Hi,
Shape refers to localized differences. For example, atrophy or growth in a
specific region. Any anisotropic scaling and shears would remain as well.
--useRigidAlign removes the translations (difference in centroids), then
removes rotations (based on least-squared differences) between the input
and reference. --useScale calculates to RMS distance from the vertices to
the centroid, the scale factor is calculated by the ratio of this RMS
distance of the reference to that of the input. Effectively, shape in this
context is anything not removed by the above.
Cheers,
Brian
>
> Dear Mark,
>
> This is really helpful.
> One more quick question, would you let me know what kind of information of
> vertex is left after --useScale option.
> I know it is shape, but I can't simply imagine what 'shape' means. Is it
> an angle (to the center of the structure) of the vertex?
>
> Thanks million!
>
> Buyean Lee
>
>
>
>
>
> -----Original Message-----
> From: Mark Jenkinson <[log in to unmask]>
> To: [log in to unmask]
> Sent: Wed, Nov 25, 2009 12:12 am
> Subject: Re: [FSL] MNI vs. Native in FIRST
>
>
> Dear Buyean,
>
> The --useScale option does a 7 DOF registration for the
> individual meshes. That is, it takes them from the native
> space and removes all pose (translation and rotation)
> as well as isotropic scaling (with --useScale). This will
> naturally be different from removing it based on the size
> of the head (or ICV), especially if the size of the caudate
> has changed. It is also different from the MNI space
> version, as in the MNI space the whole brain has
> been used to optimise a 12 DOF registration, allowing
> different scaling on different axes and also skews (or shears).
>
> What you should use is a tricky question, and that's why
> we have the different options available as there is no single
> "correct" answer. If you want to normalise by TICV then
> the third option is best. If you want to adjust for global size
> and shape differences of the whole brain, including anisotropic
> scaling and skews, then use MNI. If you want to eliminate the
> overall size changes in the structure of interest, and just look
> at shape changes on top of this, then use the first option.
> Or there is the fourth option of doing it in native space without
> any correction for brain size or structure size. It is possible
> that this might be interesting in some cases too, especially
> if you are concerned that the normalisation is not well estimated
> and creating additional noise in the measurements.
>
> I'm afraid that's the best I can do for guidance on this issue.
> I hope it helps.
> All the best,
> Mark
>
> On 25 Nov 2009, at 05:42, Buyean Lee wrote:
>
>> Hi Mark and Brian,
>>
>> I compared size-and-volume of Caudate between two groups and got the >
>> following three output files (*.vtk).
>>
>> 1. Native*.vtk generated with --useScale (is it the same as Native >
>> with 7 dof?).
>> 2. MNI*.vtk generated without --useScale and without covariate > (i.e.,
>> TICV)
>> 3. Native*.vtk generated with the design matrix that has TICV as a >
>> covariate; I didn't use '--useScale' option.
>>
>> I used the command lines in the FIRST website.
>>
>> To me, the output images should be similar because the three vertex >
>> analyses correct for global size difference (actually, I am not >
>> positive if --useScale corrects for head size or size of caudate >
>> differences). But, the output images look quite different, > especially
>> between Native*.vtk and MNI*.vtk.
>>
>> I wonder if there is any guideline which output images one should use.
>>
>> Or, does MNI*.vtk include only shape difference after correcting for >
>> head size?
>>
>> Thank you,
>>
>> Buyean Lee
>>
>>
>
>
>
|