Research And Application Of Binary B-spline Registration Algorithm For Medical Images

Medical images have always been an important basis for clinical diagnosis.With the rapid development of computer hardware and software technology,medical images containing various information are constantly emerging.During the process of image acquisition by the device,due to factors such as the environment,the angle of the device,and the movement of the patient,linear and nonlinear deformation of the medical image is often caused.Among them,the non-rigid registration technology mainly performs nonlinear position transformation for image pixels,which can well solve the nonlinear deformation of the image.Therefore,a new non-rigid registration algorithm is proposed by applying the non-tensor product B-spline theory to the medical image field.Mainly completed the following work:1)A non-tensor product quadratic B-spline non-rigid registration framework based on 2-type triangulation is established.Aiming at the problem of insufficient image information extracted by tensor product B-spline rectangular subdivision,a non-tensor product quadratic B-spline algorithm of 2-type triangulation is applied to the search space of the medical image registration framework.The deformation that occurs on the X-axis and Y-axis of the image and the diagonal of the rectangle can be extracted.The normalized cross-correlation algorithm is applied to the similarity measure,and the gradient descent algorithm is applied to the search strategy.2)A non-tensor product type B-spline non-rigid registration framework for finer meshing is established.By improving the fineness of mesh division,the registration algorithm of uniform tensor product type cubic B-spline and non-tensor-product type quadratic B-spline is given.Then,the registration experiments are carried out on the high-resolution medical image data.3)A non-uniform tensor product and non-tensor product type registration framework is established.According to the idea of improving the degree of freedom of meshing,the non-uniform tensor product type quadratic B-spline,cubic B-spline and non-tensor product quadratic B-spline algorithms are given by reducing the number of nodes.The deformation equation of image pixel point offset is constructed.It also provides ideas for grid design of different types of medical images.Experimental results show that the registration framework based on the non-tensor product quadratic B-spline under uniform partition can obtain higher accuracy under the same experimental data.Figure 33;Table 5;Reference 59...