Nonrigid registration of medical image by Maxwell model of viscoelasticity

Nonrigid medical image registration has many potentially applications for diagnosis and monitoring disease progression in the clinic, and is very hot in computational anatomy. However, there are not very efficient methods to solve the problem until now. In this paper, we proposed a nonrigid medical image registration approach based on the physics laws. There are two novelties of proposed method. One is that we model the template image as a viscoelastic matter since the mechanical behaviors of brain are shown to be viscoelastic. The local shape variations are assumed to meet the property of Maxwell model of viscoelasticity, and the deformable fields are constrained by the corresponding partial differential equations. The other is that an adaptive force is introduced to decrease the computation cost. We applied the proposed algorithm to 2D simulated data and 3D real data of different subjects, and compared the results of the proposed approach to those obtained by affine registration and other physics based method, fluid method. We found the results of proposed method were satisfied in accuracy and speed.