口腔CT图像中独立牙齿轮廓分割算法研究

发布时间：2018-10-20 10:06

【摘要】：错颌畸形指牙齿排列不整齐、上下牙齿咬合不正确等现象,其不仅影响颌面外观而且影响颌面健康,给患者带来身心伤害。错颌畸形主要通过临床口腔正畸治疗进行矫治。基于患者口腔CT图像,可重构出牙齿三维数字模型,为正畸医生制定矫治方案和实施正畸治疗提供重要信息。而口腔CT图像中各个独立牙齿轮廓的分割是重构牙齿三维模型的关键环节。由于相邻牙齿通常粘连一起、牙根镶嵌在牙槽骨中、牙冠牙根拓扑结构不一致等问题,口腔CT图像中独立牙齿轮廓分割一直是计算机辅助口腔正畸领域中的一个技术难点,本文拟对该问题展开研究。首先在口腔CT图像中进行独立牙齿区域分离。将各层口腔CT图像沿牙弓曲线展开,投影合成全景图。检测全景图中上下颌之间、相邻牙齿之间灰度较低的像素形成牙齿上下颌分离线及相邻牙齿分离线。再根据相邻牙齿分离线确定三维空间中相邻牙齿间的分离曲面,实现独立牙齿区域的分离。将独立牙齿区域在两个相互垂直投影面上投影,通过提取投影图像中的牙齿轮廓得到二维轴线,再把两根二维轴线合成的三维轴线作为独立牙齿的长轴方向。然后根据长轴方向旋转独立牙齿区域并进行图像转换,使转换后的独立牙齿区域相邻图像间的牙齿轮廓具有相似性。采用混合水平集算法分割旋转后的独立牙齿区域中的牙齿轮廓。先从独立牙齿区域的牙颈部位选择一张切片并手工分割该切片上的牙齿轮廓,其结果作为初始牙齿轮廓。然后从初始切片开始分别沿两个相反的方向进行牙齿轮廓自动传播。轮廓传播时以上一张切片的轮廓作为先验,通过构建的混合水平集能量函数对牙齿轮廓的传播进行控制。使用5个患者的口腔CT图像对提出的分割算法进行了测试,测试结果表明,本文提出的独立牙齿轮廓分割算法可以得到良好的分割结果。
[Abstract]:Malocclusion refers to the irregular arrangement of teeth and incorrect occlusion of upper and lower teeth, which not only affects the appearance of maxillofacial, but also affects the health of maxillofacial, and brings physical and mental injury to patients. Malocclusion is mainly treated by orthodontic treatment. Based on the CT images of patients with oral cavity, the three-dimensional digital model of teeth can be reconstructed, which can provide important information for orthodontists to formulate orthodontic treatment plans and implement orthodontic treatment. The segmentation of individual tooth contours in oral CT images is the key to reconstruct the 3D model of teeth. Due to the conglutination of adjacent teeth, the root embedded in the alveolar bone and the inconsistent topologic structure of the crown and root, the contour segmentation of the independent teeth in CT images has always been a technical difficulty in the field of computer-aided orthodontics. This paper intends to study this problem. At first, CT images were used to separate the region of individual teeth. The CT images of each layer of oral cavity were expanded along the curve of dental arch, and the panoramic images were projected and synthesized. The pixels in panoramic images with lower grayscale between upper and lower mandibles and between adjacent teeth form the line of separation between upper and lower teeth and adjacent lines of teeth. Then according to the separation line of adjacent teeth, the separation surface between adjacent teeth in three dimensional space is determined to realize the separation of independent tooth regions. The independent tooth region is projected on two vertical projection planes of each other. By extracting the tooth contour from the projection image, the 2D axis is obtained, and the three-dimensional axis composed of the two 2D axes is taken as the long axis direction of the independent tooth. Then the independent tooth region is rotated according to the long axis and the image conversion is carried out to make the tooth contour of adjacent images of the transformed independent tooth region similar. A hybrid level set algorithm is used to segment the tooth contour in the rotating independent tooth region. First, a slice is selected from the neck of the independent tooth region and the tooth contour on the slice is manually segmented, and the result is taken as the initial tooth contour. Then the tooth contour propagates automatically in two opposite directions from the initial section. The profile of the section above is used as a priori to control the propagation of the tooth contour by constructing a mixed level set energy function. The proposed segmentation algorithm is tested using oral CT images of five patients. The test results show that the proposed algorithm can achieve good segmentation results.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R783.5

【参考文献】