Please see the full message below--my first attempt to send failed because I was slightly over the size limit. I will send the images in a subsequent email so you can see the flip for yourself.
Tyler
---------- Forwarded message ---------- From: Tyler Seibert tseibert@ucsd.edu Date: Thu, Mar 17, 2011 at 4:02 PM Subject: bbregister bug? with init-spm option--flips left and right To: freesurfer@nmr.mgh.harvard.edu
Hello,
I hope this is the correct email list--this is my first time posting (just subscribed).
I have been using bbregister to calculate a tranformation matrix between a T2 image (T2brain.nii.gz) and the subject's T1.mgz for subsequent use with mri_vol2surf. I have found that the init-fsl option frequently gives very poor registrations (e.g., registered T2 is rotated 30+ degrees from T1). The init-spm option seemed to be working great. However, upon close inspection, I became suspicious this option was actually flipping the T2 so that the left hemisphere became right and vice versa. I am using FreeSurfer version 4.5 and have spm8 referenced in my matlab path.
I used a directory ~/bbreg_test in my home directory for the test (linux), which consisted simply of running bbregister with init-fsl and init-spm options. The attached log file (log.txt) contains the commands I used. The terminal output from each test was saved in text files as well, also attached.
I have also attached two screenshots of the tkregister2 output for these tests. In each screenshot the far left image is the registered T2 using init-spm, and the far right image is the registered T2 using init-fsl. As the init-fsl version is clearly wrong, the middle image in each screenshot is the init-fsl output again, but manually rotated several degrees to the left to better emphasize the mirror-image relationship between the init-spm (far left) and init-fsl (middle and far right) output images.
Please let me know if there is anything further you need from me to investigate this or if there is any insight you can give me as far as how to avoid this error. I noticed at line 209 of the spm_test.txt log file there is a message stating that RAS coordinate system is assumed (see below, where I have copied lines 208-213 for convenience). This may very well be the problem, as the T2brain.nii.gz volume is in LAS, according to mri_info output. However, I do not see where I can override this RAS assumption.
>> >> >> >> >> >> /home/tylers/apps/spm8/spm_coreg.m > >> >> INFO: Assuming RAS coordinate system > >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> using
/home/tylers/bbreg_test/init-spm/tmp.bbregister.15515/spmregister/refvol.spmregister.img as the target image
using
/home/tylers/bbreg_test/init-spm/tmp.bbregister.15515/spmregister/movvol.spmregister.img as the source image
> >> >> >> >> >> >> >> >> >> >> >>
Thank you,