The 4D fmri timeseries have 230 time points. I looked into the transformed
4D power spectrum and found they have 115 frequency points. Another point,
the 4D fmri data were priorly filtered by using a band-pass filter with a
internal of 0.01Hz ~ 0.1Hz.
On Mon, 30 Jun 2008 17:54:40 +0200, Andreas Bartsch
<[log in to unmask]> wrote:
>how many time points have you sampled?
>Von: FSL - FMRIB's Software Library im Auftrag von Xi-Nian Zuo
>Gesendet: Mo 30.06.2008 12:46
>An: [log in to unmask]
>Betreff: Re: [FSL] AW: [FSL] TICA on resting state
>Do you mean the TICA described in Damoiseaux 2007 Cerebral Cortex paper can
>not be directly done in MELODIC_GUI? Could you give some further details
>about this? Following as you suggested, I firstly transformed my 4d fmri
>data to 4d power spectrum data by using AFNI's plugin and then fed them into
>melodic command. However, melodic gave me only two components by using the
>automatic estimation option. What's wrong with my command as follows:
>melodic -i ./.filelist --outdir=Tconcat-pICA3.gica
>-mask=mask/finalmask.nii.gz --tr=2.16 --report --bgimage=bgimage.nii.gz
>--Opca --Owhite --Omean --Ostats -a concat
>Any suggestion is deeply appreciated!
>On Wed, 18 Jun 2008 14:30:50 +0200, Andreas Bartsch
><[log in to unmask]> wrote:
>>it depends what you want but yes, possibly.
>>You can use fslspec to get the power spectra of your 4D data prior to
>feeding them into TICA. If it runs too slow for you use matlab or R with the
>appropriate import / export.
>>>does this mean there is power spectrum information in each voxel?
>>>Are networks then identified by shared power spectra?
>>>And when group differences are evaluated in contrasts, a significant
>difference means that the spatial maps identified within the power spectrum
>for that component are different between groups?
>>No!;) It means that the power spectrum of a given component is attenuated
>or augmented, i.e. that there is significantly more (or less) power to that
>very component. It can be due to more / less prominent peaks, phase shifts,
>contaminating frequencies aside from the main peaks ... Note that
>quantifying power spectra across subjects is certainly tricky. So what you
>would typically do is to go back to your data and extract the relevant
>information (using either fslmaths/meants and / or a formal GLM testing).
>>Hope that helps,
>>Von: FSL - FMRIB's Software Library im Auftrag von Christina Hugenschmidt
>>Gesendet: Mi 18.06.2008 12:51
>>An: [log in to unmask]
>>Betreff: [FSL] TICA on resting state
>>Hi all (and Christian in particular),
>>I am interested in running a TICA analysis comparing resting state and
>>also video watching between two groups. From reading the posts and
>>information on the FSL website, I think I should be using the same
>>methodology described in the Damoiseaux 2007 Cerebral Cortex paper, where
>>the data is fourier transformed prior to running the TICA analysis. I was
>>wondering if you could tell me the process used to transform the data
>>prior to running the TICA?
>>I would also like to be sure that I understand exactly what is being
>>compared between groups in this analysis. When the data is transformed,
>>does this mean there is power spectrum information in each voxel? Are
>>networks then identified by shared power spectra? And when group
>>differences are evaluated in contrasts, a significant difference means
>>that the spatial maps identified within the power spectrum for that
>>component are different between groups?
>>Thanks for any insights you can offer,