[FieldTrip] LCMV beamformer analysis: uncertainties

[Anette Giani]

Dear Fieldtrip users,

I am currently trying to localize averaged time courses of condion1 and
condition2 in 20 subjects using LCMV beamformer and to compare source
activity of both conditions statistically. However, results look suspicious
to me and I am wondering if I am doing things correctly. It will be of great
help if you could comment on some issues and check the parameters I am
using.

Things I am wondering about:

(1) doing cluster statistics I get 1 big cluster that is highly significant
(clusterstat: 6.5829e+006, prob: 0.0020) and includes all!! (but one) voxels
that are labeled as "inside". I tried to decrease the alpha level to 0.01,
but it didn't really help.

(2) How can I check the quality of alignment of the individual anatomicals?
So far I am just using ft_volumenormalise, hoping that alignment works well.
However, I am getting suspicious by the warning:

 

1377485 voxels are inside the brain of all subjects

2557673 voxels are inside the brain of some, but not all subjects

4551464 voxels are outside the brain of all subjects

Warning: marking only voxels inside the brain of all subjects as 'inside'

 

If MRI's would be aligned, why aren't inside voxels of all subjects
identical?  How does the fact that voxels are not exactly aligned affect my
analysis?

 

Any comments & and ideas are welcome

Thank you so much for your help!!

Anette

 

--

PARAMETERS

%% Calculating the cross spectral density matrix

cfg                  = [];

cfg.covariance       = 'yes';

cfg.channel          = 'MEG';

cfg.removemean       = 'no';

cfg.trials           = cond1;

cfg.covariancewindow = [0.15 0.25];

tlckav_cond1     = ft_timelockanalysis(cfg, OrigSPMdata); 

    

cfg.trials           = cond2;

cfg.covariancewindow = [0.15 0.25];

tlckav_cond2     = ft_timelockanalysis(cfg, OrigSPMdata); 

    

cfg.trials           = [cond1 cond2];

cfg.covariancewindow = [0.15 0.25];

tlckav_combined   = ft_timelockanalysis(cfg, OrigSPMdata); 

 

%% lcmv beamformer using common filters

cfg        = [];

cfg.method = 'lcmv';

cfg.lambda = '5%';

cfg.grid   = grid;

cfg.grad   = grad;

cfg.vol    = vol;

cfg.lcmv.projectnoise = 'yes'; % necessary?

cfg.lcmv.lambda       = '5%'; 

cfg.lcmv.keepfilter   = 'yes'; 

cfg.lcmv.realfilter   = 'yes'; 

sourceAll = ft_sourceanalysis(cfg, tlckav_combined)

 

cfg.grid.filter = sourceAll.avg.filter;

source_cond1  = ft_sourceanalysis(cfg, tlckav_cond1);

source_cond2 = ft_sourceanalysis(cfg, tlckav_cond2);

 

%% reslice

mri_re = ft_volumereslice([], mri2); 

source_cond1.anatomy = mri_re.anatomy;

source_cond2.anatomy = mri_re.anatomy;

    

    

% interpolate

cfg            = [];

cfg.units      = 'mm';

cfg.downsample = 2; % taken from tutorial

cfg.parameter  = 'avg.pow';

source_cond1_Int = ft_sourceinterpolate(cfg, source_cond1 , mri_re);

source_cond2_Int = ft_sourceinterpolate(cfg, source_cond2 , mri_re);

    

% normalise

cfg = [];

cfg.coordsys      = 'ctf';

cfg.nonlinear     = 'no';

source_cond1_Int_Norm = ft_volumenormalise(cfg, source_cond1_Int);

source_cond2_Int_Norm = ft_volumenormalise(cfg, source_cond2_Int);

 

 

 

--------------------------- STATISTICS
--------------------------------------------------------------------------

% merge all subject's data 

cfg = [];

cfg.keepindividual = 'yes';

grandavg_cond1 = ft_sourcegrandaverage(cfg, sSub6_cond1, sSub8_cond1, . tbc

grandavg_cond2 = ft_sourcegrandaverage(cfg, sSub6_cond2, sSub8_cond2, . tbc

 

% -------------------------------------------------------------------------

% run statistics over subjects 

cfg = [];

cfg.dim         = grandavg_cond1.dim;

cfg.method      = 'montecarlo';

cfg.statistic   = 'depsamplesT';

cfg.parameter   = 'pow'; 

cfg.correctm    = 'cluster';

cfg.numrandomization = 500; 

cfg.tail        = 0;

cfg.alpha       = 0.05;

cfg.clusteralpha = 0.05;

cfg.correcttail = 'alpha';

nsubj=length(grandavg_cond1.trial);

cfg.design(1,:) = [1:nsubj 1:nsubj];

cfg.design(2,:) = [ones(1,nsubj) ones(1,nsubj)*2];

cfg.uvar        = 1; 

cfg.ivar        = 2; 

 

stat = ft_sourcestatistics(cfg, grandavg_cond1, grandavg_cond2);

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20120614/e678ced9/attachment-0001.html>