[Mristudio-users] reslicing DTI images

Wed Jun 4 10:02:48 EDT 2008

There are two ways to do the reslice. One is to reslice the raw DWIs or
reslice images after tensor calculation. Either way, you can use the
"reslice" function in Landmarker.

1) Reslice DWIs: Because you want z reslice, you first have to convert your
1,900 Dicom to a series of 3D volume data. You can easily do it by
DtiStudio. Read all Dicom and save each 3D data as raw or analyze format (I
recommend raw). You have to read each 3D volume to Landmarker and reslice it
to the dimension and pixel size you want. The only problem of this approach
is, you have to repeat it many times. For example, if you have 2 b0 and 30
DWIs, you have to do it 32 times. Landmarker can read text script, but it
doesn't have butch processing capability. 

2) Reslice after tensor calculation: Here you have two choices. You can read
and reslice tensor elements by Landmarker. Then you can reload the resliced
tensor into DtiStudio to recalculate eivenvalues, eigenvectors, and other
maps like FA. The procedure is;

> Calculate tensor by DtiStudio. Save each element (Dxx, Dyy, Dzz, Dxy, Dxz,
Dyz) separately as a regular floating point image or save them into one file
(so called *.d file in DtiStudio).

> Load the image into Landmarker. If the 6 elements are saved separately,
you have to read them individually. If you save them into one *.d file, you
need to use "Open *.d" button to load all at once.

(Please obtain the latest Landmarker because old versions can not
interpolate tensor data. Also, the first image you open is always treated as
a template in Landmarker and "Open *.d" is inactivated. You have to open the
image you want to interpolate in the second window (first image can be
anything because we don't use it for this interpolation operation). First
please open a scalar image such as b0 or FA. Then you can find "Open *.d" in
the right column of the window)

> The interpolation button at in the bottom of "Image" section, second from
the right. Type in the new dimensions and perform relice.

> Save the 6 interpolated tensor elements into one file (new *.d file). Just
for the sake of procedure, also reslice one scalar image such as b0 and save
it.

> Load the resliced scalar image (e.g. b0 with the new dimension) to
DtiStudio. You can see "Open *.d" button in the right column of the viewing
window. Load the new interpolated *.d into DtiStudio. In the dialog window,
check "eigenvalues/vectors". 

The second option is, first calculate all the scalar maps such as FA and
interpolate them. This can be easily done by Landmarker;

> Load a FA map and interpolate using the button described above.

These three ways lead to slightly different results. Visually the tensor
interpolation tends to enhance FA contrast. For example, if there are two
adjacent pixels with FA = 0.8 but fiber angles with 90 degree difference, FA
interpolation between these two pixels gives FA = 0.8, but tensor
interpolation gives FA < 0.8. Between two different fibers with different
angles, there remain low FA regions with the tensor interpolation. I believe
the raw DWI interpolation leads to image similar to the tensor
interpolation, but not sure. 

This may give you impression that the tensor interpolation is the right
thing to do. However, if the two adjacent pixels (FA = 0.8) belong to one
continuous fiber with angle transition of 30 degree, you want to keep the
same FA  (FA = 0.8) between the two pixels. In this case, you don't want
decrease in FA every time you interpolate pixels with slight angle
difference. 

In reality, if you do fiber tracking, angle transition greater than 40-45
degree is rejected usually. Also, average angle transition is usually very
small (<10), unless 2-2.5 mm image resolution is too large with respect to
fiber curvature. So this may not pose a large practical issue.

Anyway, as long as we don't know the exact neuroanatomy of each pixel, there
is no single "right interpolation". You just have to stick with a consistent
method. Or you can do both and apply the same manual ROI or voxel-based
method to make sure there is no significant difference depending on the
interpolation (and I doubt there would be). 

Hope it helps.

Susumu

  _____  

From: mristudio-users-bounces at mristudio.org
[mailto:mristudio-users-bounces at mristudio.org] On Behalf Of Shmuel Miron Dr
Sent: Wednesday, June 04, 2008 9:16 AM
To: mristudio-users at mristudio.org
Subject: [Mristudio-users] reslicing DTI images

Hello List members

I have a large set of DTI data with a voxel dimension of  1 x 1 x 2.6 mm
acquired in the axial position . 

However, In DTI studio the resolution on the sagital and coronal views is
not satisfactory.

Is there a way to re-slice the data (around 1900 dicom images) into voxels
of   2.6 x 2.6 x 2.6 mm in DTI studio ?

Is it possible to perform the job with a different software? 

Thanks

Dr. Shmuel Miron

Multiple Sclerosis Center

**************************************************************************

This email was sent by "SHEBA MEDICAL CENTER" Messaging System.

This email and any files transmitted with it are confidential and intended

solely for the use of the individual or entity to whom they are addressed,

If you have received this email by mistake please notify the system manager.

mailto:Support at Sheba.Health.gov.il

***This email was scanned against viruses, vandals and malicious content***

***Powered By Trend Micro - InterScan Messaging Security Suite1***
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20080604/c9d9fd94/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/jpeg
Size: 3781 bytes
Desc: not available
Url : http://lists.mristudio.org/pipermail/mristudio-users/attachments/20080604/c9d9fd94/attachment-0001.jpe 