基于脉搏波的接受血液透析的终末期肾病人群冠状动脉钙化评估模型的研究  

Evaluation model of coronary artery calcification in patients with end stage renal disease undergoing hemodialysis based on pulse wave

作　　者：王充 王燕[1] 杨洁 杨琳[4] 高艳均 李颖 李紫薇 张东亮 芦铭 韩乾 孙彤彤 陈子烨 贾小月 王艳鑫 温圆圆 陈芳 WANG Chong;WANG Yan;YANG Jie;YANG Lin;GAO Yanjun;LI Ying;LI Ziwei;ZHANG Dongliang;LU Ming;HAN Qian;SUN Tongtong;CHEN Ziye;JIA Xiaoyue;WANG Yanxin;WEN Yuanyuan;CHEN Fang(School of Biomedical Engineering,Capital Medical University,Beijing 100069;Department of Medical Engineering,Beijing Jishuitan Hospital,Capital Medical University,Beijing 100096;Department of Nephrology,Beijing Jishuitan Hospital,Capital Medical University,Beijing 100096;College of Life Science and Chemistry,The Faculty of Envionment and Life Science,Beijing University of Technology,Beijing 100124;Department of Radiology,Beijing Jishuitan Hospital,Capital Medical University,Beijing 100096)

机构地区：[1]首都医科大学生物医学工程工程学院,北京100069 [2]首都医科大学附属北京积水潭医院医学工程部,北京100096 [3]首都医科大学附属北京积水潭医院肾内科,北京100096 [4]北京工业大学化学与生命科学学院,北京100124 [5]首都医科大学附属北京积水潭医院放射科,北京100096

出　　处：《北京生物医学工程》2025年第2期134-141,185,共9页Beijing Biomedical Engineering

摘　　要：目的冠状动脉钙化(coronary artery calcification,CAC)是终末期肾病(end stage renal disease,ESRD)人群的常见疾病。通过采集血液透析人群的脉搏波波形,提取脉搏波相关参数,建立基于脉搏波的血液透析人群CAC的评估模型。方法本研究的脉搏波来源于首都医科大学附属北京积水潭医院血液净化中心的血液透析人群。以胸部低剂量断层扫描作为评估受试人群CAC的影像学标准。具体根据Agaston评分规则,将受试人群按照CAC评分的不同评级划分为4类,分别为:0分、1~100分、101~400分和>400分,分别对应无钙化、轻度钙化、中度钙化和重度钙化。同时,通过无创的方法采集受试者非瘘侧桡动脉脉搏波,采集时间为:透析前、透析后1 h、2 h、3 h和透析结束后0.5 h。提取不同时段脉搏波的特征参数,并利用单因素二元Logistic回归分析筛选出风险特征参数,建立基于随机森林(random forest,RF)的接受血液透析的终末期肾病人群CAC评估模型。结果RF的机器学习模型具有较佳的表现效果(宏准确率为0.88,宏精确率为0.76,宏召回率为0.76,宏F1分数为0.75,宏AUC为0.95)。结论本研究通过机器学习的方法,完成了基于脉搏波的血液透析人群CAC的评估模型的建立与验证。运用该模型对接受血液透析的ESRD人群CAC进行评估具有一定的可行性,对临床治疗方案的选择和调整有着重要的意义。Objective Coronary artery calcification(CAC)is a prevalent condition among patients with end stage renal disease(ESRD).This study aims to establish an assessment model about CAC based on pulse wave for hemodialysis patients by collecting pulse wave signals and extracting relevant parameters.Methods Pulse wave in this study was collected in the Hemodialysis Center,Beijing Jishuitan Hospital,Capital Medical University.Low dose computed tomography of chest was used as the imaging standard to evaluate coronary artery calcification in the study population,and according to the Agaston score rules,the study population was divided into four groups:0 points,1-100 points,101-400 points and>400 points,corresponding to no calcification,mild calcification,moderate calcification,and severe calcification.At the same time,the non-fistula radial pulse wave of the study was collected.The collection time of pulse wave was selected as:before dialysis,1 hour after dialysis,2 hours after dialysis,3 hours after dialysis and 0.5 hours after ending dialysis.The characteristic parameters of pulse wave at different time periods were extracted,and the risk characteristic parameters were screened out by single factor binary logistic regression analysis,and the evaluation model of CAC for patients with ESRD who received hemodialysis was established based on random forest.Results The machine learning model based on random forest had good performance(Macro Accuracy=0.88,Macro Precision=0.76,Macro Recall=0.76,Macro F1 score=0.75 and Macro AUC=0.95).Conclusions The evaluation model of CAC in hemodialysis population based on pulse wave is established and verified.And it is feasible for predicting CAC for patients with ESRD who receives hemodialysis,and it has great significance for clinical to select and adjust treatment.

关 键 词：冠状动脉钙化 终末期肾病 血液透析 脉搏波 随机森林模型 

分 类 号：R318.04[医药卫生—生物医学工程]