血清标记物检测结合智能算法在胃癌诊断中的应用

Application of serum markers combined detection with intelligence algorithm in diagnosis of gastric cancer

目的:基于智能算法建立胃癌的辅助诊断模型。方法:以经病理学确诊的156例胃癌患者为胃癌组,以150例消化道良性病变患者和健康体检者为对照组,检测其血清中11种标记物的含量,通过比较ROC曲线下面积筛选出9种血清标记物,分别应用BP算法和支持向量机算法建立胃癌的数学辅助诊断模型,并通过40例测试集评价其效果。结果:成功建立了2种不同的胃癌辅助诊断模型,其中BP算法建立的诊断模型诊断准确率、敏感性、特异性分别为82%、85%、80%,支持向量机诊断模型的诊断准确率、敏感性、特异性分别为90%、95%、85%。结论:支持向量机诊断模型的诊断准确率、敏感性及特异性相对较高,对胃癌的早期预测及诊断有重要的参考价值。

Aim：To establish an assistant diagnostic model of gastric cancer on the basis of intelligence algorithm and e -valuate its efficiency .Methods：A total of 156 gastric cancer patients confirmed by pathology were enrolled as case group , other 150 patients with benign digestive disease or the healthy individuals were treat as control group .The level of eleven ser-um markers were tested , respectively , and nine of them were selected according to their area under the ROC curves .Then the mathematical assistant diagnostic models which based on the BP algorithm and support vector machine were established , meanwhile, forty test sets were run to evaluate their efficiency .Results：Two kinds of different assistant diagnostic models of gastric cancer were established , among which the accuracy , sensitivity and specificity of the model based on BP algorithm were 82%, 85%, 80%, respectively , and those of the model based on support vector machine were 90%, 95%, 85%. Conclusion：The model on the basis of support vector machine has a relatively high accuracy , sensitivity and specificity , which means that has vital reference value to the early stage prediction and diagnosis of gastric cancer .