基于临床检验指标建立肺腺癌患者浸润程度判别模型  

Establishment of discriminative models for predicting the infiltration degree of patients with lung adenocarcinoma based on clinical laboratory indicators

作　　者：王梦菲 杨守志 乔永霞[1] 黄琳 WANG Mengfei;YANG Shouzhi;QIAO Yongxia;HUANG Lin(School of Public Health,Shanghai Jiao Tong University,Shanghai 200025,China;Department of Clinical Laboratory Medicine,Shanghai Chest Hospital,Shanghai Jiao Tong University School of Medicine,Shanghai 200030,China;Shanghai Chest Cancer Research Institute,Shanghai Chest Hospital,Shanghai Jiao Tong University School of Medicine,Shanghai 200030,China;School of Biomedical Engineering,Shanghai Jiao Tong University,Shanghai 200030,China)

机构地区：[1]上海交通大学公共卫生学院,上海200025 [2]上海交通大学医学院附属胸科医院检验科,上海200030 [3]上海市胸部肿瘤研究所,上海交通大学医学院附属胸科医院,上海200030 [4]上海交通大学生物医学工程学院,上海200030

出　　处：《上海交通大学学报（医学版）》2024年第1期98-107,共10页Journal of Shanghai Jiao tong University:Medical Science

基　　金：国家自然科学基金(82001985);上海交通大学医学院“双百人”项目(20221714)。

摘　　要：目的·利用肿瘤标志物、凝血功能指标、血常规指标与生化指标等临床易得检验指标,建立非小细胞肺腺癌患者浸润程度的多因素判别模型。方法·回顾性选取2022年上海交通大学医学院附属胸科医院收治的肺腺癌患者,通过多因素Logistic回归分析筛选预测肺腺癌患者浸润程度的独立影响因素并建立回归模型,同时引入人工智能算法构建判别模型,采用受试者工作特征曲线下面积(area under curve,AUC)评价模型对肺腺癌患者浸润程度的判别能力。结果·共纳入肺腺癌患者202例,分为浸润前病变组(59例)以及浸润性病变组(143例)。多因素Logistic回归分析结果显示,尿素、嗜碱性粒细胞百分比、白蛋白浓度是预测肺腺癌患者浸润程度的独立影响因素(均P<0.05)。预测模型表达式为P=e^(X)/(1+e^(X)),其中X=(0.534×尿素浓度)+(1.527×嗜碱性粒细胞百分比)−(1.916×白蛋白浓度)+6.373。机器学习结果显示,纳入尿素、纤维蛋白原、白蛋白浓度、嗜碱性粒细胞百分比、前白蛋白、癌胚抗原(carcino embryonic antigen,CEA)6个指标时模型判别性能最佳。通过DeLong检验比较8种机器学习算法(分别基于岭回归、最小绝对收缩和选择算子、神经网络、随机森林、k近邻、支持向量机、决策树和自适应增强算法)的判别性能,选择AUC最高的岭回归算法,预测模型AUC为0.744(95%CI 0.656~0.832),敏感度为70.8%,特异度为70.2%。结论·使用尿素、纤维蛋白原、白蛋白浓度、嗜碱性粒细胞百分比、前白蛋白、CEA这6个指标建立综合判别模型,可有效预测肺腺癌患者肿瘤浸润程度,有望为临床肺腺癌分级判别和辅助治疗提供更精确的指导。Objective·To establish a multifactorial discriminative model for predicting the degree of infiltration in patients with non-small cell lung adenocarcinoma based on clinically accessible laboratory indicators,such as tumor markers,coagulation function indicators,routine blood count indicators,and biochemical indicators.Methods·A retrospective study was conducted on 202 patients with lung adenocarcinoma admitted to Shanghai Chest Hospital in 2022.Multifactorial Logistic regression analysis was applied to screen independent factors that influenced the predictive infiltration degree of lung adenocarcinoma and to establish a regression model.In addition,machine learning was used to construct a discriminative model,and the area under the receiver operating characteristic curve(AUC)was used to evaluate the discriminative ability of the model to discriminate the degree of infiltration in lung adenocarcinoma patients.Results·A total of 202 patients with lung adenocarcinoma were included in the study,and divided into pre-invasive lesion group(n=59)and invasive lesion group(n=143).Multifactorial Logistic regression analysis revealed that urea,percentage of basophilic granulocytes,and albumin were independent factors for predicting the degree of infiltration of lung adenocarcinoma(all P<0.05).The predictive model expression was P=e^(X)/(1+e^(X)),where X=(0.534×urea)+(1.527×percentage of basophilic granulocytes)−(1.916×albumin)+6.373.Machine learning results showed that the model performed best when urea,fibrinogen,albumin,percentage of basophilic granulocytes,prealbumin and carcino embryonic antigen(CEA)were included.After comparing the performance of 8 machine learning algorithms(based on ridge regression,least absolute shrinkage and selection operator,neural network,random forest,k-nearest neighbors,support vector machine,decision tree,and adaptive boosting algorithms)using the DeLong test,the ridge regression algorithm with the highest AUC was selected.The AUC of the predictive model was calculated to be 0.744(95%

关 键 词：肺腺癌 肿瘤标志物 凝血功能指标 临床生化指标 预测判别模型 

分 类 号：R734.2[医药卫生—肿瘤] R446[医药卫生—临床医学]