[Mristudio-users] AIR image dimensions/ Talairach WMPM/Eigen vector flipping/ROI Error

[susumu mori]

Hi Bill,


> 1. In Diffeomap I perform Air Linear Registration in order to align my
> data with that of the Template (in my case I use the Talairach atlas). My
> images are 256*256*55 and I normalise it with the template of dimensions
> 192*192*136. How much does the difference in image size affect the accuracy
> and reliability of the subsequent analysis.
>

because this is a simple upsampling, I assume that you don't have to worry
too much about the accuracy/reliability, but I don't have a definite
answer. Practically, I believe the accuracy is much more influenced by the
registration accuracy, not by the upsampling.

>
>    - As an additional to ensure the best LDDMM result, is the contrast of
>    the two images the most important in setting up the images before entering
>    the LDDMM screen?
>
>
>
when you do non-linear normalization, two images must have very similar
contrasts. Non-linear normalization aggressively deforms images to find the
best match. If the contrasts of the two images are different, the matching
becomes a very difficult task. So, please make sure that you use the same
contrast images and do intensity (histogram) matching before LDDMM.

2. When I load the Talairach WMPM from diffeomap as an ROI the regions of
> interest are loaded as object 1,2 etc. Is it possible to get the name of
> each region of interest? When I perform the intensity measurements using
> the WMPM I get the a table listing a the same amount of parts as the
> objects i.e. in the case of Talairach 170. I then use the following paper
> to find what the abbreviations stand for:
> *
> **Oishi et al,** Atlas-based whole brain white matter analysis using
> large deformation diffeomorphic metric mapping: Application to normal
> elderly and Alzheimer’s disease participants*
> euroimage 2009 June ; 46(2): 486–499. (Appendix A part 4)
>
> Is this the correct approach? Or is it the that the regions of interest in
> the ROI objects box contain these parts as can be seen when I choose the
> button under Atlas: *WMPM of division of defined ROIs*
>
>
If you load an ROI map (like WMPM) in the RoiEditor format, it loads the
ROI names too. If you load an ROI map in the binary format, you get the
generic ROI names (i.e. 1, 2, 3,,,). In this case, you need to find a
look-up table, which should be in the RoiEditor/Image directories.


>
> 3. When I load or create the tensor image file (.d file) in DIffeoMap I am
> asked if I want to flip the eigenvectors. If I save the file as
> eigenvectors being flipped when I created the file do I also have to flip
> them again while opening the file afterwards or do I assume they are now in
> the correct direction?
>
>
The "flipping" business is important only when you happened to use a table
with wrong +/- signs. I would suggest you to use a correct table and forget
about this business. In old days, before the table information was stored
in DICOM header, we had only one generic gradient table from each vendor.
These generic tables had arbitrary +/- definition and we had to figure out
the +/- sign empirically. In this case, we often calculated the tensor
using the generic table first and correct the sign later by "flipping" the
vector orientation. There are two occasions when the flipping becomes
important. One is tractography. We load the eigenvector, flip the sign, and
do tractography. The other is image transformation. We load the tensor
file, flip the sign, and transform. Here, once the transformation is done,
you don't have to do any further flipping.



> 4. Finally, when I create an ROI in either DTIStudio or ROI editor I save
> it as a binary map. When I then try to reload it in the fibre tracking
> section I am informed that the* "Image parameters in this file are not
> consistent with with current ones, continue?"* When I select yes the ROI
> loads in the expected position. Is it a normal occurance to get an error
> even though I created the ROI in the same image during the same DTIStudio
> session?
>
>
you get his message when the matrix size and file size don't match. Please
check if your file size is correct. Also, please check FAQ#15 in here (
https://www.mristudio.org/wiki/faq).

Susumu


>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20111213/6ba7257e/attachment.html