Difference between revisions of "2012 Winter Project Week:FastInterpolation"

From NAMIC Wiki
Jump to: navigation, search
 
(4 intermediate revisions by the same user not shown)
Line 3: Line 3:
 
Image:PW-SLC2012.png|[[2012_Winter_Project_Week#Projects|Projects List]]
 
Image:PW-SLC2012.png|[[2012_Winter_Project_Week#Projects|Projects List]]
 
Image:phantom_double.png|Synthetic example used for tests.
 
Image:phantom_double.png|Synthetic example used for tests.
 +
Image:2DMix.png|Ovelay of images to register.
 +
Image:2DMix_reg.png| Registered images.
 +
Image:2DDefField.png| Computed deformation field.
 
</gallery>
 
</gallery>
  
Line 14: Line 17:
  
 
<h3>Objective</h3>
 
<h3>Objective</h3>
*Goal: Create a deformable registration approach in the presence of large deformations(e.g. in the presence of abnormalities, between different patients)
+
*Goal: Create a deformable registration approach for computing large deformations(e.g. in the presence of abnormalities, between different patients)
 
*A stochastic registration algorithm has been implemented.
 
*A stochastic registration algorithm has been implemented.
 
*Bottleneck is performing image interpolation quickly.  
 
*Bottleneck is performing image interpolation quickly.  
Line 27: Line 30:
 
*Take advantage of the structure of deformation (parameterized deformation field).
 
*Take advantage of the structure of deformation (parameterized deformation field).
 
*Perform approximate interpolation.
 
*Perform approximate interpolation.
*Make algorithmic improvements to make algorithm usable on large 3D datasets.
+
*Make algorithmic improvements to allow for large 3D datasets to be used.
  
 
</div>
 
</div>
Line 34: Line 37:
  
 
<h3>Progress</h3>
 
<h3>Progress</h3>
 +
*Proposed an optimization ~100x faster than previous version
 +
*Tested on the 2D phantom image above.
  
 
</div>
 
</div>

Latest revision as of 23:14, 12 January 2012

Home < 2012 Winter Project Week:FastInterpolation

Key Investigators

  • Ivan Kolesov : Georgia Institute of Technology
  • Greg Sharp : MGH
  • Allen Tannenbaum : Boston University

Objective

  • Goal: Create a deformable registration approach for computing large deformations(e.g. in the presence of abnormalities, between different patients)
  • A stochastic registration algorithm has been implemented.
  • Bottleneck is performing image interpolation quickly.


Approach, Plan

  • Take advantage of the structure of deformation (parameterized deformation field).
  • Perform approximate interpolation.
  • Make algorithmic improvements to allow for large 3D datasets to be used.

Progress

  • Proposed an optimization ~100x faster than previous version
  • Tested on the 2D phantom image above.

Delivery Mechanism

This work will be delivered to the NA-MIC Kit as a

  1. Slicer Module (via PLUS and OpenIGTLink)


References