I did, for example a hand tapping task (30sec on 30sec off) during the 3sec*105scans.
Another task of foot movement with the same settings. Other task of listening a tale and so on.

The onset time chosen in SPM was 10:20:100 with units scans

I also use the 3 column file option with the txt file:
23 30 1
83 30 1
143 30 1
203 30 1
263 30 1

and I run Melodic and the 6th component is similar to the spm result. (I can not attach the images)
But I'm still want to run Melodic to run the other tasks and compare, because sometimes is difficult to find the correct component that correspond to the task in MELODIC

Thanks,

Best

Carolina

2011/5/9 Cornelius Werner <[log in to unmask]>
Hi Carolina

On Mon, May 9, 2011 at 4:00 PM, Carolina Valencia
<[log in to unmask]> wrote:
> Dear FSL users,
> (...)
>  different motor and languages tasks examinated in independent cycles.

What do you mean by this? How many separate tasks were involved?


The steps in SPM were:
> - Reorient the images according to the AC
> - Realign the functional data
> - Coregister the structural with the mean functional data as a reference
> - Segmentation of the structural image
> - normalise the realigned and the bias corrected data
> - smoothing the normalise, realign functional data

Did you make sure that onset times in your regressor file or .mat
reflect that you discarded 5 timepoints? This is one of the most
frequent mistakes (at least in my experience).

> The steps in FSL FEAT were:
> - Using BET in the structural Data for brain extraction
> - In data tab choosing the functional data, using a high pass filter of 90 (calculated with cut off calc) and delete 5 scans
> - Pre stats all defaults, bet extraction on, FWHM of 6
> - stats tab I use the model setup wizard with rArA with r of 30 seg and A of 30 seg
> - Post stat I left defaults
> - registration tab I used the result from Bet extraction in the main structural image

If you have separate tasks (see above) then you shouldn't use the
wizard. Actually, I recommend using the "full model setup" with a
3-column-file in all but the most simple settings. If you want to
compare output from two packages, you certainly should use the same
onset files.
Another way to find out if there is meaningful signal in the data is
running MELODIC. Here, you do not need to know about onset times. If
there is language related activation in the data, one component (or
more) will pick it up, along with a nice 30s-timecourse. So, if you
see that (and FEAT or SPM won't give you anything reasonable), then
it's most likely misspecified onset values.

>
> Thanks in advance,

You're welcome
Cornelius

--
Dr. med. Cornelius J. Werner
Department of Neurology
RWTH Aachen University
Pauwelsstr. 30
52074 Aachen
Germany