Publication Details
Yixun Liu and Nikos Chrisochoides.
Published in International Workshop on Machine Learning in Medical Imaging (MLMI), the 13th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), September, 2010
Abstract
Mutual information (MI) based registration methods are susceptible to the variation of the intensity of the image. We present a multi- modality MRI-CT non-rigid registration method by combining K-means clustering technique with mutual information. This method makes use of K- means clustering to determine variant bin sizes in CT image. The resulting clustered (labeled) CT image is non-rigidly registered with MRI by modeling the underlying movement as Free-Form Deformation (FFD). We compare this Cluster-to- Image registration method with Image-to-Image and Cluster-to-Cluster methods. The preliminary experiment shows this method can increase the accuracy of non-rigid registration.