[FieldTrip] Strange DICS Results

[Alexander Nakhnikian]

Hi Jan-Mathijs,

Thank you for your reply.  Since I posted my original question on the 
forum, I tried applying eLORETA to the same data set and the 
sensor-level power localizes to the locations we expect given the scalp 
topography.  Would the issues you discuss have the same effect on both 
localization methods or would beamformers be more susceptible to errors 
in the forward model?

Best,

Alexander

On 6/10/16 5:27 AM, Schoffelen, J.M. (Jan Mathijs) wrote:
> Hi Alex,
>
>  From what you describe I think that your results are suffering from numerical issues in the forward model. Once a source location (grid point) is close to the innermost boundary of the BEM, funky things may happen with respect to the computed leadfield, which obviously has large consequences for the shape of the beamformer spatial filter at that location. Best thing would be verify that indeed your ‘absurd’ voxels are near the edges, and mask them out.
> As a side note, typically volumetric DICS results in isolation (i.e using a single condition) are hard to interpret (even when using normalization).
>
> Best,
> Jan-Mathijs
>
>> On 04 Jun 2016, at 00:03, Alexander Nakhnikian <alexander_nakhnikian at hms.harvard.edu> wrote:
>>
>> Hello all,
>>
>> I'm hitting a wall in my analysis trying to source-localize EEG data in a visual task.  I'm trying to localize the 8 to 15 Hz power.
>>
>> Unfortunately, I don't have individual electrode coordinates or sMRIs so I've had to create an electrode coordinate file and manually align it with the scalp in FT's template head.  I'm using the standard BEM as the head model and the standard MRI instead of individual structural scans.
>>
>> I'm able to get reasonable results from ft_freqanalysis; importantly, they match the scalp topography we see using other programs, such as IDL.
>>
>> The problem is that when I try to localize these data with DICS, I get absurd results.  The source localized power exceeds 3000 at some locations and the maxima are in strange places, such as behind the eye socket and near the base of the skull.
>>
>> When I run ft_sourceinterpolate I get the following warning:
>>
>> Warning: could not reshape freq to the expected dimensions
>>> In ft_datatype_volume (line 136)
>>   In ft_checkdata (line 329)
>>   In ft_sourceinterpolate (line 163)
>>
>> Otherwise, no errors or warning messages come up.  I'm not sure what's causing the odd results.  Am I not using the template files correctly?  I've included analysis setup below.
>>
>> Thanks in advance,
>>
>> Alexander
>>
>>
>>
>> cfg = [];
>> cfg.method = 'mtmfft';
>> cfg.output = 'powandcsd';
>> cfg.tapsmofrq = 4;
>> cfg.foilim = [12 12];
>> freqData = ft_freqanalysis(cfg,data);
>>
>> cfg = [];
>> cfg.grid.xgrid = 'auto'; cfg.grid.ygrid = 'auto';cfg.grid.zgrid = 'auto';
>> cfg.resolution = 5; %units are mm
>> cfg.method = 'dics';
>> cfg.dics.normalize = 'yes';
>> cfg.headmodel = vol; %from loading 'standard_bem.mat'
>> cfg.elec = elec_aligned; %electrode coordinates (MNI) aligned to template head
>> sources = ft_sourceanalysis(cfg,freqData);
>>
>> cfg = [];
>> cfg.downsample = 2;
>> cfg.parameter = 'pow';
>> sourcesInt = ft_sourceinterpolate(cfg,sources,mri); %mri from 'standard_mri.mat' in FT's templates
>>
>> cfg = [];
>> cfg.method = 'slice';
>> cfg.parameter = 'pow;
>> ft_sourceplot(cfg,sourcesInt)
>>
>>
>>
>>
>>
>> -- 
>> Alexander Nakhnikian, Ph.D.
>> Postdoctoral Fellow
>> Clinical Research Training Program
>> VA Boston Healthcare System
>> Harvard Medical School
>>
>> _______________________________________________
>> 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

-- 
Alexander Nakhnikian, Ph.D.
Postdoctoral Fellow
Clinical Research Training Program
VA Boston Healthcare System
Harvard Medical School