[Mristudio-users] Is DTIStudio based on DICOM patient coordinate system or image coordinate system?

Tue Oct 26 01:32:21 EDT 2010

Dear Mori-san, 

Thank you so much for your detailed comments. It really helps. 

Usually, I use DICOM patient coordiante and image coordinate spaces. Now, I need to process many oblique image volumes, such as coronal or saggital, or any oblique image volumes by DtiStudio. 

I confirmed with TOSHIBA MRI vendor. They said the gradient vector acquired by my current TOSHIBA MRI scanner is based on patient coordinate. By your reply, it is very clear that I need to transform the gradient vector from patient to image coordinates by the image orientation (DICOM TAG: 0020,0037), and then input them to DtiStudio. Let us suppose this transform matrix is MT. 

And I think that during tensor calcuation, this transform MT will affect eigen vector and its derived color map and fiber tracking results, however, this transform MT will not affect eigen value. So, there are two methods to apply this transform MT. 

1> One is to apply MT to gradient vector and then input this new gradient table to DtiStudio to do tensor calculation. Then, we can get correct eigen value and eigen vector map images. 

2> Another is to use the patient-coordinate-based gradient vector to call DtiStudio to do tensor calculation directly and then apply MT to the generated eigen vector and color map. 

I think that both of methods can get the same results. Is my understanding correct? 

>

> That means, you have to rotate them based on oblique angles ("consider rotation" needs to be checked).

>

For DICOM images, I will try to check ON "consider rotation" in DtiStudio. Thank you for your information. 

Thank you for providing the method to verify the results. I will use your proposed method to verify whether it is correct or not. 

Sincerely, 

Qiqi

Date: Mon, 25 Oct 2010 12:19:14 -0400
From: susumu at mri.jhu.edu
To: mristudio-users at mristudio.org; xli16 at jhmi.edu
Subject: Re: [Mristudio-users] Is DTIStudio based on DICOM patient coordinate system or image coordinate system?

Hi Qiqi,

I want to follow up Hangyi's comment. 

First of all, we have so many ways to define coordinates and the nomenclature could be also confusing. There are two very clear definition;

1) Image Coordinates: At the very end of data acquisition process, we get an image. The brain could seem upright or tilted. It doesn't matter. We have image X, Y, and Z. We do color-coded visualization (x: red, y: green, z: blue) and fiber tracking in this Image Coordinates. In the end, this is the most important coordinates. Likewise, the gradient table should be defined in this Image Coordinates. If a line of a table says, [1, 0, 0], this should be a gradient along the first axis of the image matrix (= image X).

2) Physical (gradient hardware) Coordinates: The very beginning of the coordinate system in the data acquisition process is the Physical Coordinates. This is there even before we put a subject. It's a physical entity. In this coordinate, gradient [1, 0, 0] means, "apply X gradient hardware".

Usually, difference between Image and Physical Coordinates is only permutation. For example, X hardware gradient (Physical [1, 0, 0]) could be Image Y (Image [0, 1, 0]) depending on image prescription. However, things get complicated once you use oblique imaging as will be discussed later.

In addition, there are two more ways;
3) Patient Coordinates: I believe that this is defined by the parameters we prescribe in the console. FH is foot-head, RL is right-left, and IS is inferior-superior. This tells overall orientation of the human body. By specifying these, we can link the Physical Coordinates and the patient position. For example, if we put a patient in head-first, face-up, and axial imaging, RL aligns to X gradient, FH to Z, and IS to Y gradient systems. Note that Patient Coordinates are not really anatomical coordinates. For example, even if we specify which is the RL orientation, the actual patient RL anatomy could be tilted. The exact anatomical RL axis is not considered here. Patient Coordinates just define the overall patient position.

4) Phase/Frequency/Slice orientation: This defines axial/coronal/sagittal image orientations.

You don't have to worry all these unless you use oblique imaging. As soon as oblique angles are introduced, then these coordinate systems start to dissociate and things get very complicated. Please correct me if I'm wrong, but I believe Physical and Patient coordinates are always aligned within the relationship of permutation. Image Coordinates and Phase/Frequency/Slice are aligned within the relationship of permutation. Between these two groups, we need coordinate conversion calculation as soon as imaging plane is tilted (oblique imaging).

Important questions are;

1) Which coordinate system DtiStudio uses? 
A: Always Image Coordinates. Gradient tables have to be in Image Coordinates.
2) Which coordinate system the vendor-provided gradient tables are?
A: It depends. GE and Philips (gradient overplus = off) dynamically adjust the gradient table such that the gradient tables always follow the Image Coordinates. Namely, you can use the table as is no matter what the oblique angles are.
A: For Siemens and Philips (gradient overplus = on), the gradient tables remain in Physical Coordinates. That means, you have to recalculate the gradient table based on oblique angles you used before fed into DtiStudio. For Siemens you can click "consider rotation if applicable" checkbox (if you are using Mosaic). For Philips, you can use a gradient table calculate at http://godzilla.kennedykrieger.org/~jfarrell/OTHERphilips/GUI.html.
A: I believe Toshiba uses the GE way so that you don't need table recalculation based on oblique angles, but not 100% sure.
3) Which coordinate systems DICOM/Mosaic headers use?
A: I am not 100% sure but so far our experience says, "Physical Coordinates". DtiStudio has a function to extract a gradient table from certain DICOM/Mosaic files. When they are extracted, they seem in Physical Coordinates. That means, you have to rotate them based on oblique angles ("consider rotation" needs to be checked).
4) How can I be sure if I'm doing it correctly?
A: Please do the following test;
a> Scan a subject in a regular way. Acquire one DTI dataset without any oblique angles.
b> Tile the image plane severely, maybe 30-40 degree. Acquire one DTI dataset.
c> Ask the subject to rotate the brain severely, maybe 30-40 degree. Acquire one DTI without oblique angles.
d> Tile the image plane so that the brain looks upright. Acquire one DTI dataset.

If everything is working correctly, in all these dataset images should look right;
Image a and d: The brain looks upright. the corpus callosum is red at the mid-sagittal. All color should look symmetric. 
Image b and c: The brain looks tilted. The color of the corpus callosum do not look symmetric.
All images: the fiber tracking should appear the same in 3D view. Simply put an ROI in the genu of the corpus callosum, and you can immediately tell if they look the same or something is not working

Hope this helps.

Susumu

2010/10/24 chichesterchiche <seigeweb at hotmail.com>

Dear DTIStudio Experts, 

   We know we need to input the gradient table in DTIStudio. However,  when doing tensor map calculation, which coordinate system is DTIStudio based on? Is DTIStudio based on DICOM patient coordinate system or image coordinate system? I acquired the DWI images by TOSHIBA MRI and the gradient vectors are based on patient coordiante. Can I input the patient-coordinate-based gradient vector to DTI Studio directly, or do I need to transform the patient-coordinate-based gradient vector to image coordinate system at first? 

Thank you. 

Qiqi

_______________________________________________
mristudio-users mailing list
mristudio-users at mristudio.org
http://lists.mristudio.org/mailman/listinfo/
Unsubscribe, send a blank email to: mristudio-users-unsubscribe at mristudio.org

_______________________________________________ mristudio-users mailing list mristudio-users at mristudio.org http://lists.mristudio.org/mailman/listinfo/ Unsubscribe, send a blank email to: mristudio-users-unsubscribe at mristudio.org 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20101026/a53b0857/attachment-0001.html 