Hi Shahrzad,
It used to be the case that this was automatically calculated, however this was unfortunately lost on the last versions of FLICA.
Once you've got the k number of outputs (or modalities, in our PNAS paper we had m=3), each with n number of components, and you are in the relevant FLICA directory, you can simply do this for each k in [1,m]:
fslmaths niftiOut_mik.nii.gz -sqr /tmp/tmpk.nii.gz && fslstats -t/tmp/tmpk.nii.gz -m >/tmp/energyk.txt
And in matlab:
load /tmp/energyk.txt
energyk = energyk * diag(1./sum(energyk))
Mk=[(0:n-1)', (energyk*100)]
dlmwrite('variancek.txt', Mk, 'delimiter','\t','precision','%.2f')
You'll see in variancek.txt that the second column sums up to 100% for that given k modality.
Hope this helps,Gwenaëlle ------------------------------------------------------------------------------------------------------------------------
Gwenaëlle Douaud, PhD
University Research Lecturer & MRC Career Development Fellow
FMRIB Centre, University of Oxford
John Radcliffe Hospital, Headington
OX3 9DU Oxford UK
Switchboard: +44 (0) 1865 222 493
Fax: +44 (0) 1865 222 717
www.fmrib.ox.ac.uk/team/principal-investigators/gwenaelle-douaud
www.fmrib.ox.ac.uk/research/fmrib-interface-analysis-clinical-neuroscience-group
-----------------------------------------------------------------------------------------------------------------------
De : shahrzad kharabian <[log in to unmask]>
À : [log in to unmask]
Envoyé le : Lundi 16 février 2015 9h48
Objet : Re: [FSL] flica- dimension estimation and spatial maps.
Thanks dear Gwenaëlle,
Yes, you are right, I see it now. and thanks for the components.
I have another question:
How does one estimate how much of the variance in the data is being explained by each component?From the outputs flica gives, I guess I should be using the weights of each component over sum of weight of all components multiplied by 100. Am I right?
Thanks a lot again. Shahrzad
From: Gwenaëlle DOUAUD <[log in to unmask]>
To: [log in to unmask]
Sent: Thursday, February 12, 2015 2:48 PM
Subject: Re: [FSL] flica- dimension estimation and spatial maps.
Hi Shahrzad,
The simple answers are:
- yes, you can choose the number of ICs (see Dimensionality selection in the Methods of our previous paper: http://www.sciencedirect.com/science/article/pii/S1053811912006532);
- and yes, the map from our latest PNAS paper is freely available: http://www.fmrib.ox.ac.uk/analysis/LIFO+AD+AOS/
I haven't added the first age-related map, as it is fairly unspecific and whole-brain, but I'm happy to add it on the webpage if you feel this would be helpful!
Cheers,Gwenaëlle
------------------------------------------------------------------------------------------------------------------------
Gwenaëlle Douaud, PhD
University Research Lecturer & MRC Career Development Fellow
FMRIB Centre, University of Oxford
John Radcliffe Hospital, Headington
OX3 9DU Oxford UK
Switchboard: +44 (0) 1865 222 493
Fax: +44 (0) 1865 222 717
www.fmrib.ox.ac.uk/team/principal-investigators/gwenaelle-douaud
www.fmrib.ox.ac.uk/research/fmrib-interface-analysis-clinical-neuroscience-group
-----------------------------------------------------------------------------------------------------------------------
De : shahrzad kharabian <[log in to unmask]>
À : [log in to unmask]
Envoyé le : Jeudi 12 février 2015 11h50
Objet : [FSL] flica- dimension estimation and spatial maps.
Dear FSL group,
- In Flica program, is it possible to choose the number of independent components automatically?
My other question is with regard to the spatial maps of the components. Specifically, the recent Douaud et al. (2014, PNAS) paper presents 2 main components that are correlating significantly with age. Are these spatial maps freely available ? (like the spatial maps provided by smith et al for resting state fmri components)
Shahrzad
