I would suggest some slight changes to make life easier:
1) Within the DARTEL code, there is some spatial smoothing of the iteratively
generated template. I would suggest reducing the amount of smoothing, as it
is generally easier to obtain a better overlap of serial scans than for
cross-sectional data. You may even wish to set the "Smoothing Parameter" to
"None".
2) Specify a subject-specific template name when you run DARTEL within
subject.
3) Generate modulated warped GM using the within subject flow fields. These
should be in alignment with the subject-specific template (mwc1BL and
mwc1FU). This template will be half-way between the BL and FU scans, so
there should be less bias in terms of whether BL is warped to FU, or vica
verca.
I haven't tested this, but I think it should then be possible to use the
Template data for each subject to register all the within subject data
together. Note that the template data has multiple volumes in the same file,
so you'll need to change the volume selector (usually set to 1) so that the
second volume could be selected. If this does not work, then you may need to
seperate the templates into their individual volumes (ie make two 3D files
from a 4D file). Once you have the inter-subject flow fields, then you can
use then to generate mwmwc1BL and mwmwc1FU files. These can then be
smoothed, prior to doing some form of paired t-test analysis.
I hope the above makes sense, and works. It is a bit messy, and I will try to
get around to making it easier to work with serial scans.
Best regards,
-John
On Tuesday 01 April 2008 16:11, Simone Reinders wrote:
> Dear John / DARTEL users,
>
> To analyze a longitudinal dataset (baseline = BL and Followup = FU) I
> would like to use DARTEL. I am trying to follow the steps as offered
> in the paper by Kipps et al. 2005 (JNNP) and to interpret them in a
> DARTEL manner.
>
> I have included my steps in detail below.
>
> I would like to obtain the Jacobian determinant (step 3)) of the flow
> field that maps BL into FU, but the output of step 2) gives me only a
> deformation field while the input to obtain a Jacobian determinant is
> a flow field.
>
> Is there a way to create a flow field (instead of a deformation field)
> that that maps BL into FU which can be used as input to the Jacobian
> step?
>
> Or am I completely on the wrong track?
>
> Thank you,
> Simone.
>
> Simone Reinders, PhD
> King's College London
> Institute of Psychiatry (IoP)
>
>
>
>
> Page 652 of Kipps paper interpreted in DARTEL:
>
> 1) Both BL and FU images are manually reoriented. The BL is
> coregistered to the FU, with FU being the reference image and no
> reslicing.
>
> *) Initial DARTEL steps:
> a) Create scan specific *_seg_sn.mat files for both FU and BL as
> desribed in section 1.1 of the DARTEL manual
> b) Perform the 'Initial Import' step for both BL and FU both grey and
> white matter, creating rc1*.nii and the rc2*.nii files.
>
> 2) Find subject specific warping/flow fields that warp BL to match FU
> within subject.
> a) Run DARTEL (create Template) with the subject specific imported
> data of the FU and the BL scan (section 1.4 of the DARTEL manual).
> * Run DARTEL (create Template)
> - Images
> * Images: Select all the rc1*.nii files 1 for FU and 1 for BL (from
> step *)b)), for e.g subject pp1.
> * Images: Select all the rc2*.nii files 1 for FU and 1 for BL for pp1.
> This will create two flow fields u_rc1_pp1FU*.nii and u_rc1_pp1BL*.nii.
> The name of the flow field is 'u_rc1*' but it is the flow field for
> both the rc1 and rc2 images.
> b) Compose the resulting deformations:
> TASKS -> utils -> Deformations
> * Deformations
> - Composition
> * DARTEL flow
> - Flow field: Specify the u_rc1_pp1BL*.nii flow field.
> - Forward/Backwards: Backward.
> - Time Steps: Usually 64.
> * DARTEL flow
> - Flow field: Specify the u_rc1_pp1FU*.nii flow field.
> - Forward/Backwards: Forward.
> - Time Steps: Usually 64.
> * Identity
> - Image to base Id on: Specify the imported data of FU.nii
> (i.e.the rc1_FU and rc2_FU)
> - Save composed deformations as a file 'y_BLtoFUpp1'
> Note: y is a deformation field not a flow field.
>
> 3) Obtain Jacobian determinant from the deformation/flow field
> generated in 2) via DARTEL
> Task -> Tools -> DARTEL tools-> Jacobian determinants
> This is where I would like to use the flow field for 'y', creating
> jac_y*.nii files.
>
> ... if I have the Jacobian determinants I would like to proceed with:
>
> 4) These are the FU files from the "Initial Import" stage (from *)b), i.e.
> the rc1*.nii and the rc2*.nii files.
>
> 5) Create customised template on the basis of all FU scans only
> * Run DARTEL (create Template)
> - Images
> * Images: Select all the rc1*.nii files for FU.
> * Images: Select all the rc2*.nii files for FU.
> This will create FU flow fields u_rc1_*FU*.nii and a customized
> template on the basis of FU scans only.
>
> 6) Create warped images (DARTEL instructions: page 4) by applying the flow
> field from step 5) onto the FU images from 4) and on the Jacobian
> jac_y*.nii file from step 3) this creates wrc*.nii and wjac_y*.nii images
> (all subject specific)
>
> 7) multiply the wrc1*.nii and the wrc2*.nii with the wjac_y*.nii images
> (subject specific) => this gives a measure of tissue specific volume change
> from BL to FU, i.e. over time.
>
> 8) This step from the Kipps 2005 paper is superfluous when using
> DARTEL but I might want to transform the results from 7) into MNI
> space.
|