TubeTK/Events/2010.10.14

Danielle

• Implementation of diffusion-based regularization
  • itkImageToImageDiffusiveDeformableRegistrationFilter
  • itkImageToImageDiffusiveDeformableRegistrationFunction
    • Has two functions - a regularization function and an intensity difference function, and sums their results
  • Extends itkPDEDeformableRegistrationFilter/Function, with similar modifications as Andinet's itkAnisotropicDiffusionTensorImageFilter/itkAnisotropicDiffusionTensorFunction (which extend itkFiniteDifferenceImageFilter/itkFiniteDifferenceFunction)
• Implementation of regularization smoothing in the tangential direction:
  • Artificial deformation field:
    • Vectors are [-0.5, 0.5, 0] or [0.5, 0.5, 0], plus white noise of variance 0.1
    • Known normal, for now use w = 1 (smoothing entirely in tangential plane)

• • Registration is regularization only (20 iterations, time step 0.05)
  • Original deformation field:

• • Deformation field after regularization:
    • Note border is maintained

• Future
  • Boundary conditions
  • Regularization smoothing in normal direction
  • Add intensity distance function (simple using existing finite difference functions)
  • Compute weighting term = distance to organ border
  • Extract organ border normals from input data