The registration of my functional images to standard space has been a bit
off for a lot of my subjects. I thought it may have something to do with
fnirt but it was in fact flirt that was causing the problem. The cost
function used (corratio) doesn't do a great job for my EPI to T1 registration.
I've been overwriting the registrations using normalised correlation as the
cost function with much better results - is there a way of specifying a
change to the cost function in the design.fsf file?
|