Q: Hi All, I’m trying to align a PET volume to it’s anatomy, and am running into some problems (and could use some help). I just have a couple of questions …
(1) do I need to resample my anatomical volume to match that of the functional volume prior to calling minctracc?
(2) when I try to resample the native MRI to the functional, I usually get an empty volume out. I suspect that this is related to the difference in z-start values, but am unsure how to proceed. The 2 input volumes look like this …
- MRI **
file: lestat_t1_native.mnc
image: signed__ short 0 to 4095
image dimensions: xspace zspace yspace
dimension name length step start
-------------- ------ —— -----
- MRI **
xspace 176 1 −85.4759
zspace 256 −1 113.831
yspace 224 −1 144.386
- PET **
file: pibXmean.mnc
image: unsigned byte 0 to 255
image dimensions: zspace yspace xspace
dimension name length step start
-------------- ------ —— -----
- PET **
zspace 63 −2.425 −452.7
yspace 128 2.05941 −130.772
xspace 128 2.05941 −130.772
The mincresample command and the resulting volume looks this …
>>mincresample -verbose -clobber -like pibXmean.mnc lestat_t1_native.mnc lestat_t1_native_rsl.mnc
- resampled empty T1 volume —
file: lestat_t1_native_rsl.mnc
image: signed__ short 0 to 4095
image dimensions: zspace yspace xspace
dimension name length step start
-------------- ------ —— -----
- resampled empty T1 volume —
zspace 63 −2.425 −452.7
yspace 128 2.05941 −130.772
xspace 128 2.05941 −130.772
(3) does minctracc want both volumes to share the same data type (e.g. byte) or does this not matter?
(4) should I blur the native T1 prior to fitting? … or does minctrac deal with this?
http://www.bic.mni.mcgill.ca/pipermail/minc-users/2008-May/002059.html
A: (1)Nope.
(2) That mincresample call won’t work due to the difference in starts and steps between the two images - you’ll need to compute the transform first and then resample.
(3) This does not matter - though if you need to do mutual information registration, then it will reshape them to unsigned byte automatically.
(4)Probably - especially with those differences in initial position you’ll probably want to use a multiscale algorithm anyway; i.e. first transform with quite blurred data, second transform with less blurring, etc.