Bad news: those fixes did not help appreciably.
I eroded the magnitude fieldmap, and ran all the analyses again, to no particular help. I tweaked the brain extraction (substantially more work, since front-to-back problems are hard to address with BET's variables), and even with both of those fixes applied, I still see huge amounts of non-brain activation, especially at the second level - just like in the past.
This zipfile contains new versions of the same files as offered before - now with both of your original suggestions applied. T1 structural files, fieldmap files (_ero are the eroded versions), .fsf file, raw .nii file, input data (denoised_data.fsf, which has been unwarped and ICA-denoised), and the reg and reg_standard directories from a first-level run. Let me know if there's anything else I can provide.
https://www.dropbox.com/s/lz4689y0yk4kes3/regTestPhilip3.zip
Thanks for your help!
-Benjamin Philip
|