基于血管增强分割的三维肺结节自动检测

AUTOMATIC 3D LUNG NODULE DETECTION BASED ON ENHANCED VESSEL SEGMENTATION

在肺结节检测过程中,与肺结节灰度相似的复杂肺部血管结构是干扰肺结节准确检测的一个重要因素。针对这一问题,首先利用基于Hessian矩阵特征值的Frangi多尺度滤波器将大小、形态各异的肺血管结构增强,然后采用模糊C-均值聚类方法将血管分割出来,最后通过去除肺血管,间接得到肺结节图像。实验结果表明,该方法能有效降低血管对肺结节检测的影响,提高肺结节的检测精度。

In lung nodule detection process,the complicated pulmonary vascular structure is an important factor to interfere the lung nodule detection due to its similar gray scale. In light of this problem,we first enhance the structures of pulmonary vessels in different sizes and conformations using the Frangi multiscale filter which is based on Hessian matrix eigenvalue; Then we use fuzzy C-mean clustering method to segment the vessels; Finally,we obtain the lung nodule image indirectly by removing the pulmonary vessels. Experiment results show that the method can effectively reduce the influence of pulmonary vascellum on lung nodule detection,and improve the accuracy of lung nodule detection.