Research On Spine Feature Points Automatic Detection And Matching

Long-term use of electronic products in the incorrect posture, spine disease has become a common disease in the crowd. Research in the field of spine become a new hotspot. Computer graphics technology provides support for medical diagnosis, which included surgery preview, surgical navigation and surgical evaluation. Currently spinal disease treatment still lack of automated data collection methods, resulting in the spine data collection process can only be carry out through manual or semi-automatic way.In order to achieve the purpose of automatic data acquisition and measurement of the spine, this paper studies the automatic detection of feature points of the spine. On this basis, using feature points matching to achieve the target of automatic data acquisition of the spine. Firstly, through comparative analysis of three dimensional modeling methods: surface rendering, volume rendering and modeling method based on Simpleware. By comparing the advantages and disadvantages of the three methods, the spine L3 segment experimental model constructed using the method base on Simpleware.Then, the paper summarizes the current mainstream feature point detection methods. The feature point detection method of three-dimensional model divided into three categories, namely: feature point detection method based on 3D geometry, 2D feature point detector and feature point detection method based on semantic. For each category, one or more typical representative algorithms described in detail. We analyze the principle and process of the algorithm. Feature point detection algorithms provide the theoretical basis for this study. The main work focused on the implementation of the feature point detection and the analysis of results. Utilization VTK development kits, this paper implements three categories feature point detection methods. The algorithm is applied to the spine L3 model, obtaining a plurality feature points detection results. By reference to the existing evaluation criteria, this paper analyzes the detection effect from different aspects.Finally, based on the spine feature point detection, ICP algorithm is applied to the spine model matching. The results of spine model matching depend on the size of the input data source. On the basis of the spine model matching, by defining measuring methods of sample model, the relevant data of experimental spine models is obtained automatically.It can be found that the method this paper presents can effectively detect feature points of the spine and automatically collect the spine data, which provides supports for the relevant research on spine.