The Research On The Key Algorithm Of Orthognathic Surgery Simulation Based On The Regional Statistical Deformable Model

As the exposed part of the body, maxillofacial area is vulnerable to be injured by the wound, tumor surgery and so on. Usually, orthognathic surgery is performed to the patients with mandible deformity, or to the person who want to perform cosmetic surgery on mandible. But how to simulate the orthognathic surgery effectively and precisely is an open issue in the field of craniomaxillofacial surgery. In this thesis, we investigate the key algorithms of orthognathic surgery simulation, and propose a novel algorithm of orthognathic soft-tissue prediction based on Regional Orthognathic Statistical Deformable Model called Rosdem, and an algorithm of automated detection for malformed craniomaxillofacial regions based on Craniomaxillofacial Statistical Deformable Model called CSDM. Using the Rosdem, we address the small-sample-size problem in statistical model and achieve better precision in the orthognathic surgery simulation. Casting abnormality detection problem to the framework of CSDM recovery, the automatic detection of malformed region and the quantification of abnormality could be implemented. The main contents of this paper are listed as follows:(1) The preprocessing of mandible regions. We perform the points registration and correspondence of the mandible regions. Firstly, the mandible regions could be got from the craniomaxillofacial using the geometric template. Then the affine transformation between the samples can be eliminated by the ICP registration. Finally, the correspondence of points are established by closest Euclidean distance algorithm.(2) The establishment of the Regional Orthognathic Statistical Deformable Model (Rosdem). The Rosdem is constructed by the pretreatment of mandible data. We study on the quantification method of the Rosdem, which includ how to obtain the normal deformable range of the statistical model, how to calculate the deformable parameter of the mandible regions, and how to recover the missing data of the sample in the model.(3) Orthognathic soft-tissue prediction based on the Rosdem. The prediction problem is cast to the framework of statistical deformable model recovery. With the strong ability of the data recovery of the Rosdem, the postoperative prediction achieve precise results. The algorithm solve the small sample size problem of traditional global statistical models, and get more precise orthognathic soft-tissue prediction.(4) Computer-aided diagnosis algorithm for automated detection and quantification of craniomaxillofacial deformed region. The regional detection problem is cast to the framework of Craniomaxillofacial Statistical Deformable Model (CSDM) recovery. The deformed regions can be detected by comparing the surface of original candidate region with the corresponding recovered region. The similarity of the two surface can be as normal, it is deformed region if the similarity is large, otherwise it is not deformed region. A fitting threshold could be used for abnormality evaluation, which guarantees both the robustness and sensitivity of the algorithm.This thesis is supported by the National Natural Science Foundation (60873095)"Orthognathic Recovery and Reconstruction Based on Statistical Deformable Model and Finite Element Models", and the National Postdoctoral Science Foundation (20070421126)"The Research on Skull Appearance Reconstruction Based on the Hierarchical Statistical Deformable Model".