> I have started to use the DARTEL tool for the first time and I have some
> questions that I could not find any answer for them in the SPM manual.
>
> Background:
>
> I am investigating effect of 3 experimental manipulations on 2 groups of
> subjects (N1=27,N2=23).
>
> Each subject undergoes 1 scanning session that is composed of: 1 EPI scan
> [event related design] and 1 anatomical scan.
>
> We are interested only in functional analysis (not a VBM study).
>
> The preprocessing sequence we have used until the DARTEL step:
>
> Realignment (Estimate & Reslice; resliced images: all images+mean) à
> Coregristration (Estimate) à New Segment (save bias corrected was selected;
> native tissue: Native + DARTEL imported were selected) à DARTEL.
>
> Processing the anatomical images with DARTEL:
>
> Following the manual, due to the usage of New Segment step, first I
> processed the anatomical images. I created a template [using Run
> DARTEL(create Template) option]. I selected all the rc1*.nii and the
> rc2*.nii files in the same order – from all the subjects. Then I used the
> DARTEL(normalize to MNI space) process, selected the flow fields and only
> the gray matter c1*.nii images of all the subjects – again with maintaining
> the same order as before. I chose ‘Few subjects’ and preserve concentrations
> options. Other factors weren’t changed.
>
> My questions:
>
> 1. Regarding normalizing the functional images using DARTEL.
>
> a. Should I need to do the previous step (creating a template) also
> for the functional images? And if so, which images are to be selected: the
> post realignment images from each subject – in one single selection of all
> the images from all the subjects [as done with the anatomical one]? Or
> should I separate the process and create a new module for each subject?
How you do this will depend on what works best for your data. If your
fMRI are completely free of distortions, you can simply coregister
your fMRI to the anatomical scans and spatially normalise your fMRI
(instead of the c1*) using Dartel. Note that if you coregister after
estimating the segmentations and warps, then you MUST move the fMRI
into alignment with the anatomical data. If you move the anatomical
data into alignment with the fMRI, then you won't be able to apply the
Dartel transforms any more.
In reality, the fMRI will be spatially distorted, which means that the
rigid-body alignment between fMRI and anatomical scans may be quite a
lot less than perfect. If the dropout artifact is not too strong and
the grey and white matter is clearly visible, you could try segmenting
the data and aligning the rc1 and rc2 from the fMRI. This
occasionally seems to do quite a reasonable job.
>
> b. In the DARTEL: Normalize to MNI Space,
>
> i. Which
> DARTEL Template should I use? I assume if I don’t need to do the
> create-template step for the functional images then I should use the
> Template_6.nii that was generated from the DARTEL template of the anatomical
> images, is it correct?
Use Template_6.nii, which is the most crisp of the series.
>
> ii. In the
> manual (page 445. Section 41.2) it is written that the flow fields should be
> selected for every subject. It is written that “u_c1*.nii” files should be
> selected. I wondered is that correct? Because I thought that after the new
> segmentation process we should use the “u_rc1*.nii” files [Dartel imported
> and not the native ones]. Is it typo or did I miss something?
That must be a typo. You should not actually have any u_c1*.nii
files. Only u_rc1*.nii.
>
> c. Just to clarify, in this step should I do this step for each
> subject independently i.e. selecting each time the flow field and the
> functional images for each subject or should I load all the flow fields and
> the functional images (ordered) in the same batch?
You'll need to do this for each subject at a time, which is what the
‘Few subjects’ option does. If you can batch it (using some sort of
scripts), it may make life easier.
>
> 2. If I understood correctly the DARTEL process includes the smoothing
> step so the next thing to do is to advance to the other steps (1st analysis,
> etc.)?
Yes.
Don't tell Karl I said this, but personally I would do the first level
analysis in native space (disabling any threshold masking) in order to
compute contrast images, and then spatially normalise the contrast
images using Dartel. With one contrast image (or only a small number
of them) per subject, you can easily use the 'Many subjects' option
for writing the spatially normalised images. Then you'd do the second
level analysis on the spatially normalised contrast images.
>
> 3. In order to match each subject anatomical image with his functional
> activity - Would the images after the DARTEL will be coregitered to the
> anatomical images or an additional coregisteration process must be employed
> before the 1st analysis?
When processing is treated as a pipeline, the within subject
processing should be done before the between subject. Therefore, try
to align the native space images.
> Looking forward to learn from you about how to use the DARTEL tool
> correctly,
I'm not entirely sure there is a single correct way. The procedures
that work best will depend on the artifacts, contrast etc in your
images.
Best regards,
-John
|