东北大学学报:自然科学版 ›› 2017, Vol. 38 ›› Issue (5): 639-644.DOI: 10.12068/j.issn.1005-3026.2017.05.007

• 信息与控制 • 上一篇    下一篇

基于CV模型与改进ME模型的肺癌检测算法

朴春赫1, 曹鹏1,2, 赵海1, 朱宏博1   

  1. (1. 东北大学 信息科学与工程学院, 辽宁 沈阳110819; 2. 中国医科大学 公共基础学院, 辽宁 沈阳110122)
  • 收稿日期:2015-12-11 修回日期:2015-12-11 出版日期:2017-05-15 发布日期:2017-05-11
  • 通讯作者: 朴春赫
  • 作者简介:朴春赫(1976-),男, 朝鲜平壤人,东北大学博士研究生; 赵海(1959-),男,辽宁沈阳人,东北大学教授,博士生导师.
  • 基金资助:

    国家科技支撑计划项目(2012BAH82F04).

A Lung Cancer Detection Algorithm Based on CV Model and Improved ME Model

PAK Chun-hyok1, CAO Peng1, 2, ZHAO Hai1, ZHU Hong-bo1   

  1. 1. School of Information Science & Engineering, Northeastern University, Shenyang 110819, China; 2. School of Public Foundation, China Medical University, Shenyang 110122, China.
  • Received:2015-12-11 Revised:2015-12-11 Online:2017-05-15 Published:2017-05-11
  • Contact: CAO Peng
  • About author:-
  • Supported by:

    -

摘要:

针对CT影像中恶性肺结节病灶难以自动检测的问题,提出了一种基于CV模型与改进ME模型分割区域之间的面积差异的肺部CT影像癌症检测算法.该方法利用在肺部CT影像中结节边界的模糊程度是判断恶性肺结节的最重要指标这一特性,首先通过CV模型和改进ME模型两种交互式目标分割算法分别对肺部CT影像分割,因这两种分割方法收缩效果不同,故得到两种不同的结节区域,再计算这两种区域之间的面积差异得到该区域的模糊程度,最后计算得到模糊程度比较阈值,以此判断是否存在癌症.实验结果表明,该算法对于肺部CT影像中的癌症检测具有较高的准确率.

关键词: 图像分割, 水平集算法, CV模型, ME模型, 肺结节检测

Abstract:

According to solve the problem that it is difficult to automatically detect lung nodule lesions in CT images, a lung cancer detection algorithm was proposed based on Chan-Vese model (CV model) and improved mean square error model (ME model). As the degree of fuzziness of nodular boundary is the most important indicator of evaluating lung nodule in CT images study, two interactive image segmentation algorithms were employed in the proposed method based on CV model and improved ME model to process the CT image. Since the shrinkage of these two algorithms vary, two different nodular boundaries were got, and the degree of fuzziness of nodular according to the boundary difference was computed. Lastly, by comparing the degree of fuzziness of nodular, the threshold value to diagnose cancer was determined. The experimental evaluation demonstrates that compared with existing methods, the algorithm can detect lung cancer with higher accuracy in CT images.

Key words: image segmentation, level-set method, CV model, ME model, lung nodule detection

中图分类号: