Dear everyone,

 

I am experiencing major issues with the normalization step of my data treatment.

I have 42 subjects to normalize. Each has a T1 sequence and a functional sequence. The EPI images consist of 884 volumes of 48 slices each, 1,5mm resolution, TR 3s (the FOW does not cover the entire brain as the study focuses on the frontal areas, as shown on attached figures).

During my pre-treatment which is done in SPM8, I first realign and reslice the images, then coregister the EPI and the T1 sequences (because of the limited FOV, I have to manually point the origin so that SPM can coregister well). The coregistration is then checked by hand for each subject. Everything looks fine up to here.

 

But then, I normalize using DARTEL tools according to the procedure described in the manual:

1 - New segment with creation of DARTEL imported tissue maps (using T1)

2 - DARTEL template creation (rc1 and rc2 dependencies)

3 - Normalize to MNI Space using the DARTEL flow field. I load both T1 and EPIs for each subject (using « few subjects » method, as I have a lot of images for each subject).

The results are frustrating: about half of the subjects (19/42) get their frontal region cut off (both for EPI and T1 normalized data).

So far, my understanding is something is wrong with the deformation matrices. When I visualize DARTEL flows of a “good” and a “bad” subject, the latter is shifted in the same way as the brain gets cut. Attached the CheckReg image of a couple of normalized T1s, corresponding DARTEL flows and the DARTEL template.

In fact, the deformation matrix is also shifted when comparing to the original T1 image and segmented tissues.

 

Do you have any idea about why this might be happening?

 

Thanks a lot!

Olga Davidenko
Doctorante
INRA/AgroParisTech UMR 914 Physiologie de la Nutrition et du Comportement Alimentaire
Chaire ANCA - Aliment, Nutrition, Comportement Alimentaire
+33 (0)1 44 08 14 22