AFQ_WholeBrainTractography image looks sparse

70 views
Skip to first unread message

Steven Meisler

unread,
Apr 7, 2020, 3:19:11 PM4/7/20
to Automated Fiber Quantification (AFQ)
Hello,

I am running the AFQ_example.m code on outputs from Vistasoft's DTIInit pipeline. Across all of the subjects I have looked at, my images for the whole brain tractography are very sparse compared to running the same code on the AFQ subject data included with the installation (see attached picture for a comparison). Both images include 1,000 fibers, based on the 'num_fibers' argument to AFQ_WholeBrainTractography. The outputs from Vistasoft (MD, FA, RBG etc.) all seem reasonable, so I am wondering if anyone has had similar issues, and if so where the problem was (data acquisition, Vistasoft, or AFQ)?

Thank you,
Steven

wbt_images.png

Jason Yeatman

unread,
Apr 7, 2020, 3:24:48 PM4/7/20
to automated-fiber-q...@googlegroups.com
Looks like you probably have your gradient directions flipped so tractography isn’t working properly. Search through this archive and you will find various recs on that

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantif...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-fiber-quantification-afq/b61518d7-12fb-436a-80d1-06bc33a53c35%40googlegroups.com.
--
Sent from a mobile device. Please excuse typos

Steven Meisler

unread,
Apr 7, 2020, 4:27:02 PM4/7/20
to Automated Fiber Quantification (AFQ)
Thanks for the quick reply! I found the rotate bvec parameter for Siemen's data and reran Vistasoft with that. Unfortunately, the image still looks a bit sparse (attached image). I can try to see if other parameters are set correctly.


On Tuesday, April 7, 2020 at 3:24:48 PM UTC-4, Jason wrote:
Looks like you probably have your gradient directions flipped so tractography isn’t working properly. Search through this archive and you will find various recs on that
On Tue, Apr 7, 2020 at 12:19 PM Steven Meisler <smei...@g.harvard.edu> wrote:
Hello,

I am running the AFQ_example.m code on outputs from Vistasoft's DTIInit pipeline. Across all of the subjects I have looked at, my images for the whole brain tractography are very sparse compared to running the same code on the AFQ subject data included with the installation (see attached picture for a comparison). Both images include 1,000 fibers, based on the 'num_fibers' argument to AFQ_WholeBrainTractography. The outputs from Vistasoft (MD, FA, RBG etc.) all seem reasonable, so I am wondering if anyone has had similar issues, and if so where the problem was (data acquisition, Vistasoft, or AFQ)?

Thank you,
Steven

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantification-afq+unsub...@googlegroups.com.
subject_sparse.png

Steven Meisler

unread,
Apr 7, 2020, 7:48:07 PM4/7/20
to Automated Fiber Quantification (AFQ)
I have tried toggling both of the bvec rotation parameters (Xform and Rx) individually and rerunning the subject, and the tractography maps unfortunately still look sparse. I also made sure the phase encoding direction matched that from the T1 information file. When running AFQ, I see that 34962 fibers passed length threshold of 50.0 (out of 240768 seeds). They had a mean length of 66mm. I will continue to investigate this further, but does this amount of exclusion and the resulting average fiber length seem normal?

Thanks,
Steven

Michal Ben-Shachar

unread,
Apr 8, 2020, 7:49:40 AM4/8/20
to automated-fiber-q...@googlegroups.com
Hi Steven,
The streamlines still look wrong (not just sparse). Many fibers are excluded because they stop short, probably because the directions are wrong. This is not normal. In all likelihood, there is still a problem with the gradient direction matrix. 
Maybe this old page will help:
michal
 



On Wed, Apr 8, 2020 at 2:48 AM Steven Meisler <smei...@g.harvard.edu> wrote:
I have tried toggling both of the bvec rotation parameters (Xform and Rx) individually and rerunning the subject, and the tractography maps unfortunately still look sparse. I also made sure the phase encoding direction matched that from the T1 information file. When running AFQ, I see that 34962 fibers passed length threshold of 50.0 (out of 240768 seeds). They had a mean length of 66mm. I will continue to investigate this further, but does this amount of exclusion and the resulting average fiber length seem normal?

Thanks,
Steven

On Tuesday, April 7, 2020 at 4:27:02 PM UTC-4, Steven Meisler wrote:
Thanks for the quick reply! I found the rotate bvec parameter for Siemen's data and reran Vistasoft with that. Unfortunately, the image still looks a bit sparse (attached image). I can try to see if other parameters are set correctly.

On Tuesday, April 7, 2020 at 3:24:48 PM UTC-4, Jason wrote:
Looks like you probably have your gradient directions flipped so tractography isn’t working properly. Search through this archive and you will find various recs on that
On Tue, Apr 7, 2020 at 12:19 PM Steven Meisler <smei...@g.harvard.edu> wrote:
Hello,

I am running the AFQ_example.m code on outputs from Vistasoft's DTIInit pipeline. Across all of the subjects I have looked at, my images for the whole brain tractography are very sparse compared to running the same code on the AFQ subject data included with the installation (see attached picture for a comparison). Both images include 1,000 fibers, based on the 'num_fibers' argument to AFQ_WholeBrainTractography. The outputs from Vistasoft (MD, FA, RBG etc.) all seem reasonable, so I am wondering if anyone has had similar issues, and if so where the problem was (data acquisition, Vistasoft, or AFQ)?

Thank you,
Steven

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantif...@googlegroups.com.
--
Sent from a mobile device. Please excuse typos

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantif...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-fiber-quantification-afq/1f7a4c06-0147-44c4-aa38-9d943884298e%40googlegroups.com.

Steven Meisler

unread,
Apr 9, 2020, 11:31:26 AM4/9/20
to Automated Fiber Quantification (AFQ)
Hi Michal,

Does this look better?

Thanks,
Steven


On Wednesday, April 8, 2020 at 7:49:40 AM UTC-4, Michal Ben-Shachar wrote:
Hi Steven,
The streamlines still look wrong (not just sparse). Many fibers are excluded because they stop short, probably because the directions are wrong. This is not normal. In all likelihood, there is still a problem with the gradient direction matrix. 
Maybe this old page will help:
michal
 



On Wed, Apr 8, 2020 at 2:48 AM Steven Meisler <smei...@g.harvard.edu> wrote:
I have tried toggling both of the bvec rotation parameters (Xform and Rx) individually and rerunning the subject, and the tractography maps unfortunately still look sparse. I also made sure the phase encoding direction matched that from the T1 information file. When running AFQ, I see that 34962 fibers passed length threshold of 50.0 (out of 240768 seeds). They had a mean length of 66mm. I will continue to investigate this further, but does this amount of exclusion and the resulting average fiber length seem normal?

Thanks,
Steven

On Tuesday, April 7, 2020 at 4:27:02 PM UTC-4, Steven Meisler wrote:
Thanks for the quick reply! I found the rotate bvec parameter for Siemen's data and reran Vistasoft with that. Unfortunately, the image still looks a bit sparse (attached image). I can try to see if other parameters are set correctly.

On Tuesday, April 7, 2020 at 3:24:48 PM UTC-4, Jason wrote:
Looks like you probably have your gradient directions flipped so tractography isn’t working properly. Search through this archive and you will find various recs on that
On Tue, Apr 7, 2020 at 12:19 PM Steven Meisler <smei...@g.harvard.edu> wrote:
Hello,

I am running the AFQ_example.m code on outputs from Vistasoft's DTIInit pipeline. Across all of the subjects I have looked at, my images for the whole brain tractography are very sparse compared to running the same code on the AFQ subject data included with the installation (see attached picture for a comparison). Both images include 1,000 fibers, based on the 'num_fibers' argument to AFQ_WholeBrainTractography. The outputs from Vistasoft (MD, FA, RBG etc.) all seem reasonable, so I am wondering if anyone has had similar issues, and if so where the problem was (data acquisition, Vistasoft, or AFQ)?

Thank you,
Steven

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantification-afq+unsub...@googlegroups.com.
--
Sent from a mobile device. Please excuse typos

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantification-afq+unsub...@googlegroups.com.
dwi_flipped.png

Jason Yeatman

unread,
Apr 9, 2020, 5:56:33 PM4/9/20
to automated-fiber-q...@googlegroups.com
That looks good to me. Have you passed that through the segmentation step to see if you get good bundles?

To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantif...@googlegroups.com.
--
Sent from a mobile device. Please excuse typos

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantif...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantif...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-fiber-quantification-afq/92eef778-b465-493d-87a1-9cb4dd6296d7%40googlegroups.com.

Meisler, Steven

unread,
Apr 9, 2020, 6:22:34 PM4/9/20
to automated-fiber-q...@googlegroups.com
image.png
The segmentation results aren't as populated as expected (with no ISOF fibers being found).

To view this discussion on the web visit https://groups.google.com/d/msgid/automated-fiber-quantification-afq/CANbxo5c0wROyjf9xa%2BSVyS%2Bx33uBK5udrUmKKZybzU0R6b%2BiSw%40mail.gmail.com.


--

Steven Meisler

unread,
Apr 9, 2020, 6:23:44 PM4/9/20
to Automated Fiber Quantification (AFQ)
image.png
The segmentation results aren't as populated as expected, and no IFOF fibers were found.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantification-afq+unsub...@googlegroups.com.
--
Sent from a mobile device. Please excuse typos

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantification-afq+unsub...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Automated Fiber Quantification (AFQ)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-fiber-quantification-afq+unsub...@googlegroups.com.

Steven Meisler

unread,
Apr 9, 2020, 6:24:42 PM4/9/20
to Automated Fiber Quantification (AFQ)
Image may have not sent as first, so I am attaching it here.
segmentation.png
Reply all
Reply to author
Forward
0 new messages