Hi Bryan,
Thanks for your help. fslmaths and fsl2ascii are absolutely useful,
but I would still have the bottleneck, since The image is being
created and then transformed to an ascii file, so I was wondering if
there is any other way that we don't create the image at the first
place.
I am running a single voxel tractography and I noticed that decreasing
the number of particles doesn't change the running time that much,
sounds like the I/O overhead is much more than the tractography itself
(at least in this experiment).
Saeideh
On Fri, Sep 24, 2010 at 3:13 PM, Bryan Benson <[log in to unmask]> wrote:
> Hi Saeideh,
>
> If I'm understanding correctly, what you want is fslmaths and fsl2ascii.
> Depending on what you are interested in using the values for, you can also
> just calculate your metrics directly with fslmaths. Just type 'fslmaths'
> into the terminal to get usage help.
>
> Bryan
>
> 2010/9/24 Saeideh Bakhshi <[log in to unmask]>
>>
>> Dear FSL experts,
>>
>> I am using fdt for a large number of iterations and I am only interested
>> in the non zero voxels and their values. As far as I know fdt produces a
>> NIFTI image as an output and then I can extract the nonzero voxels using
>> load_nii in matlab.
>> I have an I/O bottleneck here, is there any way to optimize this process
>> outputting the actual data as text at the first place?
>>
>> Thanks in advance,
>>
>> Saeideh
>
>
|