SlicerDMRI / SlicerDMRI

Diffusion MRI analysis and visualization in 3D Slicer open source medical imaging platform.
http://dmri.slicer.org
Other
77 stars 27 forks source link

Different results generated by interactive Tractography Seeding and Tractography ROI Seeding #129

Open josephmje opened 4 years ago

josephmje commented 4 years ago

Hi,

I am running a nightly version of Slicer built on 2018-08-17 and Slicer dMRI built from commit the following commit.

I noticed that I get very different tractography results when using the interactive TractographySeeding module compared to the Tractography ROI Seeding module. I have attached screenshots below. I was wondering whether this difference is expected and whether an alternate method is being applied in the interactive mode?

Diffusion -> Tractography -> Tractography Seeding

image

Diffusion -> Tractography -> Region-based -> Tractography ROI Seeding

image

The ROI Seeding module seems to replicate what I get when running TractographyLabelMapSeeding in the command line using the following command:

Slicer --launch TractographyLabelMapSeeding \
    --inputroi mask.nrrd \
    --clthreshold 0.3 \
    --minimumlength 20 \
    --maximumlength 800 \
    --thresholdmode FractionalAnisotropy \
    --stoppingvalue 0.25 \
    --stoppingcurvature 0.7 \
    --integrationsteplength 0.5 \
    --label 1 \
    --useindexspace \
    --seedspacing 2 \
    dti.nrrd \
    SlicerTractography.vtk

Thank you for your help

zhangfanmark commented 4 years ago

Hi Michael,

The two modules are based on the same underling algorithm. If all the parameters are being set the same, they should give the same tracking result.

Is the same mask being use in both methods?

Also, please try to use the "line" visualization mode with 100% fibers displayed. The "tube" visualization involves a downsample process.

Eegards, Fan

josephmje commented 4 years ago

Thank you for your reply Fan! Yes, all parameters have been set the same and using the same mask.

josephmje commented 4 years ago

Hi Fan, I'm not sure if this is a dataset-specific issue. Below I've posted a screenshot of the same data but with line visualization mode. The interactive mode is on the left and the command line/ROI mode is on the right.

spasd

I also tried this with the pre-processed data from the Philadelphia Neurodevelopmental Cohort (PNC) and get similar results.

pnc