Dear Hansruedi,
I think this really depends on the data.
In case the data of the two sessions (including distortions & deformations due to inhomogenities in magnetic field) are very similar, I'd say realign the two EPI runs onto the mean (1 batch 2 sessions), realign the two structural volumes onto each other and create a mean, and then coregister this structural mean onto the functional mean. This way, you realign within modalities, which should be (a little) more precise.
In case the data differs to some kind (let's say you have a head-only scanner, and placement of the subjects inside the scanner is quite different, then deformations e.g. in occipital lobe might be quite different as well), it might be better to go with 2 batches with 1 session each. Then coregister the structural volume of session 1 onto the mean EPI of session 1, and the same for session 2. Depending on the deformations it might be better to normalize the volumes independently.
But I'm interested in other opinions myself!
Best,
Helmut
|