Hi all,
I am relatively new to FSL, and I am hoping to get some instructions on what's the best to align two brain MRI datasets.
Basically I have two 3D MRI datasets from the same patient (multi-echo spoiled gradient echo data, TE = 3 - 45 ms, even spacing) from two different time points. I would like to co-align the 2nd time point with the 1st time point. I have tried to use flirt with a dof=12, it doesn't work as well as I'd hope. Namely, I don't think flirt is performing the re-sampling process along the slice direction. From what I've read, the non-linear registration tool fnirt is suppose to work much better, but involves a much complicated process with various adjustable parameters. So for two 3D MRI datasets that are: 1. same matrix size (voxel size), 2. same person (the scan will be 6 m - 1y apart). What would you recommend to be the default parameters to align the two sets (e.g. sample command?). Does fnirt resample along the slice direction by default or that will require some other fsl command? Also would it be helpful to mask out the warpped slices along the slice direction prior to co-alignment?
One thing i noticed is that the flirt tends "to do more" with an later echo images, where tissue with a short T2 component are dark. I guess it has more "contrast" for flirt to work with thus the results tends to be better.
Thanks for your help.
|