Hi
fsl_glm accepts time courses or spatial maps as the design matrix - it works out from the dimensions if you want spatial or temporal regression. ANd yes, -thr -uthr is what you want, then merge with fslmerge -t. fsl_glm then outputs a single text matrix where each column corresponds to the required time courses. Or, if happy with separate & unweighted time courses just use fslmaths -thr -uthr and then use fslmeants.
hth
Christian
On 25 Mar 2010, at 19:23, Michael Harms wrote:
> Not sure if I follow how to get separate output values for each cluster
> of a mask using 'fsl_glm'. Looking at the syntax, doesn't fsl_glm
> require a design matrix as input?
>
> Given a cluster index mask, I can use appropriate -thr and -uthr values
> in 'fslmaths' to get a separate, distinct mask for each cluster, for use
> in fslmeants. It would be convenient though if fslmeants itself was
> "cluster-aware".
>
> cheers,
> -MH
>
>
> On Thu, 2010-03-25 at 20:04 +0000, Christian F. Beckmann wrote:
>> Hi
>>
>> if you create a 4D file with all the different clusters as binarised mask for the individual volumes then you can simply use fsl_glm to do this (incl. the weighting if desired). For the former (turn the cluster output into 4D file) you can just write a simple shell script.
>> hth
>> Christian
>>
>>
>>
>> On 25 Mar 2010, at 17:01, Michael Harms wrote:
>>
>>> As a somewhat related comment of a "nice to have" item, it would be
>>> handy if there was an option to have fslmeants recognize that the mask
>>> may contain distinct clusters (e.g., using the -o flag of 'cluster') and
>>> thus return a separate average for each cluster.
>>>
>>> cheers,
>>> -Mike H.
>>>
>>> On Thu, 2010-03-25 at 17:47 +0000, Mark Jenkinson wrote:
>>>> No, it does not do any weighting. Internally it binarises the mask.
>>>>
>>>> All the best,
>>>> Mark
>>>>
>>>>
>>>> On 25 Mar 2010, at 17:39, Greg Burgess wrote:
>>>>
>>>>> When using a non-binarized mask, does fslmeants return the weighted
>>>>> average of the timeseries for the non-zero voxels?
>>>>> Thanks,
>>>>> --Greg
>>>>>
>>>>> ____________________________________________________________________
>>>>> Greg Burgess, Ph.D.
>>>>> Research Associate, Institute of Cognitive Science
>>>>> University of Colorado - Boulder
>>>>> Email: [log in to unmask]
>>>>>
>>>>> University of Colorado - Boulder
>>>>> UCB 594
>>>>> Boulder, CO 80309-0594
>>>>>
|