中国组织工程研究 ›› 2023, Vol. 27 ›› Issue (2): 171-176.doi: 10.12307/2022.945

• 组织构建与生物力学 tissue construction and biomechanics • 上一篇    下一篇

改进测地线活动轮廓模型在肾脏CT图像分割中的应用

全美霖1,刘  奇1,陈  曦2,邓小波2,何柯辰2,刘艳丽3   

  1. 四川大学,1生物医学工程学院,2电气工程学院,四川省成都市  610065;3承德医学院生物医学工程系,河北省承德市  067000
  • 收稿日期:2022-01-11 接受日期:2022-02-24 出版日期:2023-01-18 发布日期:2022-06-20
  • 通讯作者: 刘奇,教授,四川大学生物医学工程学院,四川省成都市 610065 刘艳丽,硕士,讲师,承德医学院生物医学工程系,河北省承德市 067000
  • 作者简介:全美霖,女,1997年生,四川省成都市人,汉族,硕士,主要从事医学图像处理、计算机辅助诊断研究。

Application of improved geodesic active contour model in kidney CT image segmentation

Quan Meilin1, Liu Qi1, Chen Xi2, Deng Xiaobo2, He Kechen2, Liu Yanli3   

  1. 1College of Biomedical Engineering, 2College of Electrical Engineering, Sichuan University, Chengdu 610065, Sichuan Province, China; 3Department of Biomedical Engineering, Chengde Medical College, Chengde 067000, Hebei Province, China
  • Received:2022-01-11 Accepted:2022-02-24 Online:2023-01-18 Published:2022-06-20
  • Contact: Liu Qi, Professor, College of Biomedical Engineering, Sichuan University, Chengdu 610065, Sichuan Province, China Liu Yanli, Master, Lecturer, Department of Biomedical Engineering, Chengde Medical College, Chengde 067000, Hebei Province, China
  • About author:Quan Meilin, Master, College of Biomedical Engineering, Sichuan University, Chengdu 610065, Sichuan Province, China

摘要:

文题释义:
计算机辅助诊断:是指通过影像学、医学图像处理技术以及其他可能的生理、生化手段,结合计算机的分析计算,辅助发现病灶,提高诊断的准确率。现在常说的计算机辅助诊断技术主要是指基于医学影像学的计算机辅助技术。计算机辅助诊断技术又被称为医生的“第三只眼”, 计算机辅助诊断系统的广泛应用有助于提高医生诊断的敏感性和特异性。
水平集模型:美国加州大学的 OSHER和SETHIAN 两位教授于1988年首先提出水平集的概念,他们用哈密顿-雅克比方程模拟闭合的曲线或曲面随时间改变的状态,如火苗的外部形状变化。水平集方法是起源于曲线演化模型的经典空间连续方法,该方法基本思想是一个初始演化曲线在图像内力和外部约束力的共同作用下驱使初始曲线演化,当满足一定的收敛条件时演化曲线停止在图像目标边界,从而实现目标的分割。

背景:肾脏CT图像质量较差且腹腔CT图像中肾脏与周围组织灰度相似,用传统的图像分割方法难以准确分割出肾脏。
目的:提出一种改进的测地线活动轮廓模型,辅助肾脏疾病的诊断,提高CT图像中肾脏分割的精度。
方法:在对比分析多种传统医学图像分割算法的基础上,设计了基于改进测地线活动轮廓模型的肾脏分割算法,根据先验知识勾画出感兴趣区域,在预处理阶段中获得肾脏的初始轮廓;再以水平集方法中的测地线活动轮廓模型为基础,增强肾脏区域的边界响应并采用改进边缘指示函数,使轮廓曲线的演化结果更接近真实目标边界。
结果与结论:在328张二维肾脏CT图像上的平均Dice系数为0.974 9,平均重叠度系数为0.907 1,相较于其他水平集方法有所提高。实验结果表明,改进的测地线活动轮廓模型可以提高腹腔CT图像中肾脏区域的分割精度及分割效率。

https://orcid.org/0000-0001-9460-0506 (全美霖)

中国组织工程研究杂志出版内容重点:组织构建;骨细胞;软骨细胞;细胞培养;成纤维细胞;血管内皮细胞;骨质疏松;组织工程

关键词: 医学图像处理, 计算机辅助诊断, 水平集, 图像分割, 测地线活动轮廓模型

Abstract:

BACKGROUND: Kidney CT image with poor quality shows similar gray scale to that of surrounding tissues on abdominal CT images. Therefore, it is difficult to segment the kidney accurately by traditional image segmentation method. 

OBJECTIVE: To assist the diagnosis of renal diseases and improve the accuracy of renal segmentation in CT images based on an improved geodesic active contour model. 

METHODS: Based on the comparative analysis of various traditional medical image segmentation algorithms, a kidney segmentation algorithm based on the improved geodesic active contour model was designed. The region of interest was delineated according to prior knowledge, and the initial contour of the kidney was obtained during the pretreatment stage. Based on the geodesic active contour model of the level set method, the boundary response of the kidney region was enhanced and the improved edge indicator function was used to make the contour curve evolution result closer to the real target boundary. 

RESULTS AND CONCLUSION: The mean Dice coefficient and mean overlap degree of 328 two-dimensional CT images of the kidney were 0.974 9 and 0.907 1, respectively, which were improved compared with other level set methods. Experimental results show that this model can improve the segmentation accuracy and efficiency of the kidney region in abdominal CT images.

Key words: medical image processing, computer-aided diagnosis, level set, image segmentation, geodesic active contour

中图分类号: