Dear Georg,
>
> Currently we are working on the registration of 26 subjects. FNIRT
> did a neat job on 24, but on two of them the slices of the
> functional data look like long haired hamsters after registration to
> the standard brain - it's due to extremely warped edges. We used
> standard settings with 6 and 12 DOFs and results won't get better
> with more DOFs. I uploaded some files; maybe you could take a quick
> look (example_func2standard.png included, upload reference number:
> 809223). However, I must admit FNIRT still does a better job on
> ventricles than FLIRT does.
I have seen this behaviour, and I must admit I am not sure why this
sometimes happens. The fnirted T1's typically still look good, but
there are crazy warps just outside of the brain. Combined with a
slight functional->structural misregistration these warps are then
sometimes inside the edge of the functional data, and leads to the
problems you see.
I'll have a look at your data and see if I can figure something out.
> Another question related to this topic is: would it be possible to
> run higher level analysis with different types of registration
> without difficulties? In this case 24 people with FNIRT and 2 with
> FLIRT...
I think that as long as you are just averaging across the group there
are no principal objections (apart from a lack of tidiness). If these
are two groups (e.g. patients and controls) that you want to contrast,
then it becomes a lot more iffy and I think that should be avoided.
Good luck Jesper
|