Difference between revisions of "2017 Winter Project Week/LORDWI"

From NAMIC Wiki
Jump to: navigation, search
Line 24: Line 24:
 
<!-- Objective bullet points -->
 
<!-- Objective bullet points -->
 
* '''Objective''': Evaluate/validate a density-based non-rigid registration framework for DWI (see the link below for a paper on similarity measure used). <br>''In short: Is this a good registration?''
 
* '''Objective''': Evaluate/validate a density-based non-rigid registration framework for DWI (see the link below for a paper on similarity measure used). <br>''In short: Is this a good registration?''
−
* '''Short description''': The model is based on Free-Form Deformation B-splines where the directions are updated using the normalized Jacobian. B-spline interpolation is used spatially, the Watson Distribution is used directionally. Histogram is smoothed. Similarity is NMI and optimisation is L-BFGS.
+
* '''Short description''': The model is based on Free-Form Deformation B-splines where the diffusion gradient directions are updated using the normalized Jacobian. B-spline interpolation is used spatially, the Watson Distribution is used (gradient) directionally. Histogram is smoothed. Similarity is NMI and optimisation is L-BFGS.
 
|
 
|
 
<!-- Approach and Plan bullet points -->
 
<!-- Approach and Plan bullet points -->

Revision as of 00:43, 9 January 2017

Home < 2017 Winter Project Week < LORDWI

Key Investigators

  • Henrik Groenholt Jensen, UCPH
  • Lauren J. O'Donnell, BWH
  • Tina Kapur, BWH
  • Fan Zhang, BWH
  • Carl-Fredrik Westin, BWH

Project Description

Objective Approach and Plan Progress and Next Steps
  • Objective: Evaluate/validate a density-based non-rigid registration framework for DWI (see the link below for a paper on similarity measure used).
    In short: Is this a good registration?
  • Short description: The model is based on Free-Form Deformation B-splines where the diffusion gradient directions are updated using the normalized Jacobian. B-spline interpolation is used spatially, the Watson Distribution is used (gradient) directionally. Histogram is smoothed. Similarity is NMI and optimisation is L-BFGS.
  • Discuss best ways to validate results (tractography, biomarkers, synthetic data, phantoms, others?). So far we have visually tested inter-subject registrations of HCP data, intra-subject multi-shell, and intra-subject on child brain tumor subjects.
  • Figure out anyone is in need of DWI registration for testing, have a dataset already registered (for comparison), or can help with anatomical landmarks.
  • Consider if Slicer can be used in tandem for evaluation.
  • This method introduces Mutual Information to nonrigid registration of DWI. Discuss if other similarity measures (e.g. correlation measures) would be better suited for specific problems.

Background and References

Retrieved from "https://www.na-mic.org/w/index.php?title=2017_Winter_Project_Week/LORDWI&oldid=94508"