基于机器学习算法构建心脏死亡器官捐献供肝肝移植后急性肾损伤风险预测模型的应用价值  

Application value of risk prediction model for acute kidney injury after donation of cardiac death liver transplantation based on machine learning algorithm

作　　者：陈冠戎 陈劲岩 胡歆 陈荣高 黄映辰 蒋瑶 司中洲[5] 杨家印[6] 蔡金贞 庄莉 周之晟 郑树森 徐骁 Chen Guanrong;Chen Jinyan;Hu Xin;Chen Ronggao;Huang Yingchen;Jiang Yao;Si Zhongzhou;Yang Jiayin;Cai Jinzhen;Zhuang Li;Zhou Zhicheng;Zheng Shusen;Xu Xiao(The Fourth School of Clinical Medicine,Zhejiang Chinese Medical University,Hangzhou 310053,China;Zhejiang University School of Medicine,Hangzhou 310011,China;Department of Hepatobiliary Surgery,The First Affiliated Hospital,Zhejiang University School of Medicine,Hangzhou 310003,China;Basic Medical and Law College,Hangzhou Medical College,Hangzhou 310059,China;Department of Liver Transplantation,Xiangya Hospital,Central South University,Changsha 410011,China;Liver Transplantation Center,Department of Liver Surgery,West China Hospital,Sichuan University,Chengdu 610041,China;Organ Transplantation Center,Affiliated Hospital of Qingdao University,Qingdao 266000,China;Liver Transplantation Center,Shulan Hospital(Hangzhou),Hangzhou 310022,China;National Quality Control Center for Liver Transplantation,Hangzhou 310003,China)

机构地区：[1]浙江中医药大学第四临床医学院,杭州310053 [2]浙江大学医学院,杭州310011 [3]浙江大学医学院附属第一医院肝胆胰外科,杭州310003 [4]杭州医学院基础医学与法学院,杭州310059 [5]中南大学湘雅二医院肝脏移植科,长沙410011 [6]四川大学华西医院肝移植中心肝脏外科,成都610041 [7]青岛大学附属医院器官移植中心,青岛266000 [8]树兰(杭州)医院肝移植中心,杭州310022 [9]国家肝脏移植质控中心,杭州310003

出　　处：《中华消化外科杂志》2025年第2期236-248,共13页Chinese Journal of Digestive Surgery

基　　金：国家自然科学基金(92159202,82303379);浙江省自然科学基金(LQ23H160030)。

摘　　要：目的探讨基于机器学习算法构建心脏死亡器官捐献(DCD)供肝肝移植后急性肾损伤风险预测模型的应用价值。方法采用回顾性队列研究方法。收集2015年1月至2023年12月中国肝移植注册中心浙江大学医学院附属第一医院等5家医院收治的1001对行DCD供肝肝移植供者及受者的临床病理资料;供者男825例,女176例;受者男806例,女195例,年龄为52(18~75)岁。运用过采样技术增加281例受者后获得1282例受者,通过计算机产生随机数方法以7∶3比例分为训练集897例和验证集385例。基于机器学习算法,构建随机森林、极端梯度提升树、支持向量机、逻辑回归、决策树、K最近邻和梯度提升树7种肝移植后发生急性肾损伤预测模型。观察指标:(1)发生急性肾损伤和无急性肾损伤受者及供者临床病理特征比较。(2)发生急性肾损伤和无急性肾损伤受者随访及生存情况。(3)肝移植后发生急性肾损伤列线图预测模型的构建及评价。(4)肝移植后发生急性肾损伤机器学习预测模型的构建及评价。正态分布的计量资料组间比较采用独立样本t检验。偏态分布的计量资料组间比较采用Mann-Whitney U检验,多组间比较采用Kruskal-Wallis H检验。计数资料组间比较采用χ^(2)检验或校正χ^(2)检验。采用Kaplan-Meier法计算生存率和绘制生存曲线。采用Logistic回归模型进行单因素和多因素分析。绘制受试者工作特征(ROC)曲线,并计算ROC曲线下面积(AUC)及95%可信区间(CI)。采用DeLong检验、准确度、灵敏度、特异度评价模型的预测能力。绘制校准曲线,评估预测模型的预测概率与实际概率效能。采用基于机器学习算法与SHapley Additive exPlanations(SHAP)方法的可解释性分析方法,生成对模型决策的单独解释。结果(1)发生急性肾损伤和无急性肾损伤受者及供者临床病理特征比较。1001例受者中,肝移植后发生急性肾损伤360例,无急�Objective To investigate the application value of risk prediction model for acute kidney injury(AKI)after donation of cardiac death(DCD)liver transplantation based on machine learning algorithm.Methods The retrospective cohort study was conducted.The clinicopathological data of 1001 pairs of DCD liver transplant donors and recipients at five hospitals,including The First Affiliated Hospital of Zhejiang University School of Medicine et al,in the Chinese Liver Transplantation Registry from January 2015 to December 2023 were collected.Of the donors,there were 825 males and 176 females.Of the recipients,there were 806 males and 195 females,aged 52(range,18−75)years.There were 281 recipients included using oversampling technique,and all 1282 recipients were divided to the training set of 897 recipients and the validation set of 385 recipients by a ratio of 7∶3 using computer-generated random numbers.Seven prediction models,including Random Forest(RF),Extreme Gradient Boosting(XGBoost),Support Vector Machine(SVM),Logistic Regression(LR),Decision Tree(DT),K‐Nearest Neighbors(KNN),and Categorical Boosting(CatBoost),were constructed for AKI after liver transplantation based on machine learning algorithm.Observation indicators:(1)comparison of clinicopathological characteristics between recipients with and without AKI and donors;(2)follow‐up and survival of recipients with and without AKI;(3)construction and validation of nomogram prediction model of AKI after liver transplantation;(4)construction and validation of machine learning prediction model of AKI after liver transplantation.Comparison of measurement data with normal distribution between groups was conducted using the independent sample t test.Comparison of measurement data with skewed distribution between groups was conducted using the Mann-Whitney U test,and comparison among groups was conducted using the Kruskal-Wallis H test.Comparison of count data between groups was conducted using the chi‐square test or corrected chi‐square test.Kaplan‐Meier me

关 键 词：肝移植 心脏死亡器官捐献 急性肾损伤 预后模型 机器学习 随机森林 

分 类 号：R65[医药卫生—外科学]