CNN-Driven Quasiconformal Model for Large Deformation Image Registration
This project attempts to perform a large deformation image registration with convolution neural network. For the sake of well-posedness, there are certain mathematical constraints. For instance, the registration has to be deformable, bijective and orientation preserving. Unlike other existing method, this project will challenge some real medical examples of extreme deformation that other methods never could.
Show that our novel model can deal with extremely large deformation image registration.
Show how CNN could aid the process.
Lok Ming, LUI, The Chinese University of Hong Kong
Ka Chun, LAM, California Institute of Technology