基于一致性法的肺部CT图像血管提取

Blood Vessel Extraction Based on a Uniformity Method in Lung CT Images

在肺部CT图像中,血管与背景的对比度较低,很难分辨肿瘤和血管结节。为了解决这个问题,本文提出具有旋转不变性的一致性法。根据肺部CT影像细小血管具有局部亮度和结构光滑的纹理特征检测细小血管。首先采用OTSU等算法快速定位CT图像的肺部区域并对定位出的结果进行去噪,应用图像二值化方法分割出粗大血管,然后对没有粗大血管肺部区域的各个子区域采用一致性法进行分类计算,最后根据细小血管的纹理特征值使用支持向量机对有细小血管的子区域及有肺结节的子区域进行样本训练,判断是否是细小血管然后将其提取。实验说明该方法是有效的。

Because of the low contrast between blood vessel and background in lung CT images,it is difficult to distinguish between lung cancer and blood vessel nodules.To solve this problem,a Uniformity Method（UM） for local brightness and structure to detect the blood vessels textures in the lung CT images is proposed.We extract the skeleton of vessel by using morphological algorithm from OTSU algorithm based binarization image,and more tiny vessel tracked by using gradient information and direction depending on smoothness of vessel.The support vector machine（SVM） has been proposed as an effective method for pattern recognition.Therefore,to evaluate the performance of the Uniformity Method（UM）,we applied the SVM to recognize tiny vessel in this system.It is proved effective by experiment.