Hi,
By a scaled mask I mean the mask scaled by a value.
That is, multiply the mask by a value, don't just add a
value, because 10 and 11 are similar enough (and probably
small with respect to the residuals) that it will still struggle
to calculate a sensible mask in the processing.
Instead multiply your mask by a number that is clearly a
lot larger than the values in your residuals and then add
that result to the residuals.
Hopefully that will sort out the problem.
All the best,
Mark
Martin M Monti wrote:
> Mark,
>
> sorry but it's still unsuccessful. First, though, let us make sure I
> understand what you mean by "scaled" mask. What I did is I took the
> image mask.nii.gz (from the same analysis that yields the residuals
> I'm interested in using), added the value 10 to all voxels (so brain
> voxels are = 11, non brain voxels are =10), and then added this mask
> to the residuals image.
>
> If I run the analysis however it still gives the following error:
>
> Stats
>
> mkdir -p custom_timing_files ; /bin/cp
> /imaging/mm03/STMemory/FSL2/sub203/Regressors/sub203_TGT.txt
> custom_timing_files/ev1.txt
>
> mkdir -p custom_timing_files ; /bin/cp
> /imaging/mm03/STMemory/FSL2/sub203/Regressors/sub203_NN-TGT.txt
> custom_timing_files/ev2.txt
>
> /imaging/local/linux/bin/fsl-4.0.1/bin/film_gls -rn stats -sa -ms 5
> -sp /imaging/local/linux/bin/fsl-4.0.1/bin/susan_smooth
> filtered_func_data design.mat 1000.0
> Log directory is: stats
> parad.getDesignMatrix().Nrows()=150
> parad.getDesignMatrix().Ncols()=6
> sizeTS=150
> numTS=0
> Calculating residuals...
> Completed
> Estimating residual autocorrelation...
> Calculating raw AutoCorrs... Completed
> Warning: Number of voxels = 0. Spatial smoothing of autocorrelation
> estimates is not carried out
> Tukey M = 12
> Tukey estimates... Completed
> Completed
> Prewhitening and Computing PEs...
> Percentage done:
> Completed
> Saving results...
>
> An exception has been thrown
> Logic error:- detected by Newmat: Maximum or minimum of null matrix
>
>
> Error: FILM did not complete - it probably ran out of memory
>
>
>
> Have I misunderstood your suggestion?
>
> all the best
>
> martin
>
>
>
>
> Mark Jenkinson wrote:
>> Dear Martin,
>>
>> I think it will still have problems generating a proper mask if you
>> add a constant
>> everywhere.
>>
>> I suggest you add a scaled mask image and see if that works.
>>
>> Otherwise it is probably trying to analyse voxels where there is zero
>> variance in
>> the res4d (because it is outside the mask) and failing because of that.
>>
>> All the best,
>> Mark
>>
>>
>> On 10 Apr 2008, at 17:16, Martin M Monti wrote:
>>
>>> Mark,
>>>
>>> thank you, I did add a constant to the whole image, but it still is
>>> not completing. The error however is now a little further down, in
>>> the stats:
>>>
>>> An exception has been thrown
>>> Logic error: detected by Newmat: Minimum or maximum of null matrix
>>>
>>> Error: FILM did not complete - it probably ran out of memor
>>>
>>> I tried running the same script with/without FILM (in fact, this
>>> analysis is on residuals which should already have been
>>> pre-whitened..), but i get exactly the same error. Is there any
>>> other feature of the residuals that may be causing this?
>>>
>>> thanks
>>>
>>> martin
>>>
>>> Mark Jenkinson wrote:
>>>> Hi,
>>>>
>>>> The problem is most likely due to the fact that res4d has a zero mean.
>>>> This causes problems for mask generation and intensity normalisation,
>>>> which is where this crashes.
>>>>
>>>> Add on a scaled mask image to your res4d (to give it a non-zero mean)
>>>> before running FEAT and you'll probably be fine.
>>>>
>>>> All the best,
>>>> Mark
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> Martin M Monti wrote:
>>>>> Dear All,
>>>>>
>>>>> I'm running a 2-step regression by taking the residulas of a a
>>>>> first-level and using them into another first level.
>>>>> However, as I run the "second" first-level I consistently get the
>>>>> following error:
>>>>>
>>>>> Prestats
>>>>>
>>>>> Total original volumes = 150
>>>>>
>>>>> /imaging/local/linux/bin/fsl-4.0.1/bin/fslroi
>>>>> /imaging/mm03/STMemory/FSL2/sub203/sub203_Flobs_Res example_func 75 1
>>>>> /imaging/local/linux/bin/fsl-4.0.1/bin/fslmaths
>>>>> /imaging/mm03/STMemory/FSL2/sub203/sub203_Flobs_Res -thrp 10 -Tmin
>>>>> -bin mask -odt char
>>>>> /imaging/local/linux/bin/fsl-4.0.1/bin/fslmaths
>>>>> /imaging/mm03/STMemory/FSL2/sub203/sub203_Flobs_Res -mask mask
>>>>> /imaging/mm03/STMemory/FSL2/sub203/sub203_Flobs_Res
>>>>> /imaging/local/linux/bin/fsl-4.0.1/bin/fslmaths
>>>>> /imaging/mm03/STMemory/FSL2/sub203/sub203_Flobs_Res -ing 10000
>>>>> filtered_func_data -odt float
>>>>> ERROR:: Empty mask image
>>>>> ERROR:: Empty mask image
>>>>> ERROR:: Empty mask image
>>>>> ERROR:: Empty mask image
>>>>>
>>>>> etc ..
>>>>>
>>>>> As I inspected the input data file ("sub203_Flobs_Res.nii.gz"
>>>>> above -- which is simply a copy of the res4d.nii.gz file in the
>>>>> stats folder of the "first" first-level analysis) I noticed that
>>>>> by when the process has finished (and given the error above) it is
>>>>> empty, yet I am sure that it was not empty before launching the
>>>>> feat script.
>>>>>
>>>>> Ideas what I'm doing wrong?
>>>>>
>>>>> all the best
>>>>>
>>>>> martin
>>>>>
>>>
>
