Research On Spine MRI Image Segmentation Based On Knowledge

Along with the development of medical image technology, magnetic resonance imaging (MRI) has become a hotspot and has been widely applied to the medical image. MRI is popular for its noninvasive nature and offering high soft-tissue contrast. And along with the development of MRI, research on the subsequent processing technology becomes increasingly urgent and has become a hotspot nowadays. At present the brain MRI image processing technology has already become more and more mature, and spine MRI image processing area also gradually get some attention and development. Diseases on the spine have persecuted humans, and the number of people with spinal disease is increasing year by year. Therefore, research on spinal MRI image post-processing technology is of great significance. Image segmentation is the key technology in image processing and analysis, and medical image segmentation is the foundation of the subsequent operation on normal tissue and pathological organization such as 3D visualization, operation simulation and graphics guiding the surgical operation. The accuracy of the segmentation is quite important for doctors to determine the real situations of the disease and make corresponding diagnosis programs. In this article, we divided the boundary of the vertebra. Our main goal is to improve the accuracy of the spine MRI image segmentation and reduce human and computer interaction. Accuracy is more important.When we initialize snake, it would be more accurate if we point much more points. However, more initial points will increase the user's work largely. In this paper we choose four vertexes as initial points to reduce human and computer interaction and increase the accuracy of segmentation by improve the segmentation algorithm. Deformable model is one of the most appropriate methods on medical image analysis and GVF snake is the classic algorithm of deformable model. In this paper we apply the deformation model method, traditional GVF snake, to segment vertebra firstly. Spine MRI image is fuzzy influenced by the noise and GVF converges too fast. These problems with the nature of GVF make segmentation results not smooth and accurate. We improve algorithm based on the shape knowledge of vertebra and the problems of traditional GVF snake algorithm applied on spinal MRI. Owning to the contour of vertebra body being similar to a rectangle whose edges are concave to the center of the contour, we add a pressure force which moves the contour to the center as one of the external forces of snake, and then reset the weight coefficients of the external forces. The set of the weight coefficients can reach a target that the convergence rate is higher when the distance from the contour point to the center point is smaller and it also reduces the overall convergence speed. Finally, we program to realize the interactive segmentation system. In addition, in order to improve the accuracy and stability of the segmentation father more, we first accurately find out the four vertices of the vertebra by hand, and then fix the four vertices to guide the whole deformation.In this paper we product an algorithm of image segmentation which is appropriate for spine MRI by improving traditional GVF snake based on the shape knowledge of vertebra. Finally, we program to realize the interactive segmentation system. Lots of experimental results show that our improved algorithm makes the segmentation result smoother, more accurate and stable.