I just want to close this issue by showing my output.
It took me ~60hrs to complete tractography in a pbs_system using 2 nodes and 2 cpus. The time did not change much between using --network vs. --omatrix1 vs. --network and --omatrix1 together.
Though the *.dot matrix is 11GB!
For further info, check this screenshot: http://nmr.mgh.harvard.edu/~rdp20/fdt_probtrackx.html
Thanks Moises for your help!
Rodrigo
################################################
PREVIOUS POSTINGS:
1) That file does not contain any connectivity results, only information about the nodes of the connectivity Matrix generated ()
For each node:
<x-coord> <y-coord> <z-coord> <Number_ROI> <Number_node>
2) Did you get any other output file, or any error message?
Please, note that you are creating a huge matrix.
On 7 February 2018 at 14:28, Rodrigo Perea <[log in to unmask]> wrote:
Thanks Moises,
I def. take a look at the tools you've provided.
The output files that I get in the .../probtrackx2/* . directory are:
coords_for_fdt_matrix1
probtrackx.log
tmpnetmask
1) Could you help me undestand the output of `coords_for_fdt_matrix1`the is the following:
$cat coords_for_fdt_matrix1
92 91 26 0 1
93 90 27 0 2
94 90 27 0 3
91 91 27 0 4
.
.
.
66 84 37 113 133192
67 84 37 113 133193
66 85 37 113 133194
Is it <x-coord> <y-coord> <z-coord> <???> <number_of_connected_paths>?
2) Should I assume the command didn't complete since fdt_path is not created?
Thanks in advance,
Rodrigo
<<<<<<<<
REPLIED FROM THESE PREVIOUS POSTS:
Hi Rodrigo,
I am not sure what are you trying to compute with probtrackx, and I am not sure how many seeds you are using for tracking,
but you are asking for:
--network: Matrix NROIS x NROIS
--matrix1: Matrix Nsees X Nseeds
--opd: path distribution map
I am not sure if you need all these things. Take a look at the documentation:
https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FDT/UserGuide#PROBTRACKX_-_probabilistic_tracking_with_crossing_fibres
The first two options (network & matrix1), can be computationally very expensive. If you have an NVIDIA GPU you can use:
https://users.fmrib.ox.ac.uk/~moisesf/Probtrackx_GPU/
Moises.
On 6 February 2018 at 10:37, Rodrigo Perea <[log in to unmask]> wrote:
Hi probtrackx2 experts,
I run the following command:
/usr/pubsw/packages/fsl/5.0.9/bin/probtrackx2 --network -x .../seeds_mask_100_v1.txt -l --onewaycondition --omatrix1 -c 0.2 -S 2000 --steplength=0.5 -P 5000 --fibthresh=0.01 --distthresh=0.0 --sampvox=0.0 --forcedir --opd -s .../dmri.bedpostX/merged -m .../nodif_brain_mask --dir=.../probtrackx2
The output below (after ~15 hours) shows that there's been at least 33 volumes seeds processed.
So,
1) If seeds_mask_100_v1.txt has 100 volumes, it seems that it will take (15/33)*100 hours to finish the job?
2) If so, is there any way you could suggest to make this process run faster?
Thanks in advance,
Rodrigo
<OUTPUT>:
Log directory is: .../probtrackx2
Running in network mode
load seeds
Load bedpostx samples
1_1
1_2
1_3
2_1
2_2
2_3
nfibres : 2
nsamples : 50
Done loading samples.
Volume seeds
volume 0
volume 1
volume 2
volume 3
volume 4
volume 5
volume 6
volume 7
|