论文部分内容阅读
目的 建立8种血清肿瘤标志物(TM)人工神经网络(ANN)诊断模型,以验证其在鉴别恶性肿瘤和提示原发灶不明肿瘤的原发部位中的应用价值。方法 用酶联免疫吸附法及时间分辨荧光分析法分别测定73例肺癌、60例肝癌、76例肠癌、78例胃癌和51例乳腺癌患者, 4例原发灶不明恶性肿瘤患者血清标本中,癌胚抗原(CEA)、α- 甲胎蛋白(AFP)、糖抗原19 9 (CA199)、糖抗原242(CA242)、癌抗原72 4(CA724)、细胞角蛋白19片段(CA211)、神经元特异性烯醇化酶(NSE)和组织多肽抗原(TPA)的含量,结合ANN进行数据建模分析,并用盲法验证模型,对原发灶不明肿瘤的鉴别准确率进行评价。结果 肺癌和肝癌ANN模型,对肺癌和肝癌的鉴别准确率达90 .9%;肠癌和肝癌ANN模型,对肠癌和肝癌的鉴别准确率为88 .9%;肝癌和胃癌ANN模型,对肝癌和胃癌的鉴别准确率为93. 5%;乳腺癌和肺癌ANN模型,对乳腺癌和肺癌的鉴别准确率为80. 5%;用肠癌和肝癌ANN模型鉴别2例原发灶不明肿瘤(术后病理诊断为肠癌)为肠癌;用肝癌和胃癌ANN模型鉴别2例原发灶不明肿瘤(术后病理诊断为肠癌)为胃癌。结论 ANN结合8种TM建立的诊断模型分析法可有效鉴别肺癌等5种肿瘤,并且可能提示原发灶不明的恶性肿瘤的原发部位。
OBJECTIVE: To establish a diagnostic model of eight human serum tumor markers (TM) artificial neural network (ANN) to verify its value in identifying primary malignancies and suggesting primary sites of unknown primary tumors. Methods 73 cases of lung cancer, 60 cases of liver cancer, 76 cases of colon cancer, 78 cases of gastric cancer and 51 cases of breast cancer were detected by enzyme-linked immunosorbent assay and time-resolved fluorescence analysis. Serum samples from 4 patients with primary malignant tumor , Carcinoembryonic antigen (CEA), alpha-fetoprotein (AFP), carbohydrate antigen 19 9 (CA199), carbohydrate antigen 242 (CA242), cancer antigen 72 4 (CA724), cytokeratin 19 fragment (NSE) and tissue polypeptide antigens (TPA) were determined. Data were analyzed by ANN, and the accuracy of identification of unknown primary tumor was evaluated by blind validation model. Results The accuracy of the ANN model of lung cancer and liver cancer was 90.9% for differentiating lung cancer from liver cancer. The ANN model of colorectal cancer and hepatocellular carcinoma was 88.9% for differentiating colorectal cancer and liver cancer, and ANN model for liver cancer and gastric cancer. The accuracy rate of identification of liver cancer and gastric cancer was 93.5%. The diagnostic accuracy of breast cancer and lung cancer ANN model for breast cancer and lung cancer was 80.5%. Two models of primary tumor were identified by ANN model of intestinal cancer and liver cancer (Postoperative pathological diagnosis of intestinal cancer) for the colon cancer; with liver cancer and gastric cancer ANN model identification of 2 cases of primary tumor of unknown tumor (postoperative pathological diagnosis of intestinal cancer) for gastric cancer. Conclusion ANN combined with eight kinds of TM established diagnostic model analysis can effectively identify five kinds of tumors such as lung cancer, and may prompt the primary tumor of unknown primary site.