NITRC Camino Forum: help

procstreamlines problem - too many seeds

Joanna Goc — Tue, 22 Jul 2014 15:32:42 GMT

Hi,

So I have converted a trk file (around 889419 tracts) into Bfloat using camino-trackvis and am trying to create connectivity matrix which ends up with a lot of 0s in the excel file and in the matrix itself, which I know is not correct especially since I have also done tracking with camino and then repeated the connectivity matrix.
When I run procstreamliens on the converted trk file (so the new Bfloat) somehow it sees 1334780 seeds which is substantially more than it's supposed to be and so the processing stops and there is a communication:

Processing ROI 1 of 1

Processing seed 889419 of 1334780Jul 22, 2014 5:10:39 PM misc.LoggedException logException
SEVERE: Exception in thread "main" class misc.LoggedException: No more tracts in input after processing seed 889419 of 1334780
at apps.ProcessStreamlines.execute(ProcessStreamlines.java:1060)
at apps.EntryPoint.main(EntryPoint.java:284)
Exception in thread "main" misc.LoggedException: No more tracts in input after processing seed 889419 of 1334780
at apps.ProcessStreamlines.execute(ProcessStreamlines.java:1060)
at apps.EntryPoint.main(EntryPoint.java:284)

Does anyone have any idea why this happens and how do I fix this?

Thanks for the help in advance,

Joanna

ActiveAx / impose fiber direction

Tanguy DUVAL — Sat, 03 May 2014 0:42:13 GMT

Hi,

I want to use the mmwmdinvivo model to fit my data. Those were taken in the spinal cord with diffusion directions perpendicular to the spine (XY plane).
I think the model should work on my data if I impose the fiber direction. Is there an option or can I impose theta and phi to 0 in the code?

Best,
Tanguy

RE: recommended tensor models for 1, 2, 3 tensors

Torsten Rohlfing — Thu, 14 Nov 2013 1:28:32 GMT

Huh... so that is interesting.

Let me ask you this then - our acquisition protocol has an additional, 6-direction b=500 series (on top of the 60-direction b=1000) with reverse phase-encode for unwarping.

So far I am using only the b=0 image of that series (for said unwarping), but would it be a genuine gain for the multi-tensor fit to add the b=500 images into the recon also?

Or would we need moree than 6 directions on the second shell also to get any benefit?

Thanks Phil!

Torsten

RE: recommended tensor models for 1, 2, 3 tensors

Philip Cook — Wed, 13 Nov 2013 21:09:54 GMT

I think your strategy is sensible. FA will be underestimated by the single-tensor model, but MD might be more stable.

With a single shell of DWI, there's no unique solution to the multi-tensor parameters:

http://crl.med.harvard.edu/publications/Scherrer_ISBI2010.pdf

RE: recommended tensor models for 1, 2, 3 tensors

Torsten Rohlfing — Tue, 29 Oct 2013 4:04:25 GMT

[color=#000000]So I am going to partly answer myself here...
[/color][i] [/i]

Turns out I overlooked the need to tell "track" the fact that it's multi-tensor data, as well as the maximum number of components, and now things are making more sense.

Sorry for the noise.

[i]Originally posted by Torsten Rohlfing:[/i][quote]
These seem to produce a lot of fibers for 1-tensor field, fewer for 2 tensors, and fewer yet for 3 tensors[/quote]

RE: recommended tensor models for 1, 2, 3 tensors

Torsten Rohlfing — Mon, 28 Oct 2013 19:35:43 GMT

And a follow-up question:

Would you suggest different "track" parameters for the 1, 2, and 3-tensor fields?

I am currently playing with the following:

-curvethresh 60 -anisthresh 0.1

These seem to produce a lot of fibers for 1-tensor field, fewer for 2 tensors, and fewer yet for 3 tensors.

Would this be expected?

Thanks!

Torsten

recommended tensor models for 1, 2, 3 tensors

Torsten Rohlfing — Mon, 28 Oct 2013 18:29:50 GMT

Hi Guys:

I was wondering if you could provide some guidance on good choices for tensor models using 1, 2, 3 tensors with the goal of fiber tracking.

We are using a 60-direction, b=1000 (plus 2 b=0) acquisition scheme, 2.5mm isotropic voxels.

My intuitive choices for tensor reconstruction (i.e., using modelfit) are these:

1 tensor - nldt_pos
2 tensors - pospos
3 tensors - pospospos

Is there anything obviously wrong or suboptimal with these choices?

Also, having had little luck with the FA/MD etc. maps derived from multi-tensor fields in the past, I was planning to use the multi-tensor fields for fiber tracking, but then evaluate measures such as FA along the fibers based on the single-tensor FA map (analogous for MD, L1, LT).

Do you have any words of caution regarding this plan also?

Thanks!

Torsten

Definition of Colour Code

Olivier Tannert — Wed, 28 Aug 2013 12:21:07 GMT

Hi,
I need the definition of the colour code which is used in Camino to show the directions of the Eigenvectors. I think it might be the same definition which is also used in ColFa images and Fibretracks. Can somebody tell me where i can find it? I'd appreciate that.

greets

Olivier

How to get numerical eigenvectors?

Olivier Tannert — Thu, 01 Aug 2013 15:48:57 GMT

Hey,
I need to get numerical eigenvectors of a DTI studi processed with camino. I already managed to get the .Bdouble file of the principle eigenvector but as far as I understand the tutorial this file is good for producing images with mathlab, but not for getting the real eigenvectors in every voxel. Has anyone an idea how that could be managed?

a fast answer would be apreciated

thanks

Olivier

visualization of fiber clustering result

Birkan Tunc — Wed, 21 Nov 2012 18:25:54 GMT

hi,

I run a clustering algorithm on tracts I got from "tract" command and have a cluster id for each individual tract. I want to visualize them in paraview or something similar by colouring according to their cluster ids (a single colour for each id). How can I do that?

vtkstreamlines asks for a scalar file which makes it really hard.