Dear John
Many thanks for your quick response. It looks like I misunderstood the strategy for analysing two groups. I thought the idea of using two steps like this was to make eliminate systematic differences in the warping distance from individual to template in the two groups. If the recommended way is to treat them as one, then that makes my life easier. My groups have slightly different numbers, which may introduce bias, but the difference is small.
So now I don't need to compose flow fields, but in case it helps anyone else, here's exactly what I did (initially):
Using SPM12
Deformations tool
Composition: Two flow fields
FF1=individual subject to group
FF2=group template to combined 2-group template
Other flow field settings as default (Forwards/Backwards = Backwards; Template = blank)
Output: pushforward
Bounding box and voxel sizes: default (NaN)
Modulation & smoothing: amount & 4 4 4 (although these shouldn't matter)
First result gave me an image that was shifted superiorly (couldn't tell if the warp was correct as I could only see bottom of cerebellum). I changed the flow field to "forward" and the result looked good, but the top of the brain was chopped off. I expect this means I needed to have set the bounding boxes differently when creating the templates.
Thanks again
Paul
On Wed, 20 May 2015 11:27:55 +0100, John Ashburner <[log in to unmask]> wrote:
>I would strongly advise against aligning the two groups separately.
>Preprocessing should be done in a way that is blind to the effects that you
>intend to study - otherwise you may introduce artifactual differences. I
>would strongly suggest just aligning all the scans in your study together
>instead.
>
>There are several things that could go wrong when composing deformations
>together, as the end result depends on exactly what you did. My first
>guess is that you are using SPM12, and filling in a Dartel template name
>when you use the Deformations utility. The SPM12 manual is slightly out of
>data for the composing deformations in the Using Dartel chapter. When
>composing deformations, I'd suggest leaving this field empty (ie don't
>specify the template).
>
>Best regards,
>-John
>
>
>On 20 May 2015 at 10:39, Paul Wright <[log in to unmask]> wrote:
>
>> Dear SPMers
>>
>> I'm running DARTEL for VBM for the first time and I'm a bit confused about
>> the various steps and image types. I've worked through the tutorial just
>> fine for a single group analysis, but I want to analysis two groups
>> together by first warping individual patients and controls to two separate
>> group templates, then warping the two group templates together. (Which I
>> understand minimizes and confounds). I then want to apply the two warps in
>> a single step, so I composed the flow fields using the Deformations tool.
>> When I apply the resulting deformation to the c1 images (since that's what
>> the tutorial does) the results are misaligned.
>>
>> Should I instead use the rc1 images when applying a deformation field
>> based on two composed flow fields?
>>
>> If so, why does the tutorial go back to the original c1 images when
>> applying the final warp? If it's to avoid a reslice step, then is there a
>> way to get a deformation field to work on the original images like the flow
>> fields do?
>>
>> Cheers
>>
>> Paul
>>
>
|