| 孔聪聪,孙礼媛,梁 田,贺亚楠,曲延峻.结合CA-125, PA, 腹水，卵巢癌偏侧性建立人工神经网络模型对上皮性卵巢癌结直肠转移进行预测诊断[J].现代生物医学进展英文版,2021,(6):1093-1098. |
| 结合CA-125, PA, 腹水，卵巢癌偏侧性建立人工神经网络模型对上皮性卵巢癌结直肠转移进行预测诊断 |
| An Artificial Neural Network Model Combining CA-125, PA, Ascites and Lateralization of Ovarian Cancer was Established to Predict and Diagnose Colorectal Metastasis of Epithelial Ovarian Cancer |
| Received:August 08, 2020 Revised:September 12, 2020 |
| DOI：10.13241/j.cnki.pmb.2021.06.021 |
| 中文关键词: 上皮性卵巢癌 结直肠转移 人工神经网络 |
| 英文关键词: Epithelial ovarian cancer Colorectal metastasis Artificial neural network |
| 基金项目:黑龙江省自然科学基金项目（引导项目）（LH2019H021） |
|
| Hits: 992 |
| Download times: 663 |
| 中文摘要: |
| 摘要 目的：基于上皮性卵巢癌患者的最优临床指标建立人工神经网络模型（ANN's Model），预测卵巢癌患者结直肠转移情况，以期在降低诊断成本的同时，能够更好地进行结直肠转移风险预估，为个体化治疗方案的制定提供参考。方法：采用回顾性研究对2015年1月至2019年11月期间由哈尔滨医科大学附属第一医院、哈尔滨医科大学附属第二医院和哈尔滨医科大学附属肿瘤医院收治的801例确诊为上皮性卵巢癌患者的血清肿瘤标志物等临床常用血液指标进行分析，其中2015年1月至2017年12月所收集的附属肿瘤医院病例资料做建模组（534例）， 2018年1月至2019年11月所收集的附属一院及附属二院病例资料做外部验证组（267例）。利用Medcalc V15.2.0.0软件计算建模组534例卵巢癌患者包括血常规、尿常规、肝肾功、肿瘤标记物、凝血象等临床指标曲线下面积（area under the curve, AUC），筛选出结直肠转移相关指标；Graphpad Prism V6.0软件计算结直肠转移相关指标共线性，进一步获得结直肠转移相关优势指标来构建ANN模型，并用最优ANN模型对267例外部验证组卵巢癌患者结直肠转移进行预测，验证其能效。结果：血液学指标AUC结果显示CA-125和血清前蛋白（Prealbumin, PA）AUC值分别为0.68、0.67（P均<0.0001），为人工神经网络模型优势参数，纳入CA125、PA建立ANN I模型AUC值为0.716（P<0.0001），其对卵巢癌结直肠转移预测的敏感性及特异性分别为69.81%、62.50%。在ANNI模型基础上进而结合术中卵巢癌发生偏侧性（单侧卵巢受累或双侧卵巢受累）、有无腹水情况所建立ANN II模型AUC值为0.745（P<0.0001），其对卵巢癌结直肠转移预测的敏感度及特异性分别为65.38%、78.24%。结论：1).卵巢癌伴结直肠转移患者其血清CA-125升高更明显，PA降低更明显。伴有腹水及双侧卵巢受累的上皮性卵巢癌患者，更易出现结直肠转移。2).根据卵巢癌患者治疗前CA-125、PA所建立的人工神经网络模型ANN I及在ANNI基础上进一步结合卵巢癌偏侧性及腹水情况所建立的人工神经网络模型ANN II，其二者对上皮性卵巢癌结直肠转移有一定预测能效，且预测性能较单独的CA-125及PA更为理想，为上皮性卵巢癌结直肠转移临床预测提供参考。 |
| 英文摘要: |
| ABSTRACT Objective: This study investigated the relationship between colorectal metastasis and the clinical parameters, and then use selected parameters to build ANN（Artificial neural network）models to diagnose colorectal metastasis. At the same time of reducing the cost of diagnosis, it can better estimate the risk of colorectal metastasis, providing a reference for the development of individualized treatment plan. Methods: Receiver operating characteristic (ROC) curve analysis for all the blood texts and clinical pathological parameters of 534 ovarian cancer patients from The Tumor Hospital of Harbin Medical University in 2015.1-2017.12 was performed using Medcalc V15.2.0.0, parameters with P<0.0001 are selected, related blood parameters were used for ANN models' building after removing collinearity by using Graphpad Prism V6.0. The best ANN model was selected by AUC and then external validation for ovarian cancer colorectal metastasis was conducted using the ANN model with 267 ovarian cancer patients of The First Affiliated Hospital of Harbin Medical University, The Second Affiliated Hospital of Harbin Medical University in 2018.1-2019.11. Results: 78 blood parameters were analyzed and two blood parameters(CA-125 and PA) were selected in ANNI for the best performance of combination at last (the sensitivity, specificity, AUC were 69.81%, 62.50%, 0.716, P<0.0001, respectively). After adding ascites and ovarian focuses with these three parameters, ANN2 models had a better performance than ANN1 model with the sensitivity, specificity were 65.38%, 78.24% respectively (AUC=0.745, P<0.0001). Conclusion: 1).Ovarian cancer patients with colorectal metastasis were characterized with higher CA-125 and lower PA compared with no-colorectal metastasis group, patients with ascites and bilateral ovarian focus were more likely to have colorectal metastasis. 2). ANN1 model and ANN2 model may have a role in the prediction of epithelial ovarian cancer colorectal metastasis, and provide a reference for the clinical prediction of colorectal metastasis of epithelial ovarian cancer. |
|
View Full Text
View/Add Comment Download reader |
| Close |
|
|
|
