[Mne_analysis] regarding morphing in matlab

[Pavan Ramkumar]

Hi Hari (and others),

Thanks very much for the suggestions and apologies for the lengthy reply!

I did check that most rows are entirely zero (about 2300/2562). Secondly,
I did look at the vertex numbering and they seem to make sense. Also, I
visualized the result as stc and they do seem as sparse as the data matrix
suggests.

It seems to me that smoothing (or as Matti recommends to call it in
Section 8.3 of the manual v2.6.1 : smudging/blurring) is the root of the
difference between the mne_make_movie result and my MATLAB implemenation.

For example, I checked that the mne_make_movie result with only a single
smoothing iteration (i.e. --smooth 1) is nearly as sparse. Specifically,
in the resulting stc file 2372/2562 used vertices in stc_l.data are zero.

Here are some snapshots from a a single time frame of a single subject stc
file along with the same file morphed to a common brain with 3 different
smoothing factors (1, 2 and 5). Matti recommends a smoothing factor
between 4-7.

https://neuro.hut.fi/~pavan/temp/smudging_subject01-lh-lat.png
https://neuro.hut.fi/~pavan/temp/smudging01_common-lh-lat.png
https://neuro.hut.fi/~pavan/temp/smudging02_common-lh-lat.png
https://neuro.hut.fi/~pavan/temp/smudging05_common-lh-lat.png

As you might guess, the sparseness of the respective data matrices
decreases with increased smoothing.

I would appreciate your (or anyone else's) comments on whether the
smoothing assessment above is right and if so, what would your
recommendations be regarding the choice of the smoothing factor. Please
note that I am subjecting my morphed data to a substantial analysis
pipeline and it is not a last stage visualization of group data as is
typically the case.

Also, any comments regarding implementation of the smoothing/blurring
operation in MATLAB are welcome. Dealing with a large "vertex x vertex"
distance matrix, as well as the choice of the neighborhood parameter 'N_j'
as defined in Section 8.3 of the v2.6.1 manual are not straightforward to
me.

Thanks again,
Pavan

> Hi Pavan,
>    I don't have much experience with morphing data in MATLAB. Given that,
> the procedure looks ok to me. Did you write any of your morphed data to
> an stc file to see what it looks like?
>
> There are a few possibilities that occur to me:
> (1) Are you sure 'most rows' of leftmapRelevant and rightmapRelevant are
> infact zero? They ought to be sparse with probably 2 or 3 elements
> non-zero in the entire row of 5000 (if that's your source space size) odd
> elements..
>
> (2) Depending on how you made the sample stc file that is in
> 'common_brain' space, it might have lost track of the vertex numbering.
> Could you look at stc_l.vertices and stc_r.vertices to see if they make
> sense? If they are simply [0:2561]' or [0:10241]' for instance, the
> procedure of course will give you mostly zeros.
>
> Hope it helps.
>
> Regards,
> Hari
>
> On Mon, March 14, 2011 7:00 am, Pavan Ramkumar wrote:
>> Dear MNE users,
>>
>> Just a follow up mail to my previous request. I would like to transform
>> my
>> data from one decimated source space to another from within matlab.
>> Below
>> are the steps I have adopted to do so. If anybody has previous
>> experience
>> in doing similar things, kindly share your thoughts.
>>
>> 1/ I precomputed the morph maps from each surface to the common brain
>> with
>> mne_make_morph_maps.
>>
>> 2/ I read in the morph maps using mne_read_morph_maps as follows:
>> [leftmap,rightmap] = mne_read_morph_map('my_brain', 'common_brain');
>>
>> 3/ I read in an inverse operator for the input surface 'my_brain' into
>> the
>> structure 'inv_op'
>>
>> 4/ I read in a sample stc file for the target surface i.e.
>> 'common_brain'
>> to get the vertex info. The stc files were read into 'stc_l' and 'stc_r'
>>
>> 5/ Next, I used the following lines of code to transform my data:
>> leftmapRelevant = leftmap(stc_l.vertices,inv_op.src(1).vertno);
>> rightmapRelevant = rightmap(stc_r.vertices,inv_op.src(2).vertno);
>> morphed_dataL = leftmapRelevant*dataL;
>> morphed_dataR = rightmapRelevant*dataR;
>>
>> Note that stc_l.vertices contains all used vertices for the left
>> hemishpere of the target surface (i.e. 'common_brain') and
>> inv_op.src(1).vertno contains all used vertices for the left hemisphere
>> of
>> the source surface (i.e. 'my_brain'). The same is true for
>> stc_r.vertices
>> and inv_op.src(2).vertno respectively.
>>
>> Intuition suggests that these steps would suffice, but I did some checks
>> and figured that most rows of leftmapRelevant and rightmapRelevant are
>> entirely zero.
>>
>> I am using a morphgrade of 4 i.e. target surface consists of 2562 used
>> vertices per hemisphere. By comparison, the source surfaces are computed
>> with 5mm spacing and therefore consist of about 5000 used vertices per
>> hemisphere.
>>
>> What am I missing?
>>
>> Thank you for your time.
>> Best regards,
>> Pavan
>>
>>> Dear MNE users,
>>>
>>> I would like to morph my data from a single subject into a common
>>> brain.
>>> However, instead of writing the result out as an stc file (which takes
>>> up
>>> too much space) I would like to do the morphing from within matlab, and
>>> then subsequently do some postprocessing on the morphed data before I
>>> write out stc files.
>>>
>>> Does anybody know where to find the vertex numbers corresponding to the
>>> --morphgrade parameter used in mne_make_movie? Specifically, for
>>> morphgrade = 5, where do I find the indices of the 10242 vertices that
>>> are
>>> selectively stored into the stc file by mne_make_movie? As far as I
>>> looked, the manual does not explicitly give this information.
>>>
>>> Please advice.
>>>
>>> Thanks in advance,
>>> Pavan
>>>
>>
>> _______________________________________________
>> Mne_analysis mailing list
>> Mne_analysis at nmr.mgh.harvard.edu
>> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/mne_analysis
>>
>>
>>
>
>
> --
> Hari Bharadwaj
>
>
> The information in this e-mail is intended only for the person to whom it
> is
> addressed. If you believe this e-mail was sent to you in error and the
> e-mail
> contains patient information, please contact the Partners Compliance
> HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to you in
> error
> but does not contain patient information, please contact the sender and
> properly
> dispose of the e-mail.
>
>