> I am still unclear how mutual information coregistration deals with
> an epi volume that only has a partial brain and a full T1 image
volume.
>
> a-Does the coreg program ignore the regions somehow in the partial
> brain volume as being out of the intersection of where the 2 brains
> best fit?
The coreg routine works with joint histograms generated only for the
region of intersection of the two image volumes.
> b- Is the problem with coreg's problem with epi distortion in part due
> to geometric distortion with large enough intensities that have
> coherent information so it tries to fit this info in with the T1
brain?
There are many issues. The main one though is that a rigid-body model
does not account for nonlinear distortions. Therefore, the rigid-body
registration may have part of the images in good register, but other
parts wouldn't be registered. Other issues relate to the signal loss
etc in the images, which makes registration more difficult.
Best regards,
-John
|