3D Nonrigid Image Registration by Parzen‐Window‐Based Normalized Mutual Information and its Application on Mr‐Guided Microwave Thermocoagulation of Liver Tumors
Rui Xu, Yen‐Wei Chen, Shigehiro Morikawa, Yoshimasa Kurumi · 2015
The aim of medical image registration is to align two/several medical images to eliminate spatial differences of target objects like organs, tissues, or anatomical structures. The medical image registration has been researched in the past 30 years. This chapter introduces a voxel-similarity-based registration method called parzen-window-based normalized mutual information. The application of the voxel-similarity-based registration method is described on 3D nonrigid image registration for magnetic resonance (MR)-guided microwave thermocoagulation of liver tumors. The parzen-window-based normalized mutual information algorithm is proposed to estimate a continuous joint histogram, so that the normalized mutual information is differentiable. A new kernel is also designed for the parzen-window-based method. The designed kernel is a second-order polynomial kernel. The proposed parzen-window-based normalized mutual information was also successfully applied to develop a CT-MR nonrigid registration in order to assist the MR-guided microwave thermocoagulation of liver tumors.