Dear Joep,
>
> I tried to apply FLIRT on my betted FLAIR image with mutual information as cost function. This, however, gives poor results.
> I also tried other settings and even tried to use FNIRT, but it all does not seem to work.
> Comparing my FA image with the FLAIR image, I think the problem is that the FA image clearly shows white matter areas (as high intensities), whereas the FLAIR image does not specifically show these areas at all (it looks more "solid").
> Therefore, I tried to use MD image (instead of FA) to align with the FLAIR. With FNIRT (non-linear!), this already gives way better results (probably since MD resembles FLAIR much more compared with FA). I think for now, this could work for me since my goal is to align the FLAIR data with diffusion data, or do you think this is a bad approach?
it is definitely a good idea to use the tensor-derived scalar image that is the most similar to your FLAIR image to drive the registration.
I am not sure it is a good idea to use FNIRT. Have you used any other form of distortion correction for your diffusion data? If so, you should use FLIRT with 6 (possibly 7) dof.
If your diffusion data is distorted relative your FLAIR data it becomes a little more of a grey zone. Some people have reported decent results from a “fnirting distortion correction” like that, but there are also lots of ways in which such a scheme can go wrong.
Jesper
>
> Kind regards,
> Joep
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Jesper Andersson
> Sent: maandag 12 september 2016 15:05
> To: [log in to unmask]
> Subject: Re: [FSL] Registration of FLAIR with FA
>
> Dear Joel,
>
>>
>> I would like to register a patient's FLAIR data with that same patient's FA data.
>> Is this possible and if so, is there a standard way to do this ? (i.e. do I use FLIRT or FNIRT?
>
> when registering within-subject-between-modailty we generally recommend FLIRT.
>
>> What cost function do I use?
>
> For FLAIR to FA I think mutualinfo or normmi are your best bets. You should also BET your flair images first.
>
>> Do I use cost weighting for white matter hyperintensities in the FLAIR data?
>
> If these are small and localised I suspect it won't make a big difference. Though it can of course never do any harm to.
>
>> Do I use the FA configuration file for FNIRT?)
>
> Always use a config file with FNIRT, but not in this case since you are using FLIRT.
>
> Jesper
>
>> Thanks in advance.
>>
>> Kind regards,
>> Joep Kamps
>>
|