A Generic Mutual Information Based Algorithm for Spatial Registration of Multi-Modal Medical Images

作者：Feng Lin^*; Datta Abhik

来源：Journal of Medical Imaging and Health Informatics, 2011, 1(2): 131-138.

DOI：10.1166/jmihi.2011.1025

• Image Registration
• Multiple Modalities
• Mutual Information
• Distance Metric
• Optimization

Spatial registration of multi-modal medical images can generate a new one that can be used to better study the functioning of the tissues and organs. In this study, two broad categories of registration techniques in the literature are reviewed, i.e., using image intensity gray level information, and using landmarks and image features. Previous research has been in the use of features such as the corresponding corners and boundaries (landmarks) of the structures for registration, but the results have been unsatisfactory, especially when the images present low Signal-to-Noise ratio. To solve the problem, we propose a new metric for registration of multi-modal images. Both intensity and feature space mutual information (MI) are used, and the Kappa Statistics is combined. Based on that, we have invented a generic algorithm that effectively combines image features with the intensity information. We present a detailed description of the construction of the new metric, in which mutual information is used for intensity and other feature spaces. Then we discuss the key techniques of the registration algorithm using the new metric. We consider the factors that can affect the registration process, such as effective bin size estimation, interpolation, optimization techniques, orientation and image contrast. The effectiveness of the measure for PET-CT registration is then discussed in detail along with experimental results and comparisons.

• 出版日期2011-6
• 单位南阳理工学院