Keren:
I do this routinely, but often it is very difficult. I use the brain
extraction steps used in fsl_anat, but have tweaked them to match infant
particpants. One issue is that the GM/WM/meninges intensity separation is
different in infants in adults, and bet has a hard time with this; one
issue is that the default flirt or fnirt does not register an adult
template to an infant head very well (one step of fsl_anat). A couple of
tricks I have used are to 1) use an age-appropriate average MRI template
rather than the MNI152; 2) manually check a variety of BET values to get
the Œbest¹ I can with the FSL tools; 3) ultimately I often have to do
some manual editing on the resulting brain. I suspect that the default
fsl_anat sh pipeline won¹t work with infants, you will need to deconstruct
it to duplicate the steps with non-default parameters.
You might check out the ANTS antsBrainExtraction routine. It has similar
problems with infant brains, but you specifically choose the template
(rather than default MNI template). I think it uses a routine based on
segmenting GM/WM/CSF and then filing holes/expanding holes and
dilating/eroding a brain mask. Again, this could be an issue with infants
< 7 mo whose non-myelinated axons confuse segmenting programs based on
adult heads; and so I substitute an age-appropriate infant template (with
head, brain, brain mask, segmenting priors). I have had some success in
doing this for infants who the fsl_anat approach does not work.
In general I have found bet to work very well with infants about 12 months
or older, and not very effective with 3 month olds. In general a routine
using tissue segmentation / fill holes / dilate-erode can solve just about
any infant brain extraction.
John
***********************************************
John E. Richards Carolina Distinguished Professor
Department of Psychology
University of South Carolina
Columbia, SC 29208
Dept Phone: 803 777 2079
Fax: 803 777 9558
Email: [log in to unmask]
HTTP: jerlab.psych.sc.edu
***********************************************
On 11/19/14, 3:24 PM, "Mark Jenkinson" <[log in to unmask]>
wrote:
>Hi,
>
>Have you tried using different BET options (via the ‹betfparam).
>Also, the method used by fsl_anat involves registration to the standard
>space, using the MNI152 template, and this might not be appropriate for
>your infants. So you might want to try just running BET rather than
>fsl_anat, or turning off the registration within fsl_anat.
>
>All the best,
> Mark
>
>
>
>> On 19 Nov 2014, at 15:12, Keren Lesinger <[log in to unmask]>
>>wrote:
>>
>> Hi ,
>> I am using fsl_anat on anatomical fmri scams of adults and infants.
>> it seems that the T1_biascorr_brain.nii.gz of adults extract the brain
>>ok.
>> However , the infants brain are over cutt by fsl_anat.
>> (I'm using fsl_anat for the purpose of brain extraction and the
>>registration functional images using fsl piplines with bbr (which use
>>fsl_anat outputs).
>> is there something I can do to improve the infants brain extraction?
>>
>> Thanks,
>> Keren
>
|