Lung PET-CT Medical Image Registraion And Fusion

With the development of modern medical imaging, modern medical imaging technologies is emerging, such as X-ray computed tomography (CT), magnetic resonance imaging (MRI), functional magnetic resonance imaging (FMRI), positron emission tomography (PET), single photon emission computed tomography (SPET), etc. These have become the essential medical digital imaging tools in the modern medical diagnostic. In the field of medical imaging, multi-modality medical image analysis becomes an important part of computer-aided diagnosis. Not only academia paid attention to the multimodality medical image registration and fusion, but also the medical world too, since its formation and widespread. This is mainly due to its wide applications, ranging from location the lesion, preoperative evaluation of treatment programs, to tracking pathological changes and evaluation of the treatment.In this paper, commonly used medical image registration and fusion methods were studied. In terms of medical image registration, we focused our deeply study on the development and predecessor’s improvement of mutual information methods. In the terms of medical image fusion, multi-scaled image fusion is further discussed. On the basis of this, depth exploration and experimental research are played according to the actual situation.Firstly, for the problem of local minima in PET-CT medical image registration based on mutual information, this paper solves it from the following two aspects. First a new measurement of registration combing spatial information (mutual variance) and normalized mutual information was proposed. Second, the new search strategy combining of PSO (particle Swarm Optimization) and Powell algorithm effectively reduce the generation of local extreme situations. Taking into account the complexity of the calculation based on mutual information registration method, we use a coarse to fine two-stage registration method to improve the speed of registration to some extent.Then, after the registration of lung CT and PET images, we use a non-subsampled Shearlet transform(NSST) fusion method for lung PET-CT image fusion. Firstly using NSST multi-direction, multi-scale decomposition to obtain sub-band coefficients of each band-pass, secondly using different fusion rules to fuse the coefficients, and finally through inverse NSST transformation to achieve the fused images. Comparative experiments show that this algorithm in detail and contrast of the fused image are better than fusion method based on wavelet transform.