中国血液净化

中国血液净化 ›› 2022, Vol. 21 ›› Issue (02): 98-102.doi: 10.3969/j.issn.1671-4091.2022.02.007

• 临床研究 • 上一篇    下一篇

基于计算机视觉的尿液定量分析系统在白蛋白尿筛检中的应用效能分析

吴静依1,2,王飞2,李青2,沈华2,王学晶3,鲍云非1,4,李鹏飞2,高碧霞1,4,张路霞1,2,4,5   

  1. 1北京大学第一医院肾内科，北京大学肾脏病研究所
    2浙江省北大信息技术高等研究院
    3北京大学民航总医院检验科
    4中国医学科学院免疫介导肾病诊治创新单元
    5北京大学健康医疗大数据国家研究院

  • 收稿日期:2021-09-23 修回日期:2021-11-02 出版日期:2022-02-12 发布日期:2022-02-17
  • 通讯作者: 李鹏飞 pfli@aiit.org.cn； 高碧霞 Sherrygao1021@126.com E-mail:pfli@aiit.org.cn
  • 基金资助:

    国家自然科学基金(81900665，91846101)；

    “北京大学医学部-密歇根大学医学院转化医学与临床研究联合研究所”联合研究项目(BMU2018JI012，BMU2019JI005)；

    北大百度基金(2019BD017)；

    中国医学科学院医学与健康科技创新工程项目资助(2019-I2M-5-046)

The application performance of urine quantitative analysis system based on computer vision in the screening of albuminuria

  1. 1Renal Division, Department of Medicine, Peking University First Hospital, and Institute of Nephrology, Peking University, Beijing 100034, China; 2Advanced Institute of Information Technology, Peking University, Zhejiang 311215, China;  3Clinical laboratory, Civil Aviation General Hospital, Peking University, Beijing 100025, China;  4Research Units of Diagnosis and Treatment of Immune- mediated Kidney Diseases, Chinese Academy of Medical Sciences, Beijing 100034, China; 5National Institute of Health Data Science, Peking University, Beijing 100191, China
  • Received:2021-09-23 Revised:2021-11-02 Online:2022-02-12 Published:2022-02-17
  • Contact: Pengfei Li E-mail:pfli@aiit.org.cn

PDF

282

摘要： 【摘要】目的探讨基于计算机视觉算法的尿液定量分析系统在白蛋白尿筛检中的应用效果。方法选取2021 年2 月在北京大学第一医院接受尿液检验分析，不同尿微量白蛋白肌酐比值范围的143 例受试者纳入研究。采集受试者的随机尿液样本，分别送到医院临床检验实验室和使用尿液定量分析系统进行检测分析。以实验室检测结果为金标准，从筛检的真实性、可靠性、预测值和受试者工作特征曲线下面积(AUC)4 个方面评价基于计算机视觉算法的尿液定量分析系统用于白蛋白尿筛检的效果。结果共纳入143 例受试者，白蛋白尿A1 分级的59(41.3%)例，A2 分级的39(27.3%)例，A3 分级的45(31.5%)例。基于计算机视觉算法的尿液定量分析系统在白蛋白尿筛检的真实性、可靠性、预测值和AUC 方面均表现良好。以白蛋白尿A1 分级为阴性，A2 和A3 分级为阳性，尿液定量分析系统准确率达到88.8%，灵敏度和特异度分别达到94.0%和81.4%，阳性预测值和阴性预测值分别达到87.8%和90.6%，AUC 达到0.962。结论基于计算机视觉算法的尿液定量分析系统在白蛋白尿筛检中的应用效果较好，诊断准确率、AUC 和灵敏度均较高。由于其便捷性与低成本，该分析系统在我国大规模人群调查用于慢性肾脏病初筛中有较大的应用推广价值。

关键词: 尿液定量分析系统, 计算机视觉算法, 白蛋白尿, 慢性肾脏病, 筛检效果分析

Abstract: 【Abstract】Objectives This study aimed to evaluate the diagnostic performance of a computer visionbased urine quantitative analysis system for albuminuria screening. Methods A total of 143 participants with various levels of urinary albumin to creatinine ratio (uACR) recruited from the patients subjected to urinary analysis at Peking University First Hospital during February 2021 were included in this study. Randomly selected spot urine samples were collected from these participants and measured using both clinical laboratory method and the computer vision-based urine quantitative analysis system. With the results of clinical laboratory
method as golden criteria, the diagnostic performance of the computer vision-based urine quantitative analysis system in albuminuria screening was evaluated in terms of validity, reliability, predictive value, and area under the receiver operating curve (AUC). Results In the 143 participants, albuminuria A1, A2 and A3 accounted for 59 case (41.3%), 39 cases (27.3%) and 45 cases (31.5%), respectively. The computer visionbased urine quantitative analysis system achieved better performance in albuminuria screening in terms of validity, reliability, predictive value and AUC. When albuminuria A1 was set as negative albuminuria and albuminuria A2 and A3 were set as positive albuminuria, the urine quantitative analysis system achieved an accuracy of 88.8%, a sensitivity of 94.0% and a specificity of 81.4%; the positive and negative predictive values were 87.8% and 90.6%, respectively, with an AUC of 0.962. Conclusions The computer vision-based urine quantitative analysis system had better diagnostic performance in albuminuria screening with higher accuracy, sensitivity and AUC. Due to its convenience and low cost, the computer vision-based urine quantitative analysis system is especially suitable for the preliminary screening of chronic kidney disease in large scales of population.

Key words: Urine quantitative analysis system, Computer vision, Albuminuria, Chronic kidney disease, Screening effect analysis

中图分类号: