AW: [FIELDTRIP] MNI volumenormalise sourceinterpolation coordinates.

jan-mathijs schoffelen j.schoffelen at PSY.GLA.AC.UK
Thu Jul 16 17:04:14 CEST 2009


Hi Hanneke,

source.pos should contain the MNI-based coordinates before calling  
sourceinterpolate; I suspect that your single subject source data  
contains the original (neuromag-space) positions. This is of course  
how it should be for the sourceanalysis, because there the grid  
points have to be defined in neuromag space. Yet, for the  
interpolation this is wrong, because sourceinterpolate assumes that  
the coordinate frames of both mri and functional volume are the same.  
Because the neuromag coordinates are shifted with respect to the MNI- 
coordinates, this still could explain the discrepancy. I suggest to  
replace sd.pos with template_grid.pos before calling sourceinterpolate.

Best,

Jan-Mathijs

On Jul 16, 2009, at 3:21 PM, Hanneke Van Dijk wrote:

> Hi Jan-Mathijs,
>
> Thanks for your suggestions and your time, but I haven't been able  
> to fix it....
>
> I agree that this work-around shouldn't be neccesary but I guess we  
> first have to know what to fix....
>
> You suggested to use the template_mri in MNI coordinates to  
> interpolate to that seemed simple so I tried it immediately,  
> however it didn't help. It didn't change a thing.
>
> So your other suggestion about the origin being different should be  
> looked at, but this raised a few questions..
> First, I thought because I had realigned the mri to ctf coordinates  
> using volumerealign, the NM coordinates wouldn't matter anymore.
> Second, the headmodel I have created using the realigned, segmented  
> and flipped mri looks very good to me relative to the gradiometer  
> locations and the grid as well.
> (see http://fieldtrip.fcdonders.nl/example/ 
> read_neuromag_mri_and_create_single- 
> subject_grids_in_individual_head_space_that_are_all_aligned_in_mni_spa 
> ce). Additionally, the alpha source that I found using this  
> headmodel and looking at the grid only (without the interpolation)  
> looked believable as well.
>
> I have also segmentated the volumenormalised mri, which actually  
> makes the later produced single_shell headmodel go downward and  
> posterior as well. While the grid stays the same and  
> 'goodlooking' ;-).
> So I am inclined to think that something strange is happening  
> during the volumenormalise or sourceinterpolation procedure moving  
> the mri slightly downward and more posterior. I would fix it myself  
> but I don't know where I should look... it also puzzles me why this  
> only happens 'for from NM- to ctf coordinates realigned' mri's. It  
> seems that fieldtrip remembers that the mri was originally a NM...
>
> Hope you or someone else can help me further...
>
> Groetjes Hanneke
>
>
> -----Urspr√ľngliche Nachricht-----
> Von: FieldTrip discussion list im Auftrag von jan-mathijs schoffelen
> Gesendet: Mo 13.07.2009 10:42
> An: FIELDTRIP at NIC.SURFNET.NL
> Betreff: Re: [FIELDTRIP] sourceinterpolation coordinates.
>
> Hi Hanneke,
>
> This looks like a difference in origin to me. Note that the MNI
> coordinate system has the cardinal axes pointing in the same
> directions as the neuromag headcoordinate system, but that the origin
> is different. The origin of the NM-headcoordinate system is on the
> interauricular line, the point where the y-axis hits it. This is
> approximately somewhere close or in the pons. The MNI origin is the
> AC, which is somewhere 'higher' in the volume. I suspect that the
> preparation of the grids did not work out without flaws.
> Alternatively, which I think is the most likely explanation, is that
> you should use the MNI-template as an MRI for the interpolation,
> rather than a normalized single subject MRI. I don't know about the
> single subject MRIs, but at least the MNI-template is explicitly in
> MNI-space, whereas the normalized MRI could be in head-space. This
> could explain the shift.
> Anyway, if the latter hunch is the explanation, then the
> sourceanalysis results might be trusted, if the first hunch was the
> one, the sourceanalysis results are most certainly wrong.
>
> As a matter of fact the whole swapping around with coordinates should
> not be necessary if the basic fieldtrip-code is adjusted such that it
> can handle neuromag-based headcoordinate systems in an appropriate
> way. This relates to an earlier remark on the list, which rightfully
> states that the approach you (we) have been taken so far seems a bit
> tedious. Ideally, in other words, the middle man has to be taken out,
> which in this case is the intermediate transformation to a ctf-based
> coordinate system.
>
> Hope this helps,
>
> JM
>
>
> Quoting Hanneke van Dijk <Hanneke.vanDijk at MED.UNI-DUESSELDORF.DE>:
>
>> Now there is an attachment too! sorry
>>
>> On Fri, Jul 10, 2009 at 4:08 PM, Hanneke Van Dijk <
>> Hanneke.vanDijk at med.uni-duesseldorf.de> wrote:
>>
>>> Dear all,
>>>
>>> I have entered the next step: sourceanalysis. But I have encountered
>>> another problem, I think in sourceinterpolation or maybe  
>>> volumenormalise.
>>>
>>> After running the script as I have described before (to make the  
>>> headmodel
>>> and the grid) I do the following;
>>>
>>> load f %load frequency data
>>> cfg                = [];
>>> cfg.grid           = grid;
>>> cfg.frequency      = 10;
>>> cfg.vol            = hdm;
>>> cfg.grad           = f.grad;
>>> cfg.projectnoise   = 'yes';
>>> cfg.lambda         = 5e-28;
>>> cfg.method         = 'dics';
>>> cfg.feedback       = 'textbar';
>>>
>>> source = sourceanalysis(cfg,f);
>>>
>>> then:
>>>
>>> sd = sourcedescriptives([],source);
>>>
>>> then:
>>>
>>> load norm %normalized to MNI brain of single subject
>>> sdint = sourceinterpolate([], sd , norm);
>>> clear norm
>>> %plotting the results
>>> cfg.funparameter = 'avg.nai';
>>> cfg.anaparameter = 'anatomy';
>>> cfg.funcolormap  = 'jet';
>>> cfg.method       = 'ortho';
>>> cfg.projmethod   = 'nearest';
>>> cfg.sphereradius = 20;
>>> cfg.surffile     = 'surface_l4_both';
>>> figure;
>>> sourceplot(cfg,sdint)
>>>
>>> The result is attached to this email, the gird seems to have  
>>> moved upwards
>>> and somewhat to the front relative to the normalised mri. But it  
>>> seems to
>>> have the right size. I have been playing around with sourceunits  
>>> mriunits
>>> x-y coordinates but nothing helps, or it makes it even worse.
>>>
>>> Does anybody have an idea what's happening?
>>>
>>> Yours,
>>> Hanneke
>>>
>>> www.linkedin.com/in/hannekevandijk
>>> tel. +49 (0)211 81 13074
>>>
>>>
>>>
>>> ----------------------------------
>>> The aim of this list is to facilitate the discussion between  
>>> users of the
>>> FieldTrip  toolbox, to share experiences and to discuss new ideas  
>>> for MEG
>>> and EEG analysis. See also
>>> http://listserv.surfnet.nl/archives/fieldtrip.html and
>>> http://www.ru.nl/neuroimaging/fieldtrip.
>>>
>>
>> ----------------------------------
>> The aim of this list is to facilitate the discussion between users
>> of the FieldTrip  toolbox, to share experiences and to discuss new
>> ideas for MEG and EEG analysis. See also
>> http://listserv.surfnet.nl/archives/fieldtrip.html and
>> http://www.ru.nl/neuroimaging/fieldtrip.
>>
>
>
>
> ------------------------------------------------------------------
> The University of Glasgow, Department of Psychology WebMail system
> ------------------------------------------------------------------
>
> ----------------------------------
> The aim of this list is to facilitate the discussion between users  
> of the FieldTrip  toolbox, to share experiences and to discuss new  
> ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/ 
> archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.
>
>
> ----------------------------------
> The aim of this list is to facilitate the discussion between users  
> of the FieldTrip  toolbox, to share experiences and to discuss new  
> ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/ 
> archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.



More information about the fieldtrip mailing list