[FieldTrip] spatial and temporal ICA after beamformer

Tyler Grummett tyler.grummett at flinders.edu.au
Sun Apr 6 07:43:25 CEST 2014


Hi Jorn,

I think the fundamental issue with this is that the spatial filter has three dimensions, each corresponding to either x y or z.

I must admit I am confused at this point, because I dont know whether to choose x, y or z or somehow combine them or have them as additional channels.

>> source_inside.avg.filter

ans = 

    [3x123 double]
    [3x123 double]
    [3x123 double]
    [3x123 double]
    [3x123 double]
    [3x123 double]
    [3x123 double]
    [3x123 double]
    [3x123 double]
    [3x123 double]
etc....

Currently I am chosing only the first dimension ie source_inside.avg.filter{ 1}( 1, :) and then multiplying that by the data.

            for j = 1:numel( source_inside.avg.filter),
                temp{ j} = source_inside.avg.filter{ j}( 1, :) * data.trial{ 1}(chans,:);
                virtual_chans.trial( j, :) = temp{ j};
            end

to end up with:

virtual_chans = 

    label: {1x642 cell}
     time: {[1x19767 double]}
    trial: [642x19767 double]

Like I said, I dont know what to with the 2nd and 3rd dimension filters.

Tyler

*************************

Tyler Grummett ( BBSc, BSc(Hons I))
PhD Candidate
Brain Signals Laboratory
Flinders University
Rm 5A301
Ext 66124

________________________________________
From: fieldtrip-bounces at science.ru.nl <fieldtrip-bounces at science.ru.nl> on behalf of "Jörn M. Horschig" <jm.horschig at donders.ru.nl>
Sent: Saturday, 5 April 2014 7:21 PM
To: FieldTrip discussion list
Subject: Re: [FieldTrip] spatial and temporal ICA after beamformer

Hi Tyler,

> However as I have 642 source.avg.filter, which filter do I chose? Or
> do I calculate EEG channels for each voxel I landed with (I only had
> one trial).

Yep, you will (should) get 642 virtual channels, where each of these
virtual channels correspond to the activity at the grid-point you
specified when computing the spatial filter.


Best,
Jörn


On 4/4/2014 4:19 PM, Tyler Grummett wrote:
>
> Hello All,
>
>
> So once I have calculated the preprocessing data, timelock data,
> headmodel, lead field and source.
>
>
> I have the following code to create virtual EEG channels of data
> (similar to the tutorial):
>
> % find sensor names and indices
>             chans = ft_channelselection( 'EEG', data.label);
>             chans = match_str( data.label, chans);
>             % compute virtual channels
>             virtual_chans = [];
>             virtual_chans.label = data.label;
>             virtual_chans.time = data.time;
>             virtual_chans.trial = source.avg.filter{ 1} *
> timelock.trial{ 1}(chans,:);
>
> However as I have 642 source.avg.filter, which filter do I chose? Or
> do I calculate EEG channels for each voxel I landed with (I only had
> one trial).
>
> Btw, the source was calculated with the following code:
>
> **
> cfg            = [];
>             cfg.method   = 'lcmv';
>             cfg.grid.inside  = grid.inside;
>             cfg.grid.outside = [];
>             cfg.vol          = vol;
>             cfg.keepfilter = 'yes';
>             source = ft_sourceanalysis( cfg, timelock);
>
> Thankyou for all the help you have provided me thus far.
>
> tyler
>
>
>
>
>
> *************************
>
> /Tyler Grummett ( BBSc, BSc(Hons I))/
> /PhD Candidate/
> /Brain Signals Laboratory/
> /Flinders University/
> /Rm 5A301/
> /Ext 66124/
> ------------------------------------------------------------------------
> *From:* fieldtrip-bounces at science.ru.nl
> <fieldtrip-bounces at science.ru.nl> on behalf of jan-mathijs schoffelen
> <jan.schoffelen at donders.ru.nl>
> *Sent:* Friday, 4 April 2014 7:02 PM
> *To:* FieldTrip discussion list
> *Subject:* Re: [FieldTrip] spatial and temporal ICA after beamformer
> Hi all,
>
> Allow me to chime in here. As I understand it Tyler wants to do ICA on
> the output of an LCMV beamformer. This is in general allowed, and can
> be done if you know how to handle the data. In Fieldtrip there is no
> pre-cooked way of doing it. What you need is to fool
> ft_componentanalysis into believing that your source reconstructed
> data is of the type 'raw'. To get started, you can have a look at
> http://fieldtrip.fcdonders.nl/tutorial/shared/virtual_sensors. In this
> particular tutorial, only a few 'virtual channels' are created. You
> probably want many of those, defined on a 3D grid. The only constraint
> there is probably the amount of RAM on your computer. Note that you
> want to do (and will do) temporal ICA in this case. Also, note that
> you probably should pre-PCA your data (this will be a parameter in the
> specs for the options for the ICA algorithm you are going to use). If
> you want to apply the ICA algorithm on amplitude envelope timeseries
> of band-limited activity, rumour has it that the fastica method works
> quite well.
>
> Cheers,
> Jan-Mathijs
>
> On Apr 4, 2014, at 10:17 AM, Jörn M. Horschig wrote:
>
>> Hi Tyler,
>>
>> sorry, I realized too late that you want to use LCMV on ICA
>> components - mostly because your approach resulted in many different
>> erros ;). Afaik, it is not possible to use beamforming on ICA
>> components. On Wednesday we talked about this in our weekly FieldTrip
>> meeting and identified some occasions where FieldTrip is not handling
>> IC-data well. This can lead to unexpected results and missing
>> warnings or errors. However, you can use dipole modelling (like
>> EEGlab does) or MNE if you want to. These should work with ICs.
>>
>> Any reason why you want to use LCMV on ICs? In theory, both steps
>> constitute a spatial filter, so you can try to be smart and first to
>> LCMV on all your channel level data, and then apply the LCMV spatial
>> filter on the individual backprojected ICs. Somewhat a complicated
>> procedure, but I think it would work?
>>
>> Best,
>> Jörn
>>
>>
>>
>> On 4/4/2014 7:23 AM, Tyler Grummett wrote:
>>>
>>> Hello fieldtrip,
>>>
>>>
>>> I was wondering whether it was possible in the fieldtrip toolbox to
>>> run a spatial and temporal ICA on data after it has been run through
>>> ft_sourceanalysis (We can a LCMV beamformer on EEG data that had
>>> been timelocked).
>>>
>>>
>>> Currently it is saying that I need raw data for it to be processed.
>>>
>>>
>>> Tyler
>>>
>>>
>>> *************************
>>>
>>> /Tyler Grummett ( BBSc, BSc(Hons I))/
>>> /PhD Candidate/
>>> /Brain Signals Laboratory/
>>> /Flinders University/
>>> /Rm 5A301/
>>> /Ext 66124/
>>>
>>>
>>> _______________________________________________
>>> fieldtrip mailing list
>>> fieldtrip at donders.ru.nl <mailto:fieldtrip at donders.ru.nl>
>>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>>
>>
>> --
>> Jörn M. Horschig
>> PhD Student
>> Donders Institute for Brain, Cognition and Behaviour
>> Centre for Cognitive Neuroimaging
>> Radboud University Nijmegen
>> Neuronal Oscillations Group
>> FieldTrip Development Team
>>
>> P.O. Box 9101
>> NL-6500 HB Nijmegen
>> The Netherlands
>>
>> Contact:
>> E-Mail: jm.horschig at donders.ru.nl <mailto:jm.horschig at donders.ru.nl>
>> Tel:    +31-(0)24-36-68493
>> Web: http://www.ru.nl/donders
>>
>> Visiting address:
>> Trigon, room 2.30
>> Kapittelweg 29
>> NL-6525 EN Nijmegen
>> The Netherlands
>>
>> _______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl <mailto:fieldtrip at donders.ru.nl>
>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>
> Jan-Mathijs Schoffelen, MD PhD
>
> Donders Institute for Brain, Cognition and Behaviour,
> Centre for Cognitive Neuroimaging,
> Radboud University Nijmegen, The Netherlands
>
> Max Planck Institute for Psycholinguistics,
> Nijmegen, The Netherlands
>
> J.Schoffelen at donders.ru.nl <mailto:J.Schoffelen at donders.ru.nl>
> Telephone: +31-24-3614793
>
> http://www.hettaligebrein.nl
>
>
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip


--
Jörn M. Horschig
PhD Student
Donders Institute for Brain, Cognition and Behaviour
Centre for Cognitive Neuroimaging
Radboud University Nijmegen
Neuronal Oscillations Group
FieldTrip Development Team

P.O. Box 9101
NL-6500 HB Nijmegen
The Netherlands

Contact:
E-Mail: jm.horschig at donders.ru.nl
Tel:    +31-(0)24-36-68493
Web: http://www.ru.nl/donders

Visiting address:
Trigon, room 2.30
Kapittelweg 29
NL-6525 EN Nijmegen
The Netherlands

_______________________________________________
fieldtrip mailing list
fieldtrip at donders.ru.nl
http://mailman.science.ru.nl/mailman/listinfo/fieldtrip



More information about the fieldtrip mailing list