betway必威登陆网址 (betway.com )学报››2023,Vol. 44››Issue (11): 830-834.DOI:10.3969/j.issn.2097-0005.2023.11.006

• 临床研究 •上一篇

不同算法在UBM小梁网区域分割中的应用

翟玉喜1(), 刘学彦2, 高建鲁1, 汪鑫1()

  1. 1.聊城市人民医院眼科,山东 聊城 252000
    2.聊城大学数学科学院,山东 聊城 252000
  • 收稿日期:2023-08-22出版日期:2023-11-25发布日期:2024-01-22
  • 通讯作者:汪鑫
  • 作者简介:翟玉喜,硕士,副主任医师,研究方向:青光眼、白内障等眼前节疾病,E-mail:zhaiyuxi1986@sina.com
  • 基金资助:
    山东省医药卫生科技发展计划(202007020086)

Different segmentation algorithms for image analysis of trabecular meshwork in UBM

Yuxi ZHAI1(), Xueyan LIU2, Jianlu GAO1, Xin WANG1()

  1. 1.Department of Ophthalmology,Liaocheng People’s Hospital,Liaocheng,252000,China
    2.Department of Mathematics,Liaocheng University,Liaocheng,252000,China
  • Received:2023-08-22Online:2023-11-25Published:2024-01-22
  • Contact:Xin WANG

摘要:

目的探讨不同算法在超声生物显微镜(ultrasonic biological microscope,UBM)图像中小梁网区域自动或半自动分割中的应用,实现小梁网-舒林氏管(trabecular meshwork-schlemm 's canal,TM-SC)边界的自动化识别,为后续的TM-SC定量、定性分析提供基础。方法采用原发性开角型青光眼(primary open angle glaucoma,POAG)患者80 MHz UBM房角图像,应用Ostu、K-means和Level set 3种分割算法和开源医学图像处理软件ImageJ对小梁网区域图像进行分割,提取感兴趣的TM-SC区域。结果Ostu、K-means算法不能准确识别TM-SC边界,Level set算法能够准确分割出TM-SC边界,ImageJ软件可手动分割小梁网区域,其效果主要取决于操作者的经验,但速度慢、重复性差。结论Level set算法可准确分割出UBM图像中TM-SC边界,TM-SC后续的几何度量对阐明青光眼发病机制具有重要的临床价值和指导意义。

关键词:图像分割,超声生物显微镜,Level set算法,Schlemm’s管,小梁网

Abstract:

ObjectiveTo explore the application of different algorithms in automatic or semi-automatic image segmentation of trabecular meshwork in ultrasonic biological microscope (UBM) images and accomplish automatic identification of trabecular meshwork-schlemm’ s canal(TM-SC),and provide the basis for the subsequent quantitative and qualitative analysis.Methods80 MHz UBM was used to collect the images of the anterior chamber angle from patients with primary open-angle glaucoma (POAG). The region of interest in the trabecular meshwork area was extracted from the UBM image. Three segmentation algorithms (Ostu, K-means and Level set) and the open-source image processing software ImageJ were used to segment the trabecular meshwork area.ResultsOstu algorithm and K-means algorithm could not identify the boundary of TM-SC. The level-set algorithm could accurately segment the boundary of TM-SC. The disadvantages of manual segmentation of UBM images by ImageJ were time-consuming, energy-consuming and poor repeatability. Besides, the segmentation effect depended on the operator’s experience.ConclusionThe level-set algorithm can accurately segment the TM-SC boundaries in the UBM images, the geometric measurement of which has important clinical reference values and guiding significance for monitoring the progression of glaucoma disease.

Key words:image segmentation,ultrasonic biological microscope,level-set algorithm,Schlemm’s canal,trabecular meshwork