基于分层模版种子点的分水岭分割孤立性肺结节的方法

A detection method for solitary pulmonary nodules in PET-CT

针对肺部结节的分割问题,该文提出了一种基于分层模版种子点的分水岭分割方法。该方法在PET图像中采用基于SUV均值的分层次模版匹配算法检测出可疑区域,标记出分割种子点,同时在对应CT图像中使用改进的分水岭算法将可疑肺结节分割出来。将该方法与特征提取结合应用于肺结节的辅助诊断中。大量的实验结果表明:与当前单独采用CT或PET图像特征分割结果相比,该方法在确保真阳性以及分类准确性的基础上,极大降低了假阳性,从而表明了该方法在肺结节临床分割方面的有效性。

Combined features of solitary pulmonary nodules（SPNs）in both PET and CT images were considered to develop a method for solving SPNs detection problems.A hierarchical template matching algorithm and an improved watershed algorithm were used to detect the suspicious area in PET and CT images respectively. The developed method was applied to SPNs diagnoses to show that the method outperforms these methods based on CT features only or PET features only in terms of the false-positive rate reduction while guaranteeing the true-positive rate.Therefore,the results show the validity of the method in clinical SPNs segmentation.