[FieldTrip] BEMCP

Thu Mar 17 15:21:26 CET 2016

What about correlations between local spheres and the FEM? If it's only the
BEM that's broken I'd expect that you should get decent values for those.
Your plots don't look so good. I did something similar a couple of years
ago and got much better results. Perhaps it'd be a good idea to re-check
the BEM again.

Vladimir

On Thu, Mar 17, 2016 at 2:08 PM, Debora Desideri <deb.desideri at gmail.com>
wrote:

> Dear Vladimir and John,
> attached you can find the results of the source localizaiton with the FEM
> head model. Moreover, I also report the correlation of the lead fields for
> the FEM and concentric spheres models and BEM and concentric spheres model
> (the latter one using an inwardshift of 6 mm). Unfortunately, I could not
> produce the correlation trend we expected (I tryed several values of
> inwardshifts, ranging form 1mm to 10 mm and the plot is of course not
> identical to the one I posted, but it is qualitatively similar).
> There must still be something I don't fully get probably, since the
> correlation between the concentric spheres lead fields and the BEM lead
> fields is always so low, even close to the center of the head, regardless
> of the inwardshift value.
>
> All considered, I would rather go for a FEM model at least for the moment
> being.
> Thanks a lot for your support!
>
> Best,
> Debora
>
>
>
>
> On Mon, Mar 14, 2016 at 12:11 PM, Vladimir Litvak <
> litvak.vladimir at gmail.com> wrote:
>
>> Dear Debora,
>>
>> cfg.moveinward option is useful when your source space is a mesh as it
>> actually deforms the mesh to push it inwards. When your source space is a
>> grid you should use cfg.inwardshift to just exclude the points that are
>> closer than 6 mm to the boundary.  It might be a useful exercise for you to
>> compute the lead fields with concentric spheres and with a BEM for the same
>> grid and then look at correlations between these lead fields. You should
>> see that the correlations should be high at depth but might be low for some
>> points close to the surface. Once you hit the right value of inwardshift
>> the low correlations should disappear.
>>
>> Best,
>>
>> Vladimir
>>
>> On Mon, Mar 14, 2016 at 10:46 AM, Debora Desideri <deb.desideri at gmail.com
>> > wrote:
>>
>>> Dear Vladimir,
>>> thanks for your explanation. I tryed to use the option cfg.moveinward in
>>> ft_prepare_sourcemodel ( I set it to 6 mm,, with inwardshift = 0), but this
>>> did not solve the issue. As you can see from the attachment the cortex is
>>> now partially uncovered. I am not sure I am using this option properly,
>>> maybe there is a combination of values for cfg.moveinward and
>>> cfg.inwardshift that works fine. I tryed several, but none of them gave me
>>> the source localization that I expect.
>>> I don't know whether the fact that some dipoles fall outside the brain
>>> matters. For the attached figures I had no dipoles outside the brain,
>>> whilst for some of the others attempts I had.
>>>
>>> I actually found this post on the mailinglist where both you and John
>>> already pointed out this issue using BEM:
>>>
>>>
>>> http://mailman.science.ru.nl/pipermail/fieldtrip/2015-November/009783.html
>>>
>>> Also I don't know whether this link (still under
>>> construction/discussion) may be of interest in this contex:
>>>
>>>
>>> http://www.fieldtriptoolbox.org/faq/is_it_good_or_bad_to_have_dipole_locations_outside_of_the_brain_for_which_the_source_reconstruction_is_computed
>>>
>>> Furthermore, this other page (also under construction), uses the option
>>> cfg.inwardshift for the template grid to be used in the creation of the
>>> individual subject's grid, but without an explanation of why the option
>>> is/should be used:
>>>
>>>
>>> http://www.fieldtriptoolbox.org/tutorial/sourcemodel#subject-specific_grids_that_are_equivalent_across_subjects_in_normalized_space
>>>
>>> In ft_prepare_sourcemodel I am using the standard grid (6 mm resolution)
>>> provided by Fiedltrip, I don't know whether craeting my own template_grid
>>> using the cfg.inwardshift option would actually fix the problem. I would
>>> like to perform some further attempts, as soon as I have some time to go
>>> through the function and understand how this options actually interplay.
>>>
>>> Meanwhile, I am still working on producing the results using the FEM
>>> head model. It is much more time-consuming than the BEM, and maybe that's
>>> one of the reasons to choose BEM models. As soon as they are ready I will
>>> post them as well :)
>>>
>>> Anyway, thanks so much for your help. It was actually quite useful and
>>> instructive!
>>>
>>> All the best wishes,
>>> Debora
>>>
>>>
>>>
>>> On Sat, Mar 12, 2016 at 5:14 PM, Vladimir Litvak <
>>> litvak.vladimir at gmail.com> wrote:
>>>
>>>> Dear Debora,
>>>>
>>>> The critical thing with BEMCP is that your sources are not too close to
>>>> the inner skull boundary. When they are closer than the safe distance the
>>>> solution you get is completely wrong. When they are not too close, the
>>>> solution is OK and comparable at least to concentric spheres. The safe
>>>> distance is is more or less the average edge length in your  skull mesh. In
>>>> SPM meshes this is 6 mm. In ft_prepare_sourcemodel there is an option
>>>> cfg.moveinward that corrects the mesh to ensure that it is not too close.
>>>> It would be interesting if you use this option and report whether this
>>>> resolves your problem.
>>>>
>>>> This issue should occur in any BEM but depending on the implementation
>>>> the critical distance might be shorter.
>>>>
>>>> Best,
>>>>
>>>> Vladimir
>>>>
>>>> On Sat, Mar 12, 2016 at 3:30 PM, Debora Desideri <
>>>> deb.desideri at gmail.com> wrote:
>>>>
>>>>> Dear John,
>>>>> thanks a lot for your promt reply and your suggestion. I guess that I
>>>>> was sticking to the BEM head model just because I read the tutorial for EEG
>>>>> BEM head model first, but going quickly through the studies you have
>>>>> mentioned and considering the tests you have performed, I may also consider
>>>>> to switch to the FEM head model or to simply use the concentric spheres
>>>>> (which appeared to work similar to BEMCP in your comparisons). I am
>>>>> relatively new to source localization and had no crash/errors in my
>>>>> pipeline, and this combination makes always difficult to find possible
>>>>> pitfalls.
>>>>>
>>>>> As soon as I have them ready, I will post the results obtained with
>>>>> the FEM head model as well, maybe they will be useful for some other users
>>>>> :)
>>>>>
>>>>> Thanks again for your support!
>>>>> All the best,
>>>>> Debora
>>>>>
>>>>> On Sat, Mar 12, 2016 at 3:42 PM, RICHARDS, JOHN <
>>>>> RICHARDS at mailbox.sc.edu> wrote:
>>>>>
>>>>>> Deb:
>>>>>>
>>>>>> I have had problems with BEMCP.
>>>>>>
>>>>>>  1—one of the practical / technical problems is that if the
>>>>>> compartments overlap, or even come close, then the procedure would not
>>>>>> work—the ft_sourceanlaysis crashed.
>>>>>>
>>>>>>  2—There are a couple of studies comparing BEMCP and dipoli, among
>>>>>> other methods.  The BEMCP method does not fare well, the dipoli fares
>>>>>> better,
>>>>>>
>>>>>> E. g., among three or four studies I have seen.
>>>>>> Gramfort, A., et al. (2010). "OpenMEEG: opensource software for
>>>>>> quasistatic bioelectromagnetics." Biomed Eng Online *9*: 45.
>>>>>>
>>>>>> Gramfort, A., et al. (2011). "Forward field computation with
>>>>>> OpenMEEG." Comput Intell Neurosci *2011*: 923703.
>>>>>>
>>>>>> 3—I have directly compared BEMCP, a 4-shell concentric spheres,
>>>>>> dipoli, and the Simbio FEM model, by correlating results from each model
>>>>>> (e.g., correlations of all the dipole CDR estimates; or correlations across
>>>>>> anatomical ROIs).  The BEMCP and concentric spheres work similarly, and the
>>>>>> dipoli and Simbio-FEM work similarly.   This was with infants using
>>>>>> participant-computed realistic models for the compartments and the FEM.
>>>>>>
>>>>>> I am not claiming from this and superiority of the dipoli (or FEM)
>>>>>> models over the BEMCP, but I suggest looking at the research literature
>>>>>> formally comparing these methods before going much further with BEMCP.
>>>>>>
>>>>>> John
>>>>>>
>>>>>> ***********************************************
>>>>>> John E. Richards Carolina Distinguished Professor
>>>>>> Department of Psychology
>>>>>> University of South Carolina
>>>>>> Columbia, SC  29208
>>>>>> Dept Phone: 803 777 2079
>>>>>> Fax: 803 777 9558
>>>>>> Email: richards-john at sc.edu
>>>>>> HTTP: jerlab.psych.sc.edu
>>>>>> ***********************************************
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>>
>>>>> Debora Desideri, PhD Student
>>>>> BNP Lab - Neurology Department
>>>>> University Hospital Tübingen
>>>>> Eberhard Karls University Tübingen
>>>>> Hoppe-Seyler Str. 3
>>>>> D-72076 Tübingen
>>>>> Tel: +49 7071/29 80478
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> fieldtrip mailing list
>>>>> fieldtrip at donders.ru.nl
>>>>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> fieldtrip mailing list
>>>> fieldtrip at donders.ru.nl
>>>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>>>
>>>
>>>
>>>
>>> --
>>>
>>> Debora Desideri, PhD Student
>>> BNP Lab - Neurology Department
>>> University Hospital Tübingen
>>> Eberhard Karls University Tübingen
>>> Hoppe-Seyler Str. 3
>>> D-72076 Tübingen
>>> Tel: +49 7071/29 80478
>>>
>>>
>>> _______________________________________________
>>> fieldtrip mailing list
>>> fieldtrip at donders.ru.nl
>>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>>
>>
>>
>> _______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl
>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>
>
>
>
> --
>
> Debora Desideri, PhD Student
> BNP Lab - Neurology Department
> University Hospital Tübingen
> Eberhard Karls University Tübingen
> Hoppe-Seyler Str. 3
> D-72076 Tübingen
> Tel: +49 7071/29 80478
>
>
> _______________________________________________
> fieldtrip mailing list
> 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/20160317/d503d87c/attachment-0002.html>