基于肿瘤标志群的支持向量机模型在肺癌辅助诊断中的价值

Value of support vector machine model based on tumor markers in aided diagnosis of lung cancer

目的:探讨支持向量机模型结合肿瘤标志物联合检测在肺癌辅助诊断中的价值。方法:选择肺癌患者180例、肺良性疾病患者243例,采用放射免疫法检测血清中癌胚抗原(CEA)、神经元特异性烯醇化酶(NSE)和胃泌素的含量;采用改良的间苯二酚显色法测定血清中唾液酸含量;采用电感耦合等离子体质谱法测定血清中铜、锌的含量;采用邻甲酚酞络合酮比色法测定血清钙含量;采用酶联免疫吸附法测定DNA甲基转移酶(DNMT) 1、DNMT3A、DNMT3B和组蛋白去乙酰化酶1(HDAC1)含量。从全部样本中随机抽取75%作为训练集(肺癌135例,肺良性疾病182例),以全部样本作为测试集,基于上述肿瘤标志物运用支持向量机建模。结果:肺癌组血清CEA、NSE、DNMT3A和DNMT3B含量均高于肺良性疾病组(P <0. 05)。联合10项血清指标建立的支持向量机模型的准确度为90. 3%,AUC为0. 901(0. 758～0. 836),而Fisher判别模型则为79. 7%和0. 799(0. 868～0. 927)。结论:基于多项肿瘤标志物建立的支持向量机模型对肺癌预测效果较佳,应用到临床诊断更有价值。

Aim:To explore the value of support vector machine model based on tumor markers in the aided diagnosis of lung cancer.Methods:A total of 423 patients were collected,including 180 lung cancer patients and 243 patients with benign lung diseases.The levels of carcinoembryonic antigen(CEA),neuron-specific enolase(NSE)and gastrin in serum were detected by radioimmunoassay,the concentration of sialic acid was detected by the improved resorcinol chromogenic,the concentrations of serum copper and zinc were detected by inductively coupled plasma mass spectrometry,that of serum calcium,by O-cresol oxime complex ketone colorimetry,and the contents of DNA methyltransferase(DNMT)1,DNMT3A,DNMT3B and histone deacetylation enzyme 1(HDAC1)were determined by enzyme-linked immunosorbent assay.Among all the samples,75%were randomly selected as training set(135 cases of lung cancer,182 cases of benign lung diseases),and all samples were used as the test set.Based on the above tumor markers,support vector machine model was developed.Results:The levels of CEA,NSE,DNMT3A and DNMT3B in the lung cancer group were higher than those of the lung benign diseases group(P<0.05).The accuracy of the support vector machine model based on the above tumor markers was 90.3%,and the AUC was 0.901(0.758-0.836).The accuracy of the Fisher discriminant analysis model was 79.7%,and the AUC was 0.799(0.868-0.927).Conclusion:The support vector machine model based on multi-tumor markers has good predictive effect on lung cancer and is more valuable for clinical diagnosis.