open-discussion > affine registration corrections
Showing 1-5 of 5 posts
Oct 19, 2012 05:10 PM | hannah kyllo
affine registration corrections
I've been unable to run the affine registration without significant
error, and a mismatch of the resulting transformed image. Is there
any way to prevent this, or an error correction tool with bioimage?
I'm also looking for a better way to place fiducials on the
pre-registered images (MRI).
thanks for any ideas
thanks for any ideas
Oct 22, 2012 03:10 PM | Xenophon Papademetris
affine registration
corrections
Please clarify. What type of images are you using? What platform?
...
Oct 27, 2012 05:10 PM | hannah kyllo
RE: affine registration corrections
I successfully registered MRI and CT, with minimal error, using a
linear rigid registration after manually transforming the images.
However, is there a way to localize electrodes using the newly
registered CT image? I'd like to save the images in their original
DICOM format, yet when the registration is finished, the matrix
file saves, and I can't find an option to save or export the
registered image. Can I apply the .matr file at a later point,
perhaps in a different viewer? Additionally, after using the
electrode editor to localize electrodes in a CT image (DICOM
format), I'd like to use the .mgrid file (the saved electrode
grid), to visualize on the registered MRI (DICOM) image. So all in
all, I'm looking for a way to visualize electrodes on a registered
MRI image, using the reference CT.
thanks for any ideas!
thanks for any ideas!
Oct 31, 2012 03:10 PM | Xenophon Papademetris
RE: affine
registration corrections
[ Sorry for the delay, we had a small storm uphere in the northeast
to deal with.]
Unfortunately BioImage Suite does not save out DICOM images, only analyze/nifty -- you will need external software to do that.
In general you have a few options w.r.t. localization.
The easiest is to registered the MRI and the CT and save out the resliced CT image. Then localize the electrodes in this resliced CT space. Then you can load the MRI image in the editor which should match OK. (The linear registration may not be perfect but there is no great way to do a nonlinear MRI to CT mapping directly -- we use an intermediate post-implantation MRI image here and do nonlinear pre-MRI to post-MRI and linear post-MRI to post-CT)
Alternatively you can transform the electrodes to MRI space once you have localized them. To do this bring up the transform grid control from the Electrode Editor Control (under the Grid menu Grid|Transform Grid) and load the matrix in any of the three slots (they are combined) and then INVERT the matrix. The matrix maps stuff from MRI to CT and you need to do the opposite.
Good luck,
Xenios
Unfortunately BioImage Suite does not save out DICOM images, only analyze/nifty -- you will need external software to do that.
In general you have a few options w.r.t. localization.
The easiest is to registered the MRI and the CT and save out the resliced CT image. Then localize the electrodes in this resliced CT space. Then you can load the MRI image in the editor which should match OK. (The linear registration may not be perfect but there is no great way to do a nonlinear MRI to CT mapping directly -- we use an intermediate post-implantation MRI image here and do nonlinear pre-MRI to post-MRI and linear post-MRI to post-CT)
Alternatively you can transform the electrodes to MRI space once you have localized them. To do this bring up the transform grid control from the Electrode Editor Control (under the Grid menu Grid|Transform Grid) and load the matrix in any of the three slots (they are combined) and then INVERT the matrix. The matrix maps stuff from MRI to CT and you need to do the opposite.
Good luck,
Xenios
Nov 5, 2012 04:11 PM | hannah kyllo
RE: affine registration corrections
the localization seems to be more accurate when not transforming
the grid (i.e. not using the inverted matrix). Linearly registering
the MRI and CT works as well, although I've had success using
affine registration, so long as I do a manual registration first,
and then check the "use initial transform" button. The process
takes a few minutes.
thanks for the ideas!
thanks for the ideas!