Hi,
By doing one dilation and two erodes, it is just like doing one erode except that it will
fill in single-voxel holes in the middle. So it is quite safe to do this as it shouldn't
really include any non-valid voxels.
All the best,
Mark
On 22 Sep 2011, at 20:44, Naama Barnea-Goraly wrote:
> A while ago there was a discussion about the "holes" in images from the tbss_1_preproc output. It was my understanding that this is happening because some subjects have an FA value of "0" and then these voxels are excluded from the rest of the analysis. This makes sense since we would want to exclude areas of the brain with missing data.
>
> If we use the fix described below (dilate and erode the image), wouldn't that affect the analysis in those areas which shouldn't be included?
>
> Thanks,
>
> Naama
>
>
>
> On Jul 21, 2011, at 3:42 AM, Mark Jenkinson wrote:
>
>> Hi,
>>
>> Another alternative would be to replace "-ero" in the fslmaths call
>> with "-dilD -ero -ero" which should fix isolated voxels that are 0
>> inside the brain.
>>
>> All the best,
>> Mark
>>
>>
>> On 21 Jul 2011, at 10:39, Nicola Toschi wrote:
>>
>>> Hi Tyler,
>>>
>>> thanks very much for your reply. It turns out this is exactly what's happening.
>>>
>>> I will experiment starting from your suggestions.
>>>
>>> Thanks angain,
>>>
>>> Nicola
>>>
>>> On 07/21/2011 01:59 AM, Tyler Blazey wrote:
>>>> Hi Nicola,
>>>>
>>>> I had the problem you described before, and it was because a few subjects had FA values of 0 within the brain. The erode function within tbss_1_preproc would then create a 3x3x3 box of zero voxels centered around the original 0 voxel. These "holes" were then propegated to the mean FA.
>>>>
>>>> My solution was to turn off the eroding by removing the -ero flag from the fslmaths call within the tbss_1_preproc script. The downside to this was the erode function helps to get rid of some of the brighter voxels at the edge of the FA images. To correct for this, I eroded the brain mask and then applied the eroded mask to the FA image. This might not be the perfect solution, but it seemed to work ok.
>>>>
>>>> - Tyler
>>>>
>>>
>
|