3D Nasal Magnetic Resonance Image Registration by Maximization of Mutual Information

Abstract

The magnetic resonance image (MRI) is widely used in medical treatment, which is harmless for human body. However, more noise than the CT image sometimes can be a problem when doctor focus on details of some narrow structure of human body such as nasal. It is necessary to process the MRI like image fusion before using it. Subjects have to move head during a few MR scans. Image registration can help all corresponding points find their own correct position, so that it can be processed and give us more information from not only one time scan. Mutual information is a good mathematics tool in statistics and information theory. It can describe the relationship between input signal and output signal. In this paper, they are reference image data set and float image data set, and the process between them are some little bit head move and some other influence during scanning. The rigid body transformation is used to reduce the impact of head move because the nasal consisting of bones almost don’t changes shape during scan. Registered images from another scan and reference image can get together to be used in image fusion and reconstruction.