Link Progress

Back to NA-MIC Collaborations, Harvard DBP 2, DBP2:Harvard:Brain_Segmentation_Roadmap

Stochastic Tractography Progress

Status

Stochastic tractography requires careful maintenance and on-going development improving its usability and performance. 

To get the latest module documentation goes on the Slicer 3.4 documentation

A - Development of tractography workflow 

• DWI image preprocessing (summer week 2009 project)
  • denoising (Rician) - currently Gaussian (FWHM = Full Width Half Maximum)
  • Eddy currents corrections
  • EPI distortion correction
• WM mask & ROIs creation (integrated)
• tensor estimation & visualization (integrated)
• tracts reconstruction & visualization (integrated)
• connectivity maps creation & visualization (integrated)

B - Visualization of diffusion properties - test utilities 

• reveal more features related to diffusion - develop visualization
• finalize batch - testing different parameters of the stochastic tractography with different datasets - wrap up of the results

C – Study of limit cases (tract reconstruction issue) 

• transformation issues between 3 spaces (IJK, RAS, measurement frame)

Reconstruction of helix

• optimal parameter setting e.g. number of path per voxel

Arcuate reconstruction with only 1 path per voxel

• reduction of sparse density/connectivity map with noisy dataset (1.5 Tesla)

Density map for cingulum based on 1.5 Tesla scan

• sensitivity to inputs (WM/ROIs) - needing criteria definition and algorythmic extensions as shown on the following example

No connection with restrictive white matter mask and misregistered ROIS (left sagittal view)

Connection appears when ROI vicinity and brain mask (notion of unrestrictive 'white matter' mask) are applied (left sagittal view)

No connection with restrictive white matter mask and misregistered ROIS (left superior view)

Connection appears when ROI vicinity and brain mask (notion of unrestrictive 'white matter' mask) are applied (left superior view)

• further algorythmic extensions if needed for dealing with specific reconstruction cases
  • ROI optimization
  • moving ROI

Updates/Progress

A - Development of tractography workflow 

• initial skeleton incorporating already existing stochastic tractography components (done)
• test Slicer GUI access with python (done)
• implement tractography algorythm in python (based on Ola Friman Matlab implementation) (done)
  • test based on sample dataset show identical computational results between Matlab and python implementation (done)
  • multiprocessing (and clustering) supported (done)
  • roi filtering from region A to region B and B to A (done)
• module release ready (done)
• bugfix for release 3.4 (done)

B - Visualization of diffusion properties 

• create visual map showing regions defined by WM mask where FA (fractional anisotropy is upper the defined threshold e.g. 0.3) and where the mask forbids (defining as a non white matter region) the stochastic algorythm to go in (done)

C – Study of limit cases (tract reconstruction issue) 

• relate C to B

Schedule

• 08/2008 - Introduction to tractography, Slicer software, articles related to tractography.
• 09/2008 - Setup of Slicer3, first tests with stochastic tractography, algorythm implementation, work with training datasets
• 10/2008 - Introduction of python interpreter for implementation of batch utility + diffusion features visualization
• 11/2008 - Completion of batch utility, testing for NAMIC meeting preparation, tractography workflow first prototype ready
• 12/2008 - Wrap up, finalization of NAMIC meeting presentation for the stochastic tractography module
• 04/2009 - Module release ready for Slicer 3.4