CBCT图像中牙齿的分割

TOOTH SEGMENTATION IN CBCT IMAGES

江犇翔

18S140349

硕士

电子与通信工程

唐晓颖

2021-05-21

2021-05-21

哈尔滨工业大学

深圳

锥形束 CT（cone beam computed tomography, CBCT）图像被广泛地应用于口腔医疗服务行业，口腔医师通常会对 CBCT 图像进行各种手动观察和测量，对患者牙齿的形状、位置、主轴方向等领域知识有了明确的认知后，进而规划诊断和治疗的方案。通过对 CBCT 图像中的牙齿进行分割，口腔医师可以对患者的牙齿进行直观的观测，也能够在计算机上对牙齿进行高精度的移动和旋转，模拟口腔疾病的治疗过程。因此，CBCT 图像中的牙齿分割在临床上具有重要的意义，但是目前学界和业界对于牙齿分割算法的研究尚未成熟。在本篇论文中，我们将基于对牙齿分割算法的研究，进行医学图像分割理论和算法的研究。目前，我们取得了如下研究成果，包括使用基于区域生长的逐层分割算法对多齿根牙齿进行分割，以及基于水平集模型的逐层分割及三维精炼算法对各种类牙齿进行分割。在使用区域生长方法实现牙齿单层分割的研究过程中，我们充分地利用了区域生长方法的逐像素分割特点，对多齿根牙齿复杂的拓扑学结构进行高精度的分割。在使用水平集模型进行牙齿单层分割的过程中，我们通过两次水平集函数的迭代，实现了对牙齿与牙齿间消失边缘的分割以及牙齿与牙槽骨间模糊边缘的分割。我们的牙齿分割算法能够实现对各种类牙齿的高精度分割，实现了我们的研究目标。并且，在解决对牙齿间消失边缘进行分割的过程中，我们定义的水平集模型包含两次迭代，对医学图像分割领域的相关研究具有一定的参考价值。

Cone beam computed tomography (CBCT) image is widely used in the oral diagnosis and treatment. Stomatologists usually conduct various manual observations and measurements on CBCT images to achieve a clear understanding of the patient’s tooth shape, position, and orientation, and then plan diagnosis and treatment plan of patient’s oral diseases. By segmenting the teeth in the CBCT image, the stomatologists can visually observe the patient's teeth in the virtual three-dimensional space and can also move and rotate the tooth easily to simulate the treatment process of oral diseases. Therefore, tooth segmentation in CBCT images has important clinical significance. However, the current academic and industry research on tooth segmentation algorithms is not mature enough, and there has not been a published paper or the software that can achieve the accurate segmentation of any type of tooth. In this paper, we want to achieve a class of tooth segmentation algorithms with high robustness, high accuracy, and high efficiency after proceeding research on the theories and algorithms of medical image segmentation. So far, we have achieved some satisfactory research results, including the segmentation of multi-root teeth using the region growing method, and the segmentation of any kind of teeth with our defined level set model. In the process of region growing method, we make full use of the advantage of the per-pixel segmentation properties of the region growing method, which is easy to segment complex shapes, and results high-precision segmentation of tooth with multiple roots. In our defined level set model for tooth segmentation, it overcomes the two major difficulties of tooth segmentation (the segmentation of the disappearing edge between the neighboring teeth, the segmentation of the blurred edge between the tooth and its adjacent alveolar bone), and achieves excellent segmentation efficiency of any type of tooth, which meets our expectations of defining a class of tooth segmentation algorithms to satisfy the clinical needs (high robustness, high precision, and high efficiency). In addition, our level set model has a certain theoretical innovation and can be used to achieve high_x0002_precision segmentation of blurred or even disappeared edges, so our results have certain enlightening significance and research value for related research in the field of medical image segmentation.

CBCT 医学图像分割 水平集方法 口腔正畸

computed tomography images medical image segmentation level set method orthodontics

中文

联合培养

哈尔滨工业大学

江犇翔. CBCT图像中牙齿的分割[D]. 深圳. 哈尔滨工业大学,2021.