[FieldTrip] Cortical Sheet for MNE source reconstruction

Thu Sep 13 15:07:08 CEST 2018

Hi Annika,

It is difficult to predict, what will happen, if source points are 
outside the assumed source region.

In some BEM implementations, it is explicitly assumed that sources are 
inside the skull. If they are, however, outside, the forward solution is 
totally wrong. If the BEM is implemented so that it allows sources 
outside the brain, the results for such sources do not make sense 
physiologically.

So, in any case, before doing MNE or scanning with dipoles, you should 
make sure that source space is inside the skull (for FieldTrip BEMS and 
local-spheres model), and none of the source points are "too close". If 
I remember right, FieldTrip should automatically check that. If it 
doesn't, the user needs to do that... (if a source point happens to be 
in scalp, a lot of activity will project there).

Cheers,
  Matti

On 2018-09-13 14:35, Annika Thierfelder wrote:
> Hi Jan-Mathijs,
> 
> Thank you very much for your response! There is quite some story behind 
> the reason why I extrapolated it to the 3D grid, but I'll try to explain 
> shortly so you can also see how both my emails relate.
> 
> When I let the MNE algorithm run for the first time, I noticed that 
> there was a consistent activity overshoot at always the same voxel in 
> all my conditions. So I checked the code and I tracked it down to the 
> part in the sourcemodel outside of the headmodel, as described in the 
> other email. I shifted the sourcemodel a little bit upwards manually and 
> got rid of the artefact (although that might not be the best solution).
> 
> However, I realized that the misalignment in the source- and the 
> headmodel is, as you also said, in the left temporal cortex. But my 
> artifacts were in the right temporal cortex. Also, the activation did 
> not seem to be on the side where we expected it to be.
> 
> So, I found out that this might be due to the coordinate systems, so I 
> wanted to make sure if it is the same if I plot it on the MRI. And this 
> then looked like you saw in the image, so I got very confused if there 
> was something essentially wrong in the whole pipeline I used. It is very 
> reassuring to hear that this probably is not the case. Is there any way 
> to check this?
> 
> So, as a final question: Is it probable that the source activity is 
> mirrored in the surface sourceplots?
> 
> Best and thanks again for the help,
> 
> Annika
> 
> 
> On 9/13/2018 12:13 PM, Schoffelen, J.M. (Jan Mathijs) wrote:
>> Hi Annika,
>>
>> Part of what you are looking at is an extrapolation artifact, which is 
>> caused by ft_sourceinterpolate. Is there a specific reason why you 
>> want the data visualized on a 3D grid, rather than as the original 
>> surface? Using ft_sourceplot (or ft_sourcemovie) should work with the 
>> original source data (provided it contains a description of the 
>> cortical mesh, in the fields pos and tri).
>>
>> Best wishes,
>> Jan-Mathijs
>>
>>
>>
>>
>>
>>> On 11 Sep 2018, at 15:04, Annika Thierfelder 
>>> <athierfelder at tuebingen.mpg.de> wrote:
>>>
>>> Dear everyone,
>>>
>>> I'm still encountering problems with my source analysis that I don't 
>>> know how to solve. When I try to interpolate the source back to the 
>>> MRI, it does not look at all the way it should (the sources seem to 
>>> explode outside of the brain). You can have a look at it here: 
>>> https://drive.google.com/file/d/1WZ2WYkKkI8FfyNHmCgk7OnbPLb6iCZYT/view?usp=sharing 
>>>
>>>
>>> I'm using the cortical sheet provided by fieldtrip 
>>> (cortex_5124.surf.gii), along with the standard MRI and BEM. When I 
>>> construct a source model from the BEM (which is not what I want since 
>>> it distributes the sources evenly spaced but I tried it for testing), 
>>> the sources are all neatly inside the brain, so I seem to be using 
>>> that cortical sheet in a wrong way. Do I have to process it further? 
>>> All tutorials I found just do it like this.
>>>
>>> Below you can find the processing pipeline I am currently using. Any 
>>> help is appreciated!
>>>
>>> Best and thank you,
>>>
>>> Annika
>>>
>>> ======================================================================================= 
>>>
>>>
>>> sourcemodel =  load('cortex_5124.surf.gii');
>>>
>>> % compute the leadfield
>>> leadfield = [];
>>> cfg = [];
>>> cfg.grid = sourcemodel;
>>> cfg.headmodel = headmodel;
>>> cfg.elec = chanlocs;
>>> leadfield = ft_prepare_leadfield(cfg);
>>>
>>> % I put the data through ft_timelockanalysis to have it in the right 
>>> format, although it doesn't do anything.
>>> % The data I work with is resting state data, that's why I don't 
>>> compute the noise covariance here.
>>> cfg = [];
>>> data_prepared = ft_timelockanalysis(cfg, data);
>>>
>>> % source reconstruction with MNE
>>> cfg = [];
>>> cfg.method = 'mne';
>>> cfg.grid = leadfield;
>>> cfg.headmodel = headmodel;
>>> cfg.mne.prewhiten = 'yes';
>>> cfg.mne.lambda = 0.1;
>>> cfg.mne.scalesourcecov = 'yes';
>>> cfg.elec = chanlocs;
>>> source = ft_sourceanalysis(cfg, data_prepared);
>>>
>>> % save into single structure to save memory
>>> source = ft_struct2single(source);
>>>
>>> % interpolate to the template MRI
>>> cfg = [];
>>> cfg.interpmethod = 'nearest';
>>> cfg.parameter = 'avg.pow';
>>> interp = ft_sourceinterpolate(cfg, source, mri);
>>>
>>> % and do the source plot
>>> cfg = [];
>>> cfg.funparameter = 'pow';
>>> cfg.method = 'ortho';
>>> ft_sourceplot(cfg, interp);
>>> On 9/7/2018 12:50 PM, Annika Thierfelder wrote:
>>>> Dear fieldtrippers,
>>>>
>>>> I have a short question about the template files provided by 
>>>> fieldtrip. I am doing MNE source reconstruction with the template 
>>>> BEM headmodel and the cortical sheets provided in the sourcemodel 
>>>> templates. However, I noticed that the sourcemodel is outside of the 
>>>> BEM brain surface at some points. So I plotted it against the MRI 
>>>> template and it doesn't really seem to fit. It seems quite small and 
>>>> shifted, so it's hard to judge the shape.
>>>>
>>>> Are those not produced from the same template MRI? And would it 
>>>> therefore make sense to go through the process of producing a 
>>>> sourcemodel from the standard MRI myself as described here 
>>>> [http://www.fieldtriptoolbox.org/tutorial/sourcemodel#construction_of_a_source_model_based_on_a_surface_description_of_the_cortical_sheet] 
>>>> ?
>>>>
>>>> Thanks in advance for your advice,
>>>>
>>>> Best,
>>>>
>>>> Annika
>>>>
>>>> _______________________________________________
>>>> fieldtrip mailing list
>>>> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>>> https://doi.org/10.1371/journal.pcbi.1002202
>>> _______________________________________________
>>> fieldtrip mailing list
>>> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>> https://doi.org/10.1371/journal.pcbi.1002202
>>
>> _______________________________________________
>> fieldtrip mailing list
>> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>> https://doi.org/10.1371/journal.pcbi.1002202
> 
> _______________________________________________
> fieldtrip mailing list
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> https://doi.org/10.1371/journal.pcbi.1002202