With the development of computer science and imaging technology, medical image processing and analysis is now obtaining more and more attention. It has now become a rapidly developing new area of interdisciplinary sciences and a very important component of modern medicine. Image registration is an important and fundamental research content in the field of medical image processing and analysis. The goal of image registration is to find an one to one mapping relationship of two or more image spaces with different time, different perspectives, as well as different imaging modes, thus the correspondences of same pixels or voxels in different images can be set up. Image registration now plays a crucial role in the modern clinical diagnosis and treatment: by multi-modal image registration, multi-modal images can be fused in the same visual image to express both anatomy, physiology and pathology information; by the registration of the same organ in different periods, doctors can monitor the development of the disease and make treatment plans; besides that, the registration of images with different time and different modes is also an important and indispensable step for image-guided surgery and treatment. In recent years, the development of medical image registration is very fast,general image registration algorithms have reached maturity. However, since that medical images are derived from a wide range of source with a diversity of applications, general methods are difficult to obtain satisfactory results for specific problems restricted by the image content, quality, and specific demands. Therefore, the current medical image registration algorithms are refined to focus on the specific applications to get better performances. Still, there are some major difficulties in current medical image registration algorithms, including: 1.Medical image registration is very computationally expensive: for the registration of three-dimensional volume data and non-linear transformation model based registration, this is especially time consuming. High-dimensional transformation parameters search space for the global optimal value will usually spend a few minutes or even tens of minutes. The shortcomings of a long running time limit further applications of image registration in real-time computer-aided diagnosis and treatment. 2.Fast and accurate registration for low-quality medical images: limited by the imaging technology and disease progression, some medi...
修改评论