Hi
> I am trying to replicate the methods from the Damoiseax 2008
> Cerebral Cortex
> paper, and thanks to Christian Beckmann and the other members of the
> list, I
> am getting closer.
The analysis in the Cerebral cortex paper was done not using the
command line tools as these were released later.
>
>
> I have converted smoothed, normalized fMRI data of a resting state run
> (TR2.5/300s/120vols) into power spectral information using fslpspec.
> This
> results in 60 volumes representing different frequency bins written
> in a nii.gz
> file (Christian posted how to calculate the bins - many thanks). I
> am having
> difficulty running the TICA - the gica log said it wasn't converging
> at the PCA
> step - so I backed up and tried to run single-session ica on one
> subject.
>
Possibly due to the the additional pre-processing that is done in the
GUI but not in the command line version. The GUI is a wrapper for
various FSL tools followed by melodic - in the case of running melodic
on power spectra you don't want any additional processing , so I guess
it's best to run this from the command line.
> If I enter this data into melodic in FSL 4.0, the program seems to be
> overwriting the nii.gz file, so that the power spectral data is lost
Melodic will not overwrite any of the input data, Check the time
stamps on the file to see what's going on.
> . An error is
> then thrown that either says that the PCA failed to converge or that
> the data
> set is empty.
>
I suspect that in the GUI you're using high and/or low pass filtering?
This would then flatten the power spectra and mess everything up.
> If I enter the data into melodic in FSL 3.2, it will run, but it
> usually results in
> between 1-4 components that are not apparently biologically
> meaningful. We
> also tried omitting the first volume of the spectral data, as it is
> very low
> frequency and seems to have a lot of noise, but this did not change
> the
> results.
Quickest way of getting this to work is to use the GUI to set up a
full concat ICA analysis. Once this is completed the file .fielist in
the .gica folder will contain a list of all the pre-processed input
files. You can take ths list, convert the files to power-spectra and
then run melodic from the command line again, now on the new files
containing powerspectra. check the melodic log file to see how to call
the comman line version. Also, make sure that the approach flag is no
longer set to concat
hth
Christian
>
>
> I am wondering if I am entering the correct information into the
> GUI. I have
> tried entering the actual TR and a TR of 0.0, running melodic with
> and without
> functions like BET and MCFLIRT, and using various hpf or turning it
> off
> altogether.
>
> We are planning on Monday to try creating the power spectral data in
> another
> way to see if there is a problem with the pspec files themselves,
> for some
> reason, but I was wondering if anyone had had similar problems, or
> can see an
> obvious mistake(s) that I am making.
>
> Thank you so much for any advice you can offer! I am very interested
> to see
> the results of this analysis.
>
> Thanks,
> Christina
_______________________________________________
Christian F. Beckmann, DPhil
Senior Lecturer, Clinical Neuroscience Department
Division of Neuroscience and Mental Health
Imperial College London, Hammersmith Campus
Rm 419, Burlington Danes Bldg, Du Cane Road, London W12 0NN, UK
Tel.: +44 (0)20 7594 6685 --- Fax: +44 (0)20 7594 6548
Email: [log in to unmask]
http://www.imperial.ac.uk/medicine/people/c.beckmann/
Senior Research Fellow, FMRIB Centre
University of Oxford
JR Hospital - Oxford OX3 9DU
Tel.: +44 (0)1865 222551 --- Fax: +44 (0)1865 222717
Email: [log in to unmask]
http://www.fmrib.ox.ac.uk/~beckmann
|