MRI Studio Forum » Diffeo Map

X, Y, Z flips

(4 posts)
  • Started 7 years ago by yongzhangdti
  • Latest reply from yongzhangdti

Tags:

  1. yongzhangdti
    Member

    Hello,

    I have a question about the X, Y, Z flips. When performing a linear normalization, I checked the "Flip x component" option since we used a Siemens machine and obtained the gradient table from the DICOM header. During the following non-linear normalization, I tried both conditions with x flippling and without x fliping. I produced both the colormaps and FA maps. To me, there seems no difference between two conditions in terms of FA values and fibler orientation observed on the colormaps. Just wonder if it is true. Or Do I even need to "flip x" during the linear normalization. Many thanks for any thought!

    Yong

    Posted 7 years ago #
  2. susumu
    Member

    You have to be very careful about this flipping business. In old days, there is no gradient table information in the DICOM header. We need to resort to the tables provided by manufacturers. Because the actual gradient table with respect to the imaging gradient varies dynamically depending on the image protocol (which is phase and which is readout, patient face up or face down, head first or leg first), x/y/z permutation and +/- signs need to be re-assigned for each protocol. Empirically, we knew that the x-compnent signs must be flipped for Siemens with Axial/face up/phase encode in AP/head first. Now if you read the gradient table from DICOM, these could be already correctly reassigned. If this is the case, you don't want to do any flipping business.

    As you might know, you NEED to make sure that all the vector angles are correct before you do any normalization. The best way to make sure is to do fiber tracking. If the sign is not right, the tracking doesn't make any sense. I know that this is somewhat non-theoretical solution.

    Another thing you need to know is how oblique information is handled. Siemes scanners handle it diferently between VB13 and VB15. So, you need to be careful. Currently we are testing if the information in DICOM header is before or after the oblique rotation. I hope that we will give some announcement. In any way, it is always do some testing. This you can do by scanning the same subject without and with strong oblique angles, do all the DTI processing, and perform fiber tracking and make sure the fiber looks right regardless of the oblique operation.

    Posted 7 years ago #
  3. yongzhangdti
    Member

    Hi susumu,

    Thank you very much for the detailed reply and good suggestion to perform fiber tracking to make sure +/- signs are correct! It is really helpful!

    Yong

    Posted 7 years ago #
  4. yongzhangdti
    Member

    Just wonder if the server is down today since there is no reply for the remote LDDMM processing. Thanks!

    Yong

    Posted 7 years ago #

RSS feed for this topic

Reply

You must log in to post.