103 Brain Image Registration by Maximization of Mutual Information Using Particle Swarm Optimization

Image registration has an important role in medical image analysis and is a process in which a spatial transformation is used to align a floating image to a reference image; taken from same scene by different methods. This is done by optimizing a similarity measure with a suitable algorithm. Mutual information, MI, is one of the most popular and reliable method in this respect. Using the normal local optimization methods to find the MI is not very suitable because they are usually very sensitive to initial values as well as they may be trapped in local minimums. In this paper we present a new optimization method to register MRI and CT scans of the brain. It is a general PSO algorithm to find the maximum of a new form of normalized mutual information and show the results for the images with the rigid deformation.