Attached, I send you the script and the resulting .mat files. Sorry, I
>> cannot sen the original data, since it is too big. I hope that helps
>> already! Let me know if things are still unclear.
>>
>> Best,
>> Anette
>>
>>
>>
>>
>> On Tue, 17 Jan 2012 16:41:31 +0000
>> Vladimir Litvak <[log in to unmask]> wrote:
>>> Dear Anette,
>>>
>>> I cannot reproduce these problems. It'd help if you send me your data
>>> and scripts or description of what you did exactly.
>>>
>>> Best,
>>>
>>> Vladimir
>>>
>>> On Tue, Jan 17, 2012 at 2:17 PM, Anette Giani
>>> <[log in to unmask]
>>> > wrote:
>>>
>>
>>> > Dear Vladimir,****
>>> >
>>> > ** **
>>> >
>>> > Thanks for your fast reply! I still have a couple of comments
>>> > though:****
>>> >
>>> > ** **
>>
>>> >
>>> > I think this is a result of summarizing over a sphere by extracting
>>> > the first eigenvariate. If you specify the radius as 0 you should
>>> > get the same
>>
>>> > thing you are getting in the plot.****
>>> >
>>> > ** **
>>> >
>>> > Specifying a radius of 0 gives an error:****
>>> >
>>> > ** **
>>> >
>>> > ??? Attempted to access XYZ(2,1); index out of bounds because****
>>> >
>>> > size(XYZ)=[1,3].****
>>> >
>>> > ** **
>>> >
>>> > Error in ==> spm_eeg_inv_extract at 63****
>>> >
>>> > dist = sqrt(sum([vert(:,1) - XYZ(i,1), ...****
>>> >
>>> > ** **
>>> >
>>> > Error in ==> Source_Extraction_Exp2 at 43****
>>> >
>>> > Ds = spm_eeg_inv_extract(D);****
>>> >
>>> > ****
>>
>>> >
>>> > Specifying a radius of 5 gives exactly the same results as in the
>>> > gui,
>>
>>> > except that they are flipped.****
>>> >
>>> > ** **
>>
>>> >
>>> > Lastly, I would be great if you could advise what would be the most
>>> > reasonable approach to extract source waveforms from the primary
>>> > auditory cortex? I would like to average across all vertices that
>>> > lie within A1.
>>> > Is
>>
>>> > there a possibility to use an image mask to do so? ****
>>> >
>>> > ****
>>
>>> >
>>> > I think a simpler way would be to just specify some centroid for A1
>>> > and radius large enough to include all of it. There is no
>>> > resolution in M/EEG to make fine distinctions between A1 and the
>>> > adjacent areas especially when
>>> > A1 is strongly activated. So I don't think that using a mask will
>>> > make a
>>
>>> > big difference. ****
>>> >
>>> > ** **
>>
>>> >
>>> > I had the same idea. But results change a lot with respect to the
> ‘seed’
>>> > region I take (i.e. [-40 -18 7] & [-41 -18 8]) and field intensities.
>>> > So
>>
>>> > how would you then decide on the seed region?****
>>> >
>>> > ** **
>>> >
>>> > Thanks a lot!****
>>> >
>>> > Anette****
>>> >
>>> > ** **
>>> >
>>> > Best,****
>>> >
>>> > ** **
>>> >
>>> > Vladimir****
>>> >
>>> > ** **
>>> >
>>> > ****
>>> >
>>> > Looking forward to your suggestions.****
>>> >
>>> > ****
>>> >
>>> > Anette****
>>> >
>>> > ****
>>> >
>>> > ****
>>> >
>>> > ** **
>>> >
>>
>>
>>
>>
>
|