Hi again,

    I just wanted to share how I solved my issue in case others are running into anything similar. First, I switched to using 16-bit .nrrd files. Then, I began using the following script which combines the registration and reformat scripts per Greg's suggestion:

cd "C:\CMTKregistration"
perl "C:\CMTK\bin\munger" -b "C:\CMTK\bin" -awr 010203 -A "--accuracy 0.8" -W "--accuracy 0.8" -T 10 -s "JFRC2010_CLAHE_16bit.nrrd" "images"

The most important change I made was to include an accuracy statement(-A "--accuracy 0.8" -W "--accuracy 0.8") which improves the rate of successful registration without noticeably changing result quality. Another important change I made was to pre-process my images. Before running CMTK I prepare my images as follows: 1) rotate the image to be registered to as close to the same orientation as the template as possible 2) change the voxel size of the image to be registered so that the volume of the brain to be registered matches that of the template brain as closely as possible.

Thanks to Torsten as well for the suggestion about Cygwin--I haven't yet tried it but plan to in the future. I would like to see if they can improve the speed of registration as well as the rate of success.

Thanks again,
Gabriella