Three-dimensional Reconstruction Of The Spine Image Using The Level Set Segmentation Method

As the body’s axis, the spine is very important to our body not only on its complex three dimensional structure, but on its complex and adjacent structure. Medical image processing is the use of image processing method for diagnosis of one of the most important technologies, and it has become a hotspot in recent years. For study on CT image segmentation and reconstruction of the spine, its crucial value is reflected in the computer-aided identification of information and helping clinical study of neuropathology.Image segmentation is a preprocessing step and precondition in visualization. Image segmentation and registration of the spine is key step both in image analysis and processing, and it is one of the key technology in surgical navigation. There are many methods of image segmentation at present, in which active contour segmentation models become one of the more successful and commonly used methods of image segmentation in image segmentation models in recent years. Active contour segmentation models can be divided into two categories: active contour model and region-based active contour model. Boundary-based active contour model, which by using image gradient information, makes the evolution of profile stop on the target boundaries; while region-based active contour model, extracts object by using the difference between objectives and related pixel values.Content of this article is to use a novel method based on a priori shaped level set method for CT slices to guide image segmentation and use the segmentation results to do 3D reconstruction. Main content includes: firstly, get training sample, use nuclear main components analysis algorithm to drop dimension for samples of feature space, and extract main components to express prior shape to guide segmentation; secondly, do morphology processing for the mean values of level set shape samples, and for the results as the initial contour of segmentation, thus problem to determine the initial contour of curve evolution are solved; thirdly, introduce the initial contour to RSF model, and construct new total energy functional to segment each CT image which after morphology pretreatment, and then integrate all of segmentation results to do three dimensional reconstruction. Experimental results show that the proposed method has a better result and higher efficiency than traditional methods in the segmentation of multiple CT slice images, a higher degree of automation, and a good effect for reconstructed vertebrae. Based on those, the proposed method has some significance in virtual spine surgery.