基于改进凸包算法的肺实质分割研究

肺实质的精确分割一直都是肺部疾病计算机辅助诊断的重要研究内容，传统的分割方法大多只能分割出不包含病灶的肺实质区域，为后期的图像分析与辅助决策带来很大的影响。针对具有边缘型肺结节的肺部CT图像，提出一种实现简单且实验效果较好的肺实质分割算法。首先，利用常规方法提取肺实质的粗略轮廓；然后，针对上一步骤中肺实质病灶信息等的缺失现象，提出一种改进的二维凸包算法对肺实质的外轮廓进行再修复；最后，利用区域生长和形态学运算，修复肺实质的内部轮廓。运用新算法，对200张边缘型肺结节的肺部CT图像进肺实质分割。实验结果表明：与已有的“滚球法”和凸包算法修复肺实质相比，新算法具有较高的准确率，可以达到90%以上,边缘型肺结节等病灶信息能被较为准确地表示出来，为建立高效的肺部疾病诊断系统奠定基础。

Segmentation Research of Pulmonary Parenchyma Based on Improved Convex Hull Algorithm

Accurate segmentation of pulmonary parenchyma has been one important research content of the computer aided diagnosis of lung disease. Pulmonary parenchyma area with lesions cannot be divided by most of the traditional method of segmentation, and a great impact is brought for the image analysis and computer aided decision.Thus, a lung parenchyma segmentation algorithm was proposed for lung CT image with edge type pulmonary nodules. The algorithm is easy to implement and has a better experimental results. Firstly, we used conventional method to extract the rough contour of pulmonary parenchyma. Secondly, in connection with the absence of lung parenchyma lesions in the previous step, an improved two dimensional convex hull algorithm was proposed to repair the pulmonary parenchyma contour. Finally, the pulmonary parenchyma internal contour was acquired by using regional growth and morphology comprehensively. The test results of the experiment on 200 clinical chest CT images showed that: compared with the existing ball pivoting algorithm and convex hull algorithm to repair the lung parenchyma, the algorithm proposed in this paper has higher accuracy. The accuracy rate can reach 90% or more. Lesions like borderline pulmonary nodules can be represented exactly and it is the basics of establishing the efficient pulmonary disease diagnosis system.