肿瘤标志物联合人工神经网络对大肠癌的预警  

The prediction of tumor markers with artificial neural network in colorectal carcinoma

作　　者：江永平[1] 张勇[1] 蒋宁[2] 俞春松 李其龙[4] 杨佐南 

机构地区：[1]嘉善县第一人民医院消化科,浙江嘉兴314100 [2]嘉善县第一人民医院内镜中心,浙江嘉兴314100 [3]嘉善县第一人民医院检验科,浙江嘉兴314100 [4]嘉善县肿瘤防治所,浙江嘉兴314100

出　　处：《中国卫生检验杂志》2015年第3期371-373,377,共4页Chinese Journal of Health Laboratory Technology

基　　金：2012年浙江省医学会正大青春宝肿瘤科研专项(2012-ZYC-A67)

摘　　要：目的观察胃肠肿瘤标志物联合人工神经网络对大肠癌的预警效果。方法将4项胃肠肿瘤标志物进行联合检测,应用人工神经网络技术建立大肠癌肿瘤标志物预警模型(即大肠癌-大肠息肉-正常人的人工神经网络模型),应用该模型和4项胃肠肿瘤标志物联合检测,分别对大肠癌患者进行预测并构建ROC曲线。结果大肠癌-大肠息肉-正常人的人工神经网络模型对大肠癌预测的灵敏度为80.03%,特异度为87.01%,准确度为81.77%,优于肿瘤标志物的联合检测,两者的ROC曲线下面积相比较,差异有统计学意义(P<0.05)。结论 4项胃肠肿瘤标志物检测联合人工神经网络模型,提高了对大肠癌的预测准确性,解决了大量复杂、繁琐的数据分析工作,其操作简便,易于推广和应用。Objective To observe the value of an artificial neural network model based on tumor markers in serum for predicting colorectal carcinoma. Methods A predicting model for colorectal carcinoma based on combined detection of 4 tumor markers was constructed with artificial neural network（ a colon carcinoma- colorectal polyps- normal ANN model）. Then the patients with colorectal cancer were predicted by using the model and 4 gastrointestinal tumor markers combined detection separately,receiver operating characteristic（ ROC） curve was plotted. Results The sensitivity,specificity,accuracy of the colon carcinoma-colorectal polyps- normal ANN model for colorectal carcinoma were 80. 03%,87. 01% and 81. 77%,respectively. It was superior than that of tumor markers combined detection. There was a statistically significance on difference between the area under the ROC curve of these two compared groups（ P〈0. 05）. Conclusion The artificial neural network model based on combined detection of 4 tumor markers in serum can increase the accuracy markedly for predicting colorectal carcinoma,besides,settle the problem of volume and complex data analysis. It is simple,and easy for popularization and application.

关 键 词：大肠癌 肿瘤标志物 人工神经网络 预警 ROC曲线 

分 类 号：R735.3[医药卫生—肿瘤]