[Mristudio-users] question regarding the diagonal components

[susumu mori]

Hi Amir,

This is an interesting question and allow me to share with other users
in the mailing list.

The 3 gradient systems of MRI usually do not have exactly the same
gradient strengths (it's almost impossible to make X, Y, and Z
gradients with exactly the same strengths). Therefore, it is one of
the important setup procedures to calibrate the gradient strengths.

Suppose X, Y, and Z gradients have 20, 20, and 10G/cm strength (I
exaggerated). Once calibrated and if you want to deliver 10G/cm
gradient in all three directions, even if you instruct [10, 10, 10
G/cm], the actual instruction to hardware goes [50, 50, 100%]. In this
way, gradients systems are calibrated.

If the gradient systems are not calibrated, a sphere object would have
oval appearance in imaging. To ensure the anatomical fidelity, all
gradient axes must be calibrated.

Therefore, it is usually unthinkable that diffusion gradients (b
vector) have calibration error; if so, your brain images would also
have a wrong aspect ratio.

You also mentioned about "orientation of tensor component". This is a
different question from gradient strength. Please refer to my recent
answers to one of the questions asking about physical and image
coordinates. It is your b-table that defines what coordinate you are
using. Unless you use oblique imaging, image and physical coordinates
are aligned and there is not much to worry about. When you use oblique
imaging, there are two ways to define the b-table; Physical and Image
coordinates. If you use a gradient table in physical coordinates
(gradient axes), the resultant fiber angles are defined by the
physical coordinates; if the fiber orientation is [1, 0, 0], that is
along the magnet X gradient. If you process the same data using a
table defined in the image coordinates, then the fiber angle are
defined in X, Y, Z coordinates of the image matrix; If the fiber
orientation is [1, 0, 0], then the fiber is along Image X axis.
Obviously, you want the image coordinates, I believe.

Hope this clarify your questions.


On Sat, Oct 30, 2010 at 3:19 PM, Amir Zolal <amirzolal at gmail.com> wrote:
> Dear dr. Susumu Mori,
> I have a simple question, are the components of the diffusion tensor matrix
> oriented along the physical scanner gradients or does the position of the
> scan play a role? I conducted an experiment where I expected the
> longitudinal (z) component to react differently than the x and y, because of
> the unequal strength of the gradients. This did not happen (ie. i could not
> see any significant difference between the changes in Dxx, Dyy and Dzz
> components) and I am looking for an explanation ..
> Thank you in advance for the answer,
> Amir Zolal