论文部分内容阅读
应用计算机辅助图像分析方法研究了230例乳腺浸润性导管癌(invasive ductal carcinoma of the breast,IDC)的1 150张苏木素-伊红组织病理图像。用支持向量机模型分割上皮-间质,用标记点控制的分水岭算法分割细胞核,提取出730个形态学特征。Kaplan-Meier生存分析显示12个形态学特征与8年无病生存相关(P<0.05);Cox比例风险回归模型显示其中4个参数为独立预后因子:癌巢细胞密度(HR 1.645,95%CI[1.193,2.270],P=0.002)、间质细胞结构特征(HR 1.507,95%CI[1.084,2.095],P=0.015)、癌巢特征(HR 1.361,95%CI[1.026,1.804],P=0.032)及癌巢细胞核特征(HR 0.731,95%CI[0.538,0.993],P=0.045),可作为预测IDC预后的病理新指标。
A total of 1 150 hematoxylin and eosin histopathological images of 230 cases of invasive ductal carcinoma of the breast (IDC) were studied using computer-assisted image analysis. The SVM model was used to segment the epithelial-mesenchyme, and the nucleus was divided by the watershed algorithm controlled by the mark points to extract 730 morphological features. Kaplan-Meier survival analysis showed that 12 morphological features were associated with 8-year disease-free survival (P <0.05). Cox proportional hazards regression model showed that 4 of them were independent prognostic factors: cancer nest cell density (HR 1.645, 95% CI (HR 1.507, 95% CI [1.084,2.095], P = 0.015), cancer nests (HR 1.361, 95% CI [1.026,1.804], P = 0.002) P = 0.032) and nuclear characteristics of cancer cells (HR 0.731, 95% CI [0.538,0.993], P = 0.045), which could be used as a new pathological index to predict the prognosis of IDC.