Medical image registration using a closed incompressible viscous fluid model with mutual information

Image registration is very important for a wide variety of medical image applications. It provides lots of image information for medical personnel to judge and analyze the anatomical structures and lesions. There are many image registration method being proposed. This thesis describes a non-rigid viscous model which involves a complicated, nonlinear partial differential equation. We propose to use the closed two-dimensional incompressible viscous flow model to solve the partial differential equation, We also incorporate the knowledge of mutual information into the model, from which the velocity and displacement fields are obtained. A wide variety of magnetic resonance images were used to evaluate this new method. Experimental results indicate that the proposed method successfully performed registration and provided good accuracy. Comparing to the optical-flow model, our method produced higher correlation coefficient scores. We believe that our algorithm is of potential in a wide variety of medical image registration applications.