PET-CT Medical Image Fusion Based On Clone Selection Algorithm

Medical image fusion, a subject which involves information science, computer science and medical imaging, is a new hot spot in medical image processing. In this paper, the defects are studied deeply, which is existing present PET-CT medical image fusion algorithms, and on this foundation PET-CT medical image fusion based on clone selection algorithm is proposed. Firstly, starting from space transformation, the difficult problem in PET-CT medical image registration based on mutual information is studied especially, which is that mutual information function is often unsmooth, there are lots of local extreme, which will involve local optimization. Therefore normalized mutual information is employed as registration function, and clone selection algorithm is also introduced to overcome the premature convergence. After PET-CT medical image is registrated, the PET-CT medical image fusion is realized based on pixel. Experiment result shows that the method does not require any prior assumption, medical knowledge, segmentation, feature extraction or other preprocessing. It can be applied in both rigid and non-rigid medical image fusion with the same good performance. Fusion precision which is improved highly by clone selection algorithm reaches to sub-pixel. Especially in the condition of the loss existing in one of medical images to be registrated, it also can obtain precise registration parameters. Non-evasive, most accurate and efficient are satisfied effectively in clinical application.