Difference between revisions of "2015 Summer Project Week:TrackerlessMRIUSFusion"

From NAMIC Wiki
Jump to: navigation, search
 
Line 28: Line 28:
 
* Initial good results with the LC2 metric constrained to a craniotomy site with 3 degrees of freedom.
 
* Initial good results with the LC2 metric constrained to a craniotomy site with 3 degrees of freedom.
 
* Given a pair of registered US-MRI images, the module calculates the similarity metric output perturbed between the two images.
 
* Given a pair of registered US-MRI images, the module calculates the similarity metric output perturbed between the two images.
* Plan to integrate image-features per Matthew Toews advice.
+
* Plan to integrate information from image-features per Matthew Toews advice.
* Working toward expanding the code to work with prostate dataset from Andre Federov
+
* Working toward expanding the code to work with prostate dataset from Andre
 
</div>
 
</div>
 
</div>
 
</div>

Latest revision as of 15:07, 24 June 2015

Home < 2015 Summer Project Week:TrackerlessMRIUSFusion


Key Investigators

  • Utsav Pardasani
  • Adam Rankin
  • Robert Owen, BK
  • Andrey Fedorov, BWH

Project Description

Objective

  • Work toward a real-time trackerless image-based registration that is constrained by a clinically relevant geometry. (With special emphasis on intra-operative neuroimaging)
  • Can support calibration / registration with tracked-systems

Approach, Plan

  • Evaluate and develop various image-similarity metrics in a geometrically constrained search space.
  • Rather than relying on an optimizer, the goal would be to find a fast similarity metric that enables a dense sampling of the objective function search space.
  • Make use of the BITE dataset for US-MRI images. http://www.bic.mni.mcgill.ca/Services/ServicesBITE

Progress

  • Initial good results with the LC2 metric constrained to a craniotomy site with 3 degrees of freedom.
  • Given a pair of registered US-MRI images, the module calculates the similarity metric output perturbed between the two images.
  • Plan to integrate information from image-features per Matthew Toews advice.
  • Working toward expanding the code to work with prostate dataset from Andre