Information acquired using different medical imaging techniques (e.g., MRI, PET, CT, etc.) can be combined to get a clear understanding of the overall condition of a patient for the purpose of diagnosis. Registering images from different modalities without a priori knowledge is difficult since the images may have very different intensity mappings and structural characteristics. This paper presents a novel approach to the multi-modal registration of medical images through the use of a priori knowledge to align medical images using an indirect mapping. The proposed algorithm uses stored information from successful alignment results to infer a relationship between the input images from different modalities. This relationship is then used to estimate the transformations needed to align the medical images together. Experimental results show that a high level of accuracy can be achieved using the proposed algorithm to align medical images from different modalities.
展开▼
