Hi all,
I have a set of 4 fMRI time series from a single subject that I want to register so I can
concatinate them into a single time series to run an ICA (yup, you know the data). The fMRI
data is a 64x64 matrix gradient echo EPI, 3 acquired on the same day, one on another day. I
also have a higher resolution 128x128 SE EPI for this subject. After motion correction of each
individual fMRI time series, I registered the example_func for each to the high res image
using a 6 param with the default FLIRT settings, then applied each registration matrix to the
time series. Concatinating the data and running through the time series, however, it's clear
that the registration between runs is not good. Looking at each example_func overlayed on
the high res with "slices," it's clear that the registration for each run is not great. It's the kind
of registration I'd expect to see between subjects -- good, but the ventricles aren't exactly
aligned and the brain perimeter is also just an ok match. Not what I expected from a 64 and
128 EPI of the same subject.
SO
Are there other parameters you recommend for within subject? A 7 param with the default
settings was also just so so...
What is the registration used in FLIRT? Would it be a good idea to just "motion correct" the
concatinated time series?
Thanks.
Robin
|