Hi,

> I am attempting to use the B0 unwarping option in FEAT - pre-stats to correct distortion an fMRI (EPI) study in which I am interested in both cortical and particularly brainstem results.
>
> Unfortunately, for a couple of subjects I don't have have B0 fieldmap images. It has been suggested that an average of the other subjects' maps would be much better than not applying B0 unwarping at all (which doesn't give very good results at the individual or group level).
>
> I would like to know if this a reasonable thing to try to do, and if so, how can I proceed?

Yes, I would use an average fieldmap rather than none.

> Here is one idea:
>
> 1. For each participant optimize registration parameters between their EPIs (I have three), T1s and standard space using nonlinear registration and BBR.
> 2. For those participants who have field maps, transform each fmap (radians) and magnitude images to standard space using the transformation matrix from above
> 3. Average these across all participants
> 4. For the participants who do not have fieldmaps, transform the averaged maps back into their native space using their inverse transformation matrices

Yes, that looks right, at least for the rad/s fieldmap image. You don't need to do it for the magnitude map though. You can just use the subject's own T1-weighted image as the magnitude map, along with the fieldmap (in rad/s) transformed from the standard space into the native (T1-weighted) space of the individual. The only thing to be careful about is the masking of the fieldmap and the brain extraction of the corresponding "magnitude". I would take a copy of the T1-weighted image, appropriately mask/threshold your rad/s fieldmap image (after it is transformed into the native space) and then apply that same mask to the copy of the T1-weighted image to use as the brain-extracted fieldmap magnitude image.

All the best,
Mark

>
> Best regards,
>
> Emily