Okay, thanks to you both for your help.
Ian
----- Original Message -----
From: "Brian Patenaude" <[log in to unmask]>
To: [log in to unmask]
Sent: Tuesday, 9 March, 2010 7:01:33 PM
Subject: Re: [FSL] improving first_flirt registrations?
Hi,
Watch out for the brainstem, it could effect the brainstem if the brain
extraction follows its border. The easy way around this is to input the
full head image into first using the transformation matrix found using
first_flirt on the brian extracted image. If you're using run_first_all,
you'll want to run all the `first_flirt` operations first, then input
them into run_first_all. I actually suggested this in general since it
gives you the opportunity to check the registration.
Cheers,
Brian
> Hi,
>
> The discouragement is not very strong.
> Only if you get reasonable registrations in both cases do we prefer
> the non-brain-extracted input. We found that this was slightly more
> reliable
> but it does depend on sequence, SNR and FOV.
>
> If you get good results with non-brain-extracted images (and they don't
> work otherwise) then just go with that. It will be fine. There is no
> consequence
> further on in the pipeline as FIRST only uses intensity information in
> the
> immediate vicinity of the subcortical structures.
>
> All the best,
> Mark
>
>
> On 9 Mar 2010, at 11:12, Ian Malone wrote:
>
>> Hi,
>>
>> A colleague is trying to run FIRST and has run into a problem with
>> first_flirt producing incorrect standard space registrations (for
>> example 90 degree rotations in the sagittal plane with a resulting
>> match to the wrong part of the head. Starting with brain extracted
>> images is more successful, but we've noticed the warning on the
>> FIRST page "Although it is generally discouraged, the flag *-b* will
>> allow *|first_flirt|* to also be used on brain extracted data."
>>
>> However I've checked the FMRIB technical report and Brian
>> Patenaude's thesis and haven't been able to find out why it's
>> discouraged; is it simply to avoid using the BETed images further
>> down the pipeline or a concern about the registration that would
>> result? I ask because one solution that's been suggested is to use
>> the brain mask from BET as an input to -inweight, which should
>> produce similar results for the registration, but without masking
>> out non-brain tissue in the resulting image.
>>
>> Thanks for your time,
>> Ian Malone
>>
>
>
|