Segmentation And Reconstruction Of Medical Images Based On Parametric Level Set

With the development of medical technology, optical molecular imaging occupies an important position in modem medicine. How to segment and reconstruct medical image accurately is an important subject of modern medical imaging research. Current image segmentation and optical image reconstruction technology are mainly carried out in the grid framework (such as voxel and finite element), and it exists many problems on complex topological structure evolution. The mesh-less methods is flexible in calculation, which has great advantages in dealing with topology structure evolution. With the development of medical imaging technology, it is necessary to study the application of mesh-less method in the segmentation and reconstruction of complex medical image.In this paper, we studied image segmentation and optical image reconstruction algorithm based on mesh-less parametric level set method, we first introduce the basic concept and developing situation of image segmentation and reconstruction algorithm, analyze the principle of level set algorithm, radial basis function and the inverse problem of reconstruction method, propose segmentation and reconstruction algorithm based on parametric level set using radial basis function. We design the algorithm, which is proved by computer simulations and experimental results.The segmentation algorithm of C-V model based on radial basis function is realized, and we improved the distribution of basis function centers, which reduces the interference on the segmentation boundary. The method effectively improves the segmentation boundary of parametric C-V level set model. In addition, we design a mesh-less level set reconstruction algorithm based on compactly supported radial basis functions, using coefficients of radial basis functions as the unknown parameter of the target, which reduces the target’s dimension, realize the reconstruction of FMT light source. Compared with the traditional finite element method, the method in this paper avoids the interpolation process on the node, it shows greater convergence, and performance good in accuracy and stability.