面向CT影像的肺结节计算机辅助诊断算法

Computer-aided Detection Algorithm for Pulmonary Nodule Based on CT Images

为了在早期能够发现肺癌,降低对肺结节的漏诊率,提高病人的生存率;基于模糊C均值聚类的算法,利用直方图统计特性对数据进行优化,在此基础上利用像素的邻域特性,将数据样本对各聚类中心约束条件为1改变为隶属度之和为样本总数;用改进的FCM对肺实质图像进行分割,将分割后的图像应用区域分割算法去除小面积区域,利用肺结节的关键特征,提取可疑区域;运用改进算法后,区域分割效果更好;仿真结果证明算法很好地将"线"形或分枝状结构的血管去除;改进的FCM有很好的实时性和对噪声的鲁棒性,分离血管后,将可疑区域在原图标记出来,使医生的工作更加明确。

In order to diagnose lung cancer in the early time, to reduce the rate of missed diagnosis of lung nodules, improve the survival rate of patients. The article put forward a new method for segmenting the region of interest with the improved fuzzy C means cluster algorithm. The data was optimized by making use of the statistical properties of histogram. With using of pixel~ s neighborhood feature, it change the conditions of sample data for each sample clustering center constraints to 1 for membership summation equal total samples. That using regional mark algorithm remove the smallest area. with the characteristics of pulmonary nodule , extract the suspicious areas. It can achieve the better segmentation after using the improved algorithm. It is proved that algorithm has the better capability in removing the line shape and branching structure of vascular.. It is proved that the improved algorithm perform more robust to noise and the characteristics real --time. The suspicious region was marked in the original image after the seoaration of vascular. It offer doctors convenience in the work.