[Mristudio-users] ACPC alignment

[susumu]

Hi JH,

 

You have to be careful if you want to move your DTI data. There are two
options to do post-processing ACPC alignment; reorient raw DWIs or reorient
after tensor calculation. The latter approach (reorient after tensor
calculation) also has two options; reorient tensor (or vector) or reorient a
rotation-invariant scalar images such as ADC, FA, and eivenvalues.

 

Let's go over the image analysis process;

 

Raw DWI - tensor map (Dxx, Dyy, Dzz, Dxy, Dxz, Dyz) - scalar map (FA, ADC,
eigenvalues)

 

You can't simply reorient the first two images (raw DWI and tensor) because
they are "rotationally variant". 

 

For example, if you want to rotate the raw DWIs by 10 degree to get the ACPC
alignment, you have to also rotate your b-table. If your first gradient
orientation is [0, 1, 0] (y gradient) and rotate the brain about x axis,
your gradient orientation now becomes something like [0, 0.9, 0.03].

 

If you rotate the tensor map by 10 degree, you have to make sure that all
tensor elements in each pixel also have to be rotated.

 

In the third step (scalar map), you can do any rotation you wants without
worrying any reorientation.

 

Another question is how you define the rotation angles to achieve ACPC. If
the brain orientation is not correct both in sagittal and axial images, you
get a complex situation to get the brain straight.

 

For reorientation, what we can offer is the following;

 

1) Do tensor calculation and obtain the tensor map without worrying the
brain orientation.

2) use one of our atlases as a target of the brain orientation. All our
atlases are ACPC aligned (more precisely the orientation of ICBM-152).

3) use Landmarker -> rigid rotation option to reorient your data to the
orientation of the atlas. Here you want to reorient the tensor map you
obtained in the step #1. Landmarker knows how to reorient tensor data.

4) reload the re-oriented tensor map to DtiStudio and recalculate all the
images such as FA/ADC/eigenvalues/eigenvectors

 

You can skip #4 if you reorient FA/ADC/eigenvalues (not tensor) in step #3.
But in this approach, you can't get reoriented eigenvector maps and
therefore you can't do tractography after rotation.

 

If you do manual ROI-based analysis, it is important that your brain data
are all oriented in the same way so that the axial slices of interest (or
coronal or sagittal) are oriented in a comparable manner among subjects. If
you use affine transformation (instead of rigid) in step #3, you can also
make the size of the brain the same. In this way, your manual ROI drawing
becomes even more reproducible because you can write a protocol like,
"extract axial slice #60 in the atlas coordinate and draw ROI on the
internal capsule".

 

This is not really related to your question, but hope it'll help.

 

Susumu

 

  _____  

From: mristudio-users-bounces at mristudio.org
[mailto:mristudio-users-bounces at mristudio.org] On Behalf Of Jarnet Han
Sent: Monday, October 20, 2008 7:50 PM
To: mristudio-users at mristudio.org
Subject: [Mristudio-users] ACPC alignment

 

Hi everyone,

 

I was just wondering about the easiest way to do ACPC alignment before
loading up the files into DTI studio (or does DTI studio have a built-in way
to do this?).  Currently, we are using the protocol from Mori's paper which
appears to be based on ACPC aligned brains.  Thanks in advance.

 

Best,
JH

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20081113/502566f4/attachment-0001.html