Dear Shane,
if the additional steps help this is fine with the CAT12 preprocessing.
Best,
Christian
On Thu, 29 Jul 2021 13:18:26 +0000, Shane Schofield <[log in to unmask]> wrote:
>
>Hi Christian,
>I tried having a go at the segmentation again, and the results seem to be much improved if I run ANTS N4 on the T1 before CAT12. The blue voxels are before ANTS N4. Is it advisable to run N4 before feeding the images into CAT12, since CAT12 is also doing its own intensity correction?
>
>
> On Wednesday, July 28, 2021, 11:02:33 PM GMT+1, Christian Gaser <[log in to unmask]> wrote:
>
> Dear Shane,
>
>On 28 Jul 2021, at 23:44, Shane Schofield wrote:
>
>> Thanks for the explanation. Could you elaborate more on "higher local
>> variance" and why is this not really an issue in my case?
>Statistical power is high if the effects are large and noise (variance)
>is low. Thus, you have more sensitivity to find effects. If local
>variance is higher it’s more difficult to find GM effects in these
>areas.
>
>> Is that because the amount of misclassificaion is not severe at all?
>> My goal is to derive some PET measurements in GM areas in native PET
>> space and I was thinking of using the new CAT12 function for that.
>This will be almost unaffected. The effects I have mentioned rather
>relate to GM.
>
>> In this context, my PET measurements will be attenuated because of the
>> non-PET signal in the dura / misclassified GM, right?
>> Moreover, how does this misclassification affect the cortical
>> thickness and pial surfaces? The surfaces look fine in the report but
>> I still don't know how to view the surfaces. Can I do that using
>> Freesurfer Freeview?
>It’s easier to remove non-brain areas for surfaces. Thus, these
>skull-stripping problems will be usually not visible. The volume-based
>skull-stripping approach in CAt12 is rather conservative to not cut
>brain areas.
>Surfaces can be visualized with the „Display Surfaces“ function. To
>also allow to list and select raw surfaces you have to change the
>selection filter from „*.gii“ to „“. However, you can also use
>Freesurfer (Freeview).
>
>Best
>
>Christian
>>
>>
>>
>> On Wednesday, July 28, 2021, 10:37:07 PM GMT+1, Christian Gaser
>> <[log in to unmask]> wrote:
>>
>>
>> Dear Shane,
>>
>> some MR sequences show more difficulties in skull-stripping and my
>> personal experience is that changing the skull-stripping approach in
>> CAT12 does not really help a lot. Anyway, you can try the older gcut
>> method or the SPM approach in CAT12. Usually the remaining non-brain
>> areas lead to a higher local variance. Therefore, it is more difficult
>> to find effects in these areas. In your case this is probably not an
>> issue.
>>
>> Best,
>>
>> Christian
>>
>> On 28 Jul 2021, at 23:29, Shane Schofield wrote:
>>
>> Thanks Christian, it works! Sorry, I was putting the flag after the
>> image. On a follow-up note, I see some dura being classed as GM in
>> some of the subjects (screenshot below). Would you recommend trying
>> some of the other skull-stripping options or pursue some manual edits?
>> This happens in approximately 1 out of 10 subjects I might add.
>>
>>
>>
>> On Wednesday, July 28, 2021, 03:46:33 PM GMT+1, Christian Gaser
>> <[log in to unmask]> wrote:
>>
>> Dear Shane,
>>
>> setting the number of processes to "1" should work. Please check that
>> you have defined the flags before defining the data:
>> ~/spm/spm12/toolbox/cat12/cat_batch_cat.sh -fg -p 1
>> your_nifti_files.nii
>>
>> Furthermore, you may additionally define the "-fg" flag to keep the
>> process in the foreground.
>>
>> Best,
>>
>> Christian
>>
>> On Fri, 23 Jul 2021 07:41:28 +0000, Shane Schofield
>> <[log in to unmask]> wrote:
>>
>>> Hi Prof Gaser,
>>> Hass anyone used cat_batch_cat.sh successfully within a SLURM
>>> environment?
>>> I am having difficulties running CAT12 on my school's HPC because it
>>> is taking up too much cores I think.
>>> Can I check, is the cat_batch_cat.sh an openMP or MPI application? I
>>> tried to set the -p to 1 but it does work.
>>> Thanks for the help!!
>>> Best,Shane
>
|
