中文摘要：

目的利用计算机后处理程序对孤立性困难肺结节（与胸膜、血管相连的结节及磨玻璃结节）进行分割，获得结节的形态特征，辅助医生诊断，评判计算机辅助分割在困难肺结节中的临床应用价值。方法回顾性分析经病理证实的52例困难肺结节的影像资料，其中14例与胸膜相连、18例与血管相连，20例磨玻璃结节。利用数学形态学及随机游走方法对结节进行分割，提取出结节边缘信息，并与手术病理结果进行对照，上述两者进行统计学分析。结果52例困难肺结节经上述方法去除了胸膜及血管的影响，并对磨玻璃肺结节进行交互式分割，提取出结节的精确数据；其中44例与手术病理结果进行对照无明显统计学差异。结论经病理学证实，采用数学形态学及随机游走运算对困难肺结节进行处理，实现了结节的精确分割，显示结节的边缘形态及体积大小，有利于结节良、恶性的鉴别。

英文摘要：

Objective To study the computer processing program to segment the solitary pulmonary nodules （connected with the pleura, blood vessels and ground glass nodules） ,acquire the morphological characteristics of the nodules and assist the diagnosis. To evaluate the clinical value of computer aided segmentation in difficult pulmonary nodules. Methods The imaging data of 52 cases of difficult pulmonary nodules confirmed by pathology, including 14 cases connected with the pleura, 18 cases connected with blood vessels and 20 cases with ground glass nodules were analyzed retrospectively. Mathematical morphology and random walk method to segment the nodules were used, and the edge information was extracted, which were compared with pathology results and analyzed statistically. Results Accurate data of 52 cases of difficult pulmonary nodules were extracted. By the methods of removing the effects of pleura and vessels, and interactive segmention of the ground-glass nodules, there were no significant statistical differences with pathological results in 44 cases. Conclusion Using mathematical morphology and random walk algorithm process can achieve the accurate segmentations of difficult pulmonary nodules,show the nodules＇ edge shapes and sizes. It is helpful to the differential diagnosis between benign and malignant nodules.