血清肿瘤标志物人工神经网络模型在胃癌诊断中的临床应用

The clinical application of an artificial neural network (ANN) model based on tumor markers in serum for diagnosing gastric carcinoma

目的:建立多种血清肿瘤标志物人工神经网络(ANN)模型并探讨其在胃癌诊断中的应用价值。方法:用酶联免疫吸附法分别测定69例良性胃病患者和72例胃癌患者血清标本中癌胚抗原(CEA)、糖类抗原(CA)199、CA242、CA50和CA724的含量,并用ANN建立5种血清肿瘤标记物诊断模型。结果:CEA、CA199、CA242、CA50、CA724对胃癌患者诊断的敏感性依次为58.3%、72.2%、77.8%、33.3%、30.6%,对胃癌患者诊断的特异性依次为89.9%、85.5%、84.1%、89.9%、91.3%,准确性分别为73.8%、78.7%、80.9%、61.0%、60.3%。ANN诊断模型的敏感性、特异性及准确性分别为88.9%、94.2%、91.5%。结论:与单一肿瘤标记物比较,血清肿瘤标志物ANN模型能显著提高诊断胃癌的敏感性、特异性与准确性,对胃癌的诊断具有较高的价值。

To establish an artificial neural network （ANN） model based on tumor markers in serum and to investigate its value in diagnosing gastric carcinoma. Methods The levels of CEA, CA199, CA242, CA50, and CA724 in the serum samples from 69 patients with benign gastric diseases and 72 with gastric carcinoma were assayed by ELISA, A diagnostic model based on the above 5 tumor markers was constructed with ANN. Results The sensitivity of CEA for gastric carcinoma was 58,3%, CA199 72.2%, CA242 77.8%, CA50 33.3%, and CA724 30.6%, the specificity was 89,9%, 85.5%, 84,1%, 89.9%, and 91.3%, and the accuracy was 73.8%, 78.7%, 80.9%, 61.0%, and 60.3%, correspondingly, The sensitivity of the ANN model was 88,9%, specificity 94.2%, and accuracy 91.5%. Conclusion As compared to a single tumor marker, the ANN model based on tumor markers in serum can markedly increase the sensitivity, specificity, and accuracy for detecting gastric carcinoma. It has a Tumor value in clinical practice.