I found the issue. Apparently the direction matrix is not applied to the image gradients. This is also the cause for the previous issue with the HCP subject. I fixed the problem in the current master, so all datasets should work fine without any flipping.
If you are not working on the current master, a workaround would be to set the image direction matrix to the identity matrix. But I recommend updating to the current master.
Peter
________________________________
Von: Lidia Konopleva <***@gmail.com>
Gesendet: Mittwoch, 29. März 2017 11:28
An: Neher, Peter
Betreff: Re: AW: [mitk-users] Qball multishell reconstruction
Hi!
Here is a link with data on google drive. Folder contains original dataset and my dataset, adapted foe MITK and with only one b-factor. Thank you for your help!??
https://drive.google.com/open?id=0BzcLAMqH6ZTcR2c3MUkxa3RuaTA
Lidia
2017-03-29 5:43 GMT+00:00 Neher, Peter <***@dkfz-heidelberg.de<mailto:***@dkfz-heidelberg.de>>:
Hi Lidia,
I'm having trouble downloading the datasets from the HCP website. Can you send me the dataset so I can check out your issue?
Peter
Von: ????? ????????? [mailto:***@gmail.com<mailto:***@gmail.com>]
Gesendet: Montag, 27. März 2017 21:36
An: Neher, Peter <***@Dkfz-Heidelberg.de>
Betreff: Re: AW: [mitk-users] Qball multishell reconstruction
It's MGH_1020. Yes, I tried to use different orientations of gradient directions. When I flip them, orientations of FODs in transversal and coronal planes are wrong.
27 ????? 2017 ?., ? 22:09, Neher, Peter <***@Dkfz-Heidelberg.de<mailto:***@Dkfz-Heidelberg.de>> ???????(?):
Which subject ID is your dataset? Maybe I can check what's going wrong here.
Did you try flipping the gradients?
Von: Lidia Konopleva [mailto:***@gmail.com]
Gesendet: Montag, 27. März 2017 14:19
An: Neher, Peter <***@Dkfz-Heidelberg.de<mailto:***@Dkfz-Heidelberg.de>>
Betreff: Re: [mitk-users] Qball multishell reconstruction
Thank you! it works for multishell data. I also have one more HCP dataset (year 2014), i think it has a different acquisition scheme. FODs from Qball reconstruction and also tensors principal directions are fine in coronal and transversal planes, but there is something strange with principal directions of tensors in saggital plane in gyrus . Can I fix this problem with transformation of gradient directions?
2017-03-27 9:47 GMT+00:00 Neher, Peter <***@dkfz-heidelberg.de<mailto:***@dkfz-heidelberg.de>>:
Hi Lidia,
MITK internally uses a different gradient convention than the HCP. We are planning on fixing it, but currently you have to flip the gradient vectors manually in the Y-dimension (y --> -y) before processing the HCP datasets in MITK.
If you are using the current master of MITK, you can use the "Gradients"-Tab of the Preprocessing view to flip the gradient directions. Otherwise you have to do it manually in the .bvecs file.
Did that solve your problem?
Best,
Peter
Von: ????? ????????? [mailto:***@gmail.com<mailto:***@gmail.com>]
Gesendet: Montag, 27. März 2017 10:33
An: mitk-***@lists.sourceforge.net<mailto:mitk-***@lists.sourceforge.net>
Betreff: [mitk-users] Qball multishell reconstruction
Hi!
I'm trying to import Human Connectome Project multishell Data to MITK as fsl files. But when I do Qball multi shell reconstruction of data, there are some mistakes in FODs orientation. Gradients directions are in fsl measurement frame. Could you help me with this problem, please?