口腔医学研究

口腔医学研究 ›› 2018, Vol. 34 ›› Issue (8): 876-880.DOI: 10.13701/j.cnki.kqyxyj.2018.08.018

• 颞下颌关节与颌骨研究 • 上一篇    下一篇

不同CBCT图像分割技术在TMJOA诊断中的应用对比研究

吕佳虹, 许跃, 张志光, 林韩*   

  1. 中山大学光华口腔医学院·附属口腔医院 广东省口腔医学重点实验室 广东 广州 510000
  • 收稿日期:2018-01-25 出版日期:2018-08-28 发布日期:2018-08-23
  • 通讯作者: 林韩,E-mail:linhan5@mail2.sysu.edu.cn
  • 作者简介:吕佳虹(1992~ ),女,湖南人,硕士,主要从事口腔正畸与颞下颌关节相关临床及基础的研究。
  • 基金资助:
    国家自然科学基金面上项目(编号:81771124 )

Comparison on Image Segmentation Methods for Patients with Temporomandibular Joint Osteoarthrosis

LV Jia-hong, XU Yue, ZHANG Zhi-guang, LIN Han*   

  1. School and Hospital of Stomatology, Sun Yat-sen University, Guangzhou 510055, China
  • Received:2018-01-25 Online:2018-08-28 Published:2018-08-23

PDF (PC)

466

摘要： 目的:对比不同关节CBCT图像分割算法的准确性和临床可用性,获得TMJ-OA关节结构理想的优化分割算法。方法:对46侧TMJ-OA髁突及46侧健康成年人髁突应用MIMICS 14.11软件重建髁突三维模型,分别采用手工分割法、阈值分割法及基于主动轮廓模型和区域生长的混合分割算法,对髁突进行图像分割。比较3种方法在髁突体积、表面积、分割时间及分割轮廓线的可调性、自适应性及软件界面友好度的差异。结果:对TMJ-OA组和正常对照组,混合分割算法重建的髁突与手工分割法均无统计学差异,阈值分割法重建髁突尺寸显著比其他两种方法小。混合分割算法较其他算法具有分割快、临床可用性高的优势。结论:基于主动轮廓模型和区域生长的混合分割算法在TMJ-OA CBCT图像分割中具有较高的准确性和临床可用性,优于目前临床常用的单一分割算法。

关键词: 图像分割, 颞下颌关节骨关节病, 无边界主动轮廓模型

Abstract: Objective: To compare different algorithms for TMJ CBCT image segmentation and explore an ideal optimization segmentation algorithm. Methods: The CBCT images of 46 TMJ-OA condyles and 46 healthy condyles were included and three-dimensional model was reconstructed. The images of condyles were segmented by the manual segmentation, threshold segmentation, and hybrid algorithm based on borderless active contour model and region growing. The methods were compared concerning time expenditure, adjustability, adaptability of segmentation contours, software interface friendliness score, condylar volume, and surface size. Results: No significant morphological differences were found between manual segmentation and hybrid algorithm segmentation of condyles with TMJ-OA. The volume of condyles with threshold segmentation was significantly smaller than that with the other methods. The value of volume discrepancy was 85.7±296.1mm. The hybrid algorithm segmentation was significantly faster with better usability compared with manual and threshold based segmentation method. Conclusion: The hybrid algorithm segmentation method based on the borderless active contour model and region growing algorithm is a reliable and intelligent approach for condyle contouring in patients with TMJ-OA, which is better than single threshold based segmentation.

Key words: Image segmentation, TMJ-OA, Borderless active contour model