A New Method Based on Contour Feature for Multi-modality Medical Image Registration

Multi-modality medical image registration and fusion have important applications in clinical diagnosis and therapy planning. It is essential to accurately align two images from different modalities prior to any operation of fusion. This paper presents an SVD-ICP (Single Value Decomposition-Iterative Closest Points) method to register brain images based on contour feature, which combines the advantages of the speed of SVD analytical optimization and the precision of iterative search to solve the problem of image contour points matching. It uses feature sampling and accelerating algorithm to reduce computation time. The method to extract the contour is semiautomatic so that the accuracy and reliability are assured. It is applicable to multi-modality medical image registration, the original SVD-ICP algorithm is in fact an appropriate solution to the problem of n-Dimension space points matching. Our experiments on CT-MRI and PET-MRI registration prove that this method is effective.