[FieldTrip] ft_clusterstat OUT OF MEMORY - DICS

Martin Winkels m_wink10 at uni-muenster.de
Tue Feb 24 22:57:00 CET 2015


Hey Julian,

thanks for the answer.

We are using some sort of an Intel iCore i7 with 16 GB of RAM as well as 40
GB of swap and Fedora 16.

We do interpolate the data on an MRI. In fact I'm not sure if that is the
source of the problem. We downsampled the data and it did not change
anything. The problem seems to be that there is a number generated that is
too big for MATLAB to process it with the zeros(x) instruction.

Around 1-2 years ago I did nearly the same thing in fieldtrip but including
an LCMV-Beamformer, the resulting data structures where much bigger and it
worked without a problem.

Thanks, Martin

On Mon, Feb 23, 2015 at 5:14 PM, Julian Keil <julian.keil at gmail.com> wrote:

> Dear Martin,
>
> what kind of machine are you using?
> Did you interpolate your data to an MRI?
> What is your grid resolution?
>
> You have quite a high number of grid points that you want to compare.
> So in case you run out of memory, I'd suggest not interpolating to an MRI
> (in case you have done this) but to stay on the grid-point level for your
> stats. Otherwise, you could use a less dense grid which obviously results
> in smaller data structures.
>
> Good luck,
>
> Julian
>
> ********************
> *Dr. Julian Keil*
>
> AG Multisensorische Integration
> Psychiatrische Universitätsklinik
> der Charité im St. Hedwig-Krankenhaus
> Große Hamburger Straße 5-11, Raum E 307
> 10115 Berlin
>
> Telefon: +49-30-2311-1879
> Fax:        +49-30-2311-2209
>
> http://psy-ccm.charite.de/forschung/bildgebung/ag_multisensorische_integration
>
> Am 23.02.2015 um 17:00 schrieb Martin Winkels:
>
> Dear Fieldtrippers,
>
> we encountered a problem during our DICS Beamformer-Statistics.
>
> After calculating a beamformer (DICS), normalisation and building
> grandaverages across subjects (here exemplarily 3 subjects) we try to
> calculate cluster based permutation statistic (in this study: between
> groups - one condition).
>
> The code we used is as follows:
>
>         cfg = [];
>
>         cfg.method           = 'montecarlo';
>         %cfg.statistic        = 'depsamplesT';
>         cfg.statistic        = 'ft_statfun_indepsamplesT';
>         cfg.correctm         = 'cluster';
>         cfg.clusteralpha     = 0.05; %ft default 0.05;
>         cfg.clusterstatistic = 'maxsum';
>         cfg.minnbchan        = 2;
>         cfg.tail             = 0;
>         cfg.clustertail      = 0;
>         cfg.alpha            = 0.025; %ft hat hier 0,025
>
>         cfg.parameter   = 'pow';
>         cfg.dim         =  grandavgA.dim;
>
>         cfg.numrandomization = 1;      % number of draws from the
> permutation distribution
>
>         design(1,:) = [1 1 1 2 2 2];
>         design(2,:) = [1 1 1 1 1 1];
>
>         cfg.design = design;
>        cfg.ivar   = 1;
>
>         stat = ft_sourcestatistics(cfg, grandavgA, grandavgB);
>
>
>
> The input data structure (grandavgA, grandavgB) is as follows:
>
> grandavgA =
>
>        pow: [3x116380 double]
>        dim: [46 55 46]
>     inside: [116380x1 logical]
>        pos: [116380x3 double]
>        cfg: [1x1 struct]
>
> grandavgB =
>
>        pow: [3x116380 double]
>        dim: [46 55 46]
>     inside: [116380x1 logical]
>        pos: [116380x3 double]
>        cfg: [1x1 struct]
>
>
> Fieldtrip version: current (02/23/2015)
>
>
> Thanks, Martin
>
> --
>
> M.Sc. Martin Winkels
>
> Universitätsklinikum Münster
>
> Institut für Biomagnetismus & Biosignalanalyse
>
> Malmedyweg 15
>
> 48149 Münster
>
> GERMANY
>
>
> Telefon: +49 251 83 56 846
> Web: http://biomag.uni-muenster.de
>
> _______________________________________________
> 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20150224/df1a249f/attachment-0001.html>


More information about the fieldtrip mailing list