论文部分内容阅读
目的:探讨血清肿瘤标志物检测结合决策树模型在肝癌诊断中的价值。方法:运用肿瘤标志物定量检测试剂盒对119例肝部良性疾病及98例肝癌患者血清中9项肿瘤标志[甲胎蛋白(AFP)、癌胚抗原(CEA)、CA125、CA242、CA199、神经元特异性烯醇化酶(NSE)、铁蛋白(Ferritin)、人生长激素(HGH)和CA153]水平进行检测,应用logistic回归筛选肿瘤标志物,并于筛选前后建立决策树模型和Fisher判别分析模型。结果:肝癌组9项血清肿瘤标志物水平均高于肝良性疾病组(P<0.05)。筛选前基于9项肿瘤标志物、筛选后基于3项肿瘤标志物分别建立Fisher判别分析模型、决策树模型,其预测准确度分别为76.5%、91.2%、74.4%、90.8%。筛选前后决策树模型ROC曲线的AUC分别为0.912和0.908,高于Fisher判别分析的0.745和0.727(Z=4.512和4.589,P均<0.05);但决策树模型和Fisher判别分析筛选前后自身相比,差异均无统计学意义(Z=1.855和1.122,P均>0.05)。结论:基于3项血清肿瘤标志物建立的决策树模型诊断肝癌的效果优于Fisher判别分析。
Objective: To investigate the diagnostic value of serum tumor markers combined with decision tree model in the diagnosis of liver cancer. Methods: Ninety - nine benign diseases of the liver and nine of 98 patients with hepatocellular carcinoma (AFP, CEA, CA125, CA242, CA199) were detected by using Tumor Markers Quantitative Test Kit. (NSE), ferritin, human growth hormone (HGH) and CA153 levels were detected by logistic regression analysis of tumor markers, and before and after screening to establish a decision tree model and Fisher discriminant analysis model . Results: The serum levels of nine serum tumor markers in HCC patients were significantly higher than those in benign liver disease patients (P <0.05). Based on the nine tumor markers before screening, the Fisher discriminant analysis model and the decision tree model were established based on the three tumor markers after screening. The prediction accuracy was 76.5%, 91.2%, 74.4% and 90.8% respectively. The AUC of the ROC curves of the decision tree model before and after screening were 0.912 and 0.908 respectively, which were higher than those of Fisher discriminant analysis of 0.745 and 0.727 (Z = 4.512 and 4.589, P <0.05 respectively). However, compared with the Fisher discriminant analysis , There was no significant difference (Z = 1.855 and 1.122, P> 0.05). Conclusion: The decision tree model based on three serum tumor markers is superior to Fisher discriminant in the diagnosis of liver cancer.