Hi FSL list,
I am doing group ICA, and stuck by registration phase. At first I used ApplyXFM GUI, set the example_fun2standard.mat in FEAT file as transformation matrix; the filtered_func_data as input volumes, set a 3mm reference standard image made by myself based on 2mm MNI brain image as reference volume, and I think I can finish regsitration and downsample in one step. However, after group ICA I found that many ICA components were beyond the back ground of reference image. Checking the FSL forum, I found a command line for registration you suggested:
flirt -in mapX -out mapX_hr -ref ${FSLDIR}/data/standard/MNI152T1_2mm applyxfm -init example_func2standard.mat
In case your machine can’t handle the higher-level melodic run on maps at 2mm isotropic resolution, you can downsample by replacing the last two options with:
-applyisoxfm 4 -init example_func2standard.mat
I used this command to do registration and downsample to 3mm data. After this step I tried to do group ICA, and I still found some components were beyond the brain back ground. I am very confused about this. Did I make some mistake on registration and downsample? Do I need to try FNIRT to improve my registration? I appreciate your kind response to this problem.
Thanks,
Hongyu Yang, Ph. D
Division on Addictions,
Psychiatry Department,
UT Southwestern Meidcal Center,
Dallas, TX 75235
|