Dear Natalia,
> I’m processing data (anatomie and functional) of a longitudinal study on stroke patients and I have some questions concerning the DARTEL registration. I already send these questions twice to the mailing list and nobody reply to them, so I’m not sure if this is so because nobody can really answer them or because I’m asking very basic things.
I suspect you haven't received a reply because you are performing an uncommon analysis, so it's not necessarily clear what the right approach would be. I'll have a go but don't have a lot of experience with longitudinal data myself, so hopefully others will jump in.
Just to start, it would be helpful to know exactly what you are looking at. Is the aim to, eventually, conduct a group study? Or is the focus on results in single subjects?
If you are going to be comparing a subject with another (whether another subject or a control), at some point you will need to normalize them together into a common space—that is, create a DARTEL template that contains all of the subjects you want to compare (or perhaps a representative subsample). If you are just looking within a single subject over time, you may not need to do this—this seems to be the approach you are going for, but it is worth clarifying.
Also, I wonder how far apart your sessions are, and whether you expect any substantial structural changes in subjects' brains over this time (e.g., post-stroke?). If you do not expect changes in each subject's brain, it seems to me that using DARTEL to compare sessions is probably not necessary—you could probably just coregister these together.
If you are doing a group analysis, and there was no expected change in brain shape, you could also perhaps use a single session's data (or a mean across sessions?) to enter in a group template, which may save some time and effort. If there is a chance that the subject's brain changed significantly, then I suspect that your approach of using DARTEL to obtain an average shape for that subject is probably a good one. But I would be interested to hear others comment on multi-stage DARTEL image registration.
> I try to use DARTEL to register the anatomical and functional data of the 3 sessions together. This is what I did :
> - Skull stripping of session 1, 2 and 3 separatelly : as describe in the SPM manual page 318 Chapter 33 „Using DARTEL“
Some people have reported improved segmentation results with skull stripping, but (especially with the new segment) it may not be necessary. This would save you a step if you omit it, and also keep everything related to tissue class separation within the same model, which is somewhat more elegant.
> - Segmentation of the skull stripped T1 image of session 1, 2 and 3 separatelly : with New Segment
> - Estimate warps (u_c1*.nii for each session) and Template of the 3 sessions (=Template6.nii): with Run DARTEL. I chose for Smoothing parameter None. I saw in previous studies that sometimes people do a mean image of all sessions. Is it correct to use instead the Template6.nii ?
If you are using DARTEL, you should warp to the Template_6. But see above for whether you may be able to omit the subject-level DARTEL. (It's not clear to me what previous studies have used the mean images for.)
> - Normalisation : with Normalise to MNI space tool. Is correct to use this tool even if I have a DARTEL template for each subject and not for a bigger group?.
If you want to get the results to MNI space, then this is the correct tool to use. But, if you are comparing subjects with each other (a common use of putting things in MNI space), you'll want to do this using a group template composed of all your subjects (or a balanced group).
> In this step I was also not sure how to enter the flow fields. I chose to „Few Subjects“ and enter each session as a subject. So that I apply the flow field of the first session to the funtional scans of the first session, the flow field of the second session to the functional scans of the second session and so on. I run « Normalise to MNI space » tool ones for the functional scans with « preserve concentration » and FWHM=8 ; and ones for the anatomical data (bias corrected image) with « preserve Amoun »t and FWHM=1.
This stage also depends on what you want to do in the end—i.e., if you adopt a different approach (as I've suggested above), you will only have one set of flow fields per subject, which you could then apply to all of these images.
If you are planning on doing any VBM, I would also suggest > 1 mm smoothing on the structural images, but that depends on what in particular you are doing with the data.
> - I’m not sure how to warp the sessions with each other in the DARTEL space. Can I just apply the warps that I got from Run DARTEL to the functional images? That means, can I warp the functional images of the first session with the flow field of the first session and the functional images of the second session with the flow field of the second session, and so on.
This relates to whether you need to register each subject's sessions together with DARTEL, and then to group/MNI space, or whether you can just do a single registration from subject to group/MNI space. If you do the 2-stage approach, I would think you would want to use subject-specific flow fields for each session (as you've described), and then the (average subject) -> group/MNI flow fields on this data. Although I believe you can combine these two sets of flow fields in a single step (I'm away from SPM at the moment so can't check).
> - Do I have to mask the lesion out to do the DARTEL normalization and If yes where (in which processing step) do I have to introduce the mask?
If you are normalizing just a single subject's data, I don't think masking would be necessary (although perhaps during segmentation). If you are working with group results, I'm not sure how well DARTEL will work with lesioned brains—if subjects have different shape brains, it may be difficult to compare. (That is, I assume that DARTEL will produce a solution, but it may be difficult to interpret.)
Hope this helps!
Best regards,
Jonathan
--
Dr. Jonathan Peelle
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104
USA
http://jonathanpeelle.net/
|