血常规数据判别骨髓增生异常综合征和急性髓样白血病的应用价值  

Value of blood routine data in distinguishing myelodysplastic syndrome and acute myeloid leukemia

作　　者：代晓宇 路媛 王志恒 李明卓[1,2] 司书成 李吉庆 井明 薛付忠[1,2] DAI Xiaoyu;LU Yuan;WANG Zhiheng;LI Mingzhuo;SI Shucheng;LI Jiqing;JING Ming;XUE Fuzhong(Department of Biostatistics,School of Public Health,Shandong University,Jinan 250012,Shandong,China;Institute for Medical Detaology,Shandong University,Jinan 250002,Shandong,China)

机构地区：[1]山东大学公共卫生学院生物统计学系,山东济南250012 [2]山东大学健康医疗大数据研究院,山东济南250002

出　　处：《山东大学学报（医学版）》2020年第1期20-25,38,共7页Journal of Shandong University:Health Sciences

基　　金：国家自然科学基金(81170352)。

摘　　要：目的基于血常规数据构建骨髓增生异常综合征和急性髓样白血病的判别模型。方法数据来源于山东多中心健康医疗大数据平台,共计1 681例。随机抽取70%患者为训练集,其他30%为测试集,应用随机森林模型对骨髓增生异常综合征与急性髓样白血病判别,采用受试者工作特征曲线下面积(AUC)衡量模型的辨别能力并使用十折交叉验证法检验模型的稳定性。结果随机森林模型与支持向量机模型均具有鉴别MDS与AML的能力,但随机森林模型表现效果更好,男性判别模型的AUC为0.874(95%CI:0.815~0.932),灵敏度和特异度分别为81.1%、81.9%;女性判别模型的AUC为0.831(95%CI:0.752~0.911),灵敏度和特异度分别为77.8%、74.3%。十折交叉验证的结果显示,男性AUC为0.884(95%CI:0.854~0.913),女性AUC为0.842(95%CI:0.802~0.883)。结论构建的随机森林模型在骨髓增生异常综合征和急性髓系白血病患者中具有较好的判别能力。Objective To establish a model to identify myelodysplastic syndrome(MDS) and acute myeloid leukemia(AML) based on blood routine data. Methods The data of 1 681 patients from Shandong Multi-Center Health Medical Big Data Platform were randomly divided into the training set(70%) and testing set(30%). MDS and AML were identified with random forest model. The discriminatory ability of the model was determined with the area under the receiver operating characteristic curve(AUC) and the stability of the model was tested with ten-fold cross validation. Results Both of the random forest model and support vector machine model were able to identify MDS and AML, but the former had better performance. It showed the estimated AUC for male was 0.874(95%CI: 0.815-0.932), sensitivity was 81.1%, and specificity was 81.9%;the estimated AUC for female was 0.831(95%CI: 0.752-0.911), sensitivity was 77.8%, and specificity was 74.3%. The ten-fold cross validation showed the estimated AUC was 0.884(95%CI: 0.854-0.913) for male and 0.842(95%CI: 0.802-0.883) for female. Conclusion The discriminant model is capable of identifying MDS and AML based on blood routine data.

关 键 词：血常规 随机森林 骨髓增生异常综合征 急性髓样白血病 ROC曲线 

分 类 号：R551.3[医药卫生—血液循环系统疾病]