基于PIVKA-Ⅱ、AFP检测和机器学习算法的肝癌诊断预测模型性能分析  

Performance analysis of predictive model for diagnosis of hepatocellular carcinoma based on PIVKA-Ⅱ,AFP detection and machine learning algorithms

作　　者：宋佳悦 袁权 车雨东 张毅敏[1] SONG Jia-yue;YUAN Quan;CHE Yu-dong;ZHANG Yi-min(Department of Clinical Laboratory,Zhejiang Cancer Hospital,Institute of Basic Medicine and Cancer Research,Chinese Academy of Sciences,Hangzhou,310022,China;不详)

机构地区：[1]浙江省肿瘤医院检验科,中国科学院基础医学与肿瘤研究所,浙江杭州310022 [2]温州医科大学检验医学院生命科学学院 [3]宁波市临床病理诊断中心

出　　处：《中西医结合肝病杂志》2024年第9期775-780,共6页Chinese Journal of Integrated Traditional and Western Medicine on Liver Diseases

基　　金：浙江省卫生厅资助项目(No.2021KY098,No.2016KYB034,No.2020KY067)。

摘　　要：目的:探讨基于PIVKA-Ⅱ和AFP检测及机器学习算法的肝癌辅助诊断预测模型建立及诊断应用价值。方法:选取2022年3月至2022年12月浙江省肿瘤医院健康体检者112例,肝良性疾病患者149例,以及初诊为肝癌的患者265例,评价受试者基线血清异常凝血酶原(PIVKA-Ⅱ)和血清甲胎蛋白(AFP)水平,结合机器学习算法构建辅助诊断预测模型,比较不同模型在肝癌中的诊断价值。结果:血清PIVKA-Ⅱ和AFP在肝癌组患者中的水平最高,与肝癌患者肿瘤的大小、数量、分化程度等临床特征相关。以年龄、性别、PIVKA-Ⅱ和AFP水平为特征,借助梯度提升机(GBM)算法构建的肝癌辅助诊断预测模型在诊断肝癌、早期肝癌、晚期肝癌和AFP阴性肝癌的性能均优于PIVKA-Ⅱ、AFP单项及ASAP模型。结论:以年龄、性别、PIVKA-Ⅱ和AFP水平为特征,借助GBM算法构建的肝癌辅助诊断预测模型提高了肝癌的诊断准确率。Objective:To investigate the establishment and diagnostic application value of the auxiliary diagnostic prediction model for hepatocellular carcinoma patients based on PIVKA-Ⅱand AFP detection and machine learning algorithms.Methods:A total of 112 cases of healthy check-ups,149 cases of patients with benign liver disease,and 265 cases of patients with a primary diagnosis of hepatocellular carcinoma admitted to Zhejiang Provincial Cancer Hospital from March 2022 to December 2022 were selected to evaluate the levels of serum vitamin K absenceⅡ(PIVKA-Ⅱ)and alpha-fetoprotein(AFP)and to combine with machine-learning algorithms to construct a prediction model for auxiliary diagnosis,and to compare the diagnostic value of different models in hepatocellular carcinoma.Results:Serum PIVKA-Ⅱand AFP had the highest levels in the hepatocellular carcinoma group of patients.They were correlated with clinical characteristics such as tumor size,tumor number,and tumor differentiation in hepatocellular carcinoma patients.The predictive model for the adjuvant diagnosis of hepatocellular carcinoma constructed with the aid of the gradient boost machine(GBM)algorithm,characterized by age,gender,PIVKA-Ⅱ,and AFP levels,outperformed the PIVKA-Ⅱ,AFP alone,and ASAP models in the diagnosis of hepatocellular carcinoma,early-stage hepatocellular carcinoma,advanced-stage hepatocellular carcinoma,and AFP negative hepatocellular carcinoma.Conclusion:The predictive model for hepatocellular carcinoma auxiliary diagnosis constructed with the help of the GBM algorithm characterized by age,gender,PIVKA-Ⅱ,and AFP levels improved the diagnostic accuracy of hepatocellular carcinoma.

关 键 词：肝癌 PIVKA-Ⅱ AFP 梯度提升机算法 诊断价值 

分 类 号：R735.7[医药卫生—肿瘤]