使用机器学习方法分析增强CT检查后对比剂肾病的风险因素  

Using machine learning methods to analyze the risk factors of contrast-induced nephropathy after contrast-enhanced CT

作　　者：赵凯[1] 张晓东[1] 吴静云[1] 张保翠[1] 罗健[1] 王霄英[1] ZHAO Kai;ZHANG Xiaodong;WU Jingyun;ZHANG Baocui;LUO Jian;WANG Xiaoying(Department of Medical Imaging,Peking University First Hospital,Beijing 100034,China)

机构地区：[1]北京大学第一医院医学影像科,北京100034

出　　处：《实用放射学杂志》2022年第8期1359-1361,1382,共4页Journal of Practical Radiology

摘　　要：目的使用机器学习方法分析增强CT检查后对比剂肾病(CIN)的风险因素.方法回顾性分析2373例患者的临床资料,包括基本信息、基础病史、对比剂注射信息共18项指标.以患者有无CIN为金标准,通过机器学习方法分析各指标的重要性,并使用传统分析方法进行补充和验证.结果机器学习结果显示CIN风险因素的重要性排序依次是估算肾小球滤过率(eGFR)、血清肌酐(SCR)、糖尿病病史、性别、恶性肿瘤病史;传统统计方法显示CIN阴性和阳性组间,eGFR、SCR和性别3项指标存在显著性差异.结论机器学习可以作为深度挖掘CIN影响因素的工具;除了eGFR、SCR,糖尿病、性别及恶性肿瘤都是CIN的风险因素,糖尿病患者、女性患者、恶性肿瘤患者应给予重点关注.Objective To investigate the risk factors of contrast-induced nephropathy(CIN)after contrast-enhanced CT examination by machine learning.Methods The clinical data of 2373 patients were analyzed retrospectively,including 18 indicators like basic information,basic medical history and contrast agent injection information etc.Taking CIN results as the gold standard,the importance of each indicator was analyzed by machine learning method,and the traditional analysis method was used for complement and validation.Results The CIN risk factors showed by machine learning were estimated glomerular filtration rate(eGFR),serum creatinine(SCR),history of diabetes,gender and history of malignant tumor in the order of importance,While eGFR,SCR and gender showed significant difference in traditional statistical methods between CIN negative and positive groups.Conclusion Machine learning can be used as a tool for in-depth exploration of the risk factors of CIN;in addition to eGFR,SCR,diabetes,gender,and malignant tumors are all risk factors.Diabetic patients,female patients,and patients with malignant tumors should be paid special attention.

关 键 词：机器学习 对比剂肾病 血清肌酐 计算机体层成像 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] R692[自动化与计算机技术—控制科学与工程] R814.42[医药卫生—泌尿科学]