[1]王慧霞,张玉婷,朱曼辉.基于H2O平台自动化机器学习的糖尿病视网膜病变预测模型的建立[J].医学信息,2023,36(22):8-13.[doi:10.3969/j.issn.1006-1959.2023.22.002]
　WANG Hui-xia,ZHANG Yu-ting,ZHU Man-hui.Establishment of Diabetic Retinopathy Prediction Model Based on H2O Platform Automated Machine Learning[J].Journal of Medical Information,2023,36(22):8-13.[doi:10.3969/j.issn.1006-1959.2023.22.002]
点击复制

基于H2O平台自动化机器学习的糖尿病视网膜病变预测模型的建立()
分享到：

医学信息[ISSN:1006-1959/CN:61-1278/R]

卷:
36卷
期数:
2023年22期
页码:
8-13
栏目:
医学数据科学
出版日期:
2023-11-15

文章信息/Info

Title:
Establishment of Diabetic Retinopathy Prediction Model Based on H2O Platform Automated Machine Learning
文章编号:
1006-1959（2023）22-0008-06
作者:
王慧霞; 张玉婷; 朱曼辉
（苏州大学理想眼科医院斜视与小儿眼科1，病理科2，江苏 苏州 215000）
Author(s):
WANG Hui-xia; ZHANG Yu-ting; ZHU Man-hui
(Department of Strabismus and Pediatric Ophthalmology1,Department of Pathology2,Lixiang Eye Hospital of Soochow University,Suzhou 215000,Jiangsu,China)
关键词:
糖尿病视网膜病变; 自动化机器学习; 预测模型; 混淆矩阵; SHAP可视化; 部分依赖图
Keywords:
Diabetic retinopathy; Automatic machine learning; Prediction model; Confusion matrix; Shapley additive explanations; Partial dependence plots
分类号:
R587.2；R774.1
DOI:
10.3969/j.issn.1006-1959.2023.22.002
文献标志码:
A
摘要:
目的 利用H2O平台推出的自动化机器学习（AutoML）算法，建立预测糖尿病（DM）视网膜病变（DR）模型。方法 纳入2019年1月-2021年1月于本院就诊的606例DM患者，根据眼底照相分为单纯DM组（DM组，303例）及DM合并DR组（DR组，303例）。采集两组患者基本情况、血生化检测结果及视网膜图像等数据。利用H2O AutoML算法建立针对DR二分类结局，进行变量筛选并建立机器学习预测模型，产生相应预测结果，据此绘制ROC曲线并建立混淆矩阵，绘制SHAP及部分依赖图，评价模型区分能力。结果 DR组糖尿病病程长于DM组，吸烟、饮酒、高血压、脂肪肝比例、腰臀比、BMI及收缩压高于DM组，差异有统计学意义（P＜0.05）；DR组HDL-C低于DM组，FPG、FINS、HOMA-IR、HbA1c、ALT和AST均高于DM组，差异有统计学意义（P＜0.05）。将两组特征数据载入AutoML工作环境中，得到最佳模型为通用梯度回归模型（GBM），该模型Gini值0.914，R2为0.679，LogLoss为0.260。重要性排名前3的变量包括FPG、糖尿病病程及FINS。在Train数据集中，ROC曲线下面积为0.942（95%CI：0.921~0.963）。利用混淆矩阵得到特异度为0.924，敏感度为0.959，准确度为0.942，误分类率为0.058。在Valid数据集中，ROC曲线下面积为0.831（95%CI：0.764~0.897）。利用混淆矩阵得到特异度为0.828，敏感度为0.833，准确度为0.831，误分类率为0.169。结论 本次利用AutoML算法建立的通用梯度回归DR患病预测模型可用于DM人群中DR的筛查。
Abstract:
Objective To establish a model for predicting diabetic retinopathy (DR) by using the automatic machine learning (AutoML) algorithm introduced in the H2O platform.Methods A total of 606 patients with DM who were treated in our hospital from January 2019 to January 2021 were divided into simple DM group (DM group, 303 patients) and DM combined with DR group (DR group, 303 patients) according to fundus photography. The basic conditions, blood biochemical test results and retinal images of the two groups were collected. The H2O AutoML algorithm was used to establish a machine learning prediction model for DR binary classification outcomes, and the corresponding prediction results were generated. Based on this, the ROC curve was drawn and the confusion matrix was established. SHAP and partial dependence graph were drawn to evaluate the model’s ability to distinguish.Results The duration of diabetes in the DR group was longer than that in the DM group, and the proportion of smoking, drinking, hypertension, fatty liver, waist-to-hip ratio, BMI and systolic blood pressure were higher than those in the DM group (P<0.05). HDL-C in DR group was lower than that in DM group, FPG, FINS, HOMA-IR, HbA1 c, ALT and AST in DR group were higher than those in DM group, the differences were statistically significant (P<0.05). The two sets of feature data are loaded into the AutoML working environment, and the best model was the general gradient regression model (GBM), the Gini value of the model was 0.914, R2 was 0.679, and LogLoss was 0.260. The top three important variables included FPG, duration of diabetes and FINS. In the Train dataset, the area under the ROC curve was 0.942 (95%CI:0.921-0.963). Using the confusion matrix, the specificity was 0.924, the sensitivity was 0.959, the accuracy was 0.942, and the misclassification rate was 0.058. In the Valid dataset, the area under the ROC curve was 0.831(95%CI:0.764-0.897). Using the confusion matrix, the specificity was 0.828, the sensitivity was 0.833, the accuracy was 0.831, and the misclassification rate was 0.169.Conclusion The gradient boost machine model for DR diagnosis prediction based on H2O AutoML can be used for DR screening in DM population.

参考文献/References:

[1]中华医学会糖尿病学分会视网膜病变学组.糖尿病视网膜病变防治专家共识[J].中华糖尿病杂志,2018,10(4):241-247.[2]中华医学会糖尿病学分会.中国2型糖尿病防治指南(2017年版)[J].中国实用内科杂志,2018,38(4):292-344.[3]Semeraro F,Parrinello G,Cancarini A,et al.Predicting the risk of diabetic retinopathy in type 2 diabetic patients[J].J Diabetes Complications,2011,25(5):292-297.[4]Liu Y,Yang J,Tao L,et al.Risk factors of diabetic retinopathy and sight-threatening diabetic retinopathy: a cross-sectional study of 13 473 patients with type 2 diabetes mellitus in mainland China[J].BMJ Open,2017,7(9):e016280.[5]Handelman GS,Kok HK,Chandra RV,et al.eDoctor: machine learning and the future of medicine[J].J Intern Med,2018,284(6):603-619.[6]LeDell E,Poirier S.H2O AutoML:Scalable Automatic Machine Learning[C]//7th ICML Workshop on Automated Machine Learning(AutoML).2020.[7]Pan CW,Wang S,Qian DJ,et al.Prevalence, Awareness, and Risk Factors of Diabetic Retinopathy among Adults with Known Type 2 Diabetes Mellitus in an Urban Community in China[J].Ophthalmic Epidemiol,2017,24(3):188-194. [8]Saeidian J,Mahmoudi T,Riazi-Esfahani H,et al.Automated assessment of the smoothness of retinal layers in optical coherence tomography images using a machine learning algorithm[J].BMC Med Imaging,2023,23(1):21.[9]Han R,Jin M,Xu G,et al.Progressively Decreased HCN1 Channels Results in Cone Morphological Defects in Diabetic Retinopathy[J].J Neurosci,2022,42(43):8200-8212.[10]Petersen MC,Shulman GI.Mechanisms of Insulin Action and Insulin Resistance[J].Physiol Rev,2018,98(4):2133-2223.[11]Cheung N,Mitchell P,Wong TY.Diabetic retinopathy[J].Lancet,2010,376(9735):124-136.[12]Hainsworth DP,Bebu I,Aiello LP,et al.Risk Factors for Retinopathy in Type 1 Diabetes:The DCCT/EDIC Study[J].Diabetes Care,2019,42(5):875-882.[13]Hayes AJ,Leal J,Gray AM,et al.UKPDS outcomes model 2: a new version of a model to simulate lifetime health outcomes of patients with type 2 diabetes mellitus using data from the 30 year United Kingdom Prospective Diabetes Study:UKPDS 82[J].Diabetologia,2013,56(9):1925-1933. [14]Vujosevic S,Aldington SJ,Silva P,et al.Screening for diabetic retinopathy: new perspectives and challenges[J].Lancet Diabetes Endocrinol,2020,8(4):337-347. [15]Mo R,Shi R,Hu Y,et al.Nomogram-Based Prediction of the Risk of Diabetic Retinopathy:A Retrospective Study[J].J Diabetes Res,2020,2020:7261047.[16]明帅,姚溪,谢坤鹏,等.糖尿病视网膜病变人工智能自动诊断系统在社区和医院老年糖尿病患者中的应用效果分析[J].中华眼底病杂志,2022,38(2):120-125.[17]Ji Y,Chen N,Liu S,et al.Research Progress of Artificial Intelligence Image Analysis in Systemic Disease-Related Ophthalmopathy[J].Dis Markers,2022,2022:3406890.[18]Dong L,He W,Zhang R,et al.Artificial Intelligence for Screening of Multiple Retinal and Optic Nerve Disease[J].JAMA Netw Open,2022,5(5):e229960.[19]Ma J,Jiang S,Liu Z,et al.Machine Learning Models for Slope Stability Classification of Circular Mode Failure: An Updated Database and Automated Machine Learning (AutoML) Approach[J].Sensors (Basel),2022,22(23):9166.[20]Wang Y,Hong Y,Wang Y,et al.Automated Multimodal Machine Learning for Esophageal Variceal Bleeding Prediction Based on Endoscopy and Structured Data[J].J Digit Imaging,2023,36(1):326-338.

相似文献/References:

[1]郭 明,郑继伟,马丽平.25(OH)D3与糖尿病视网膜病变的关系研究[J].医学信息,2018,31(08):99.[doi:10.3969/j.issn.1006-1959.2018.08.031]
　GUO Ming,ZHENG Ji-wei,MA Li-ping.Study on the Relationship between 25(OH)D3 and Diabetic Retinopathy[J].Journal of Medical Information,2018,31(22):99.[doi:10.3969/j.issn.1006-1959.2018.08.031]
[2]闫宇梅.血清胱抑素C及同型半胱氨酸与2型糖尿病 视网膜病变的相关性分析[J].医学信息,2018,31(17):89.[doi:10.3969/j.issn.1006-1959.2018.17.027]
　YAN Yu-mei.Correlation between Serum Cystatin C and Homocysteine and Type 2 Diabetic Retinopathy[J].Journal of Medical Information,2018,31(22):89.[doi:10.3969/j.issn.1006-1959.2018.17.027]
[3]罗妮娜,金昌德,陈文莉.中西医结合护理在老年糖尿病视网膜病变低视力患者中应用及效果观察[J].医学信息,2018,31(19):165.[doi:10.3969/j.issn.1006-1959.2018.19.052]
　LUO Ni-na,JIN Chang-de,CHEN Wen-li.Application and Effect of Integrated Traditional Chinese and Western Medicine Nursing in Elderly Patients with Diabetic Retinopathy with Low Visual[J].Journal of Medical Information,2018,31(22):165.[doi:10.3969/j.issn.1006-1959.2018.19.052]
[4]潘 浩.依帕司他片辅助治疗对糖尿病视网膜病变患者SOD、TAOC及TNF-α水平的影响[J].医学信息,2020,33(05):100.[doi:10.3969/j.issn.1006-1959.2020.05.030]
　PAN Hao.Effects of Adjuvant Epalrestat on SOD,TAOC and TNF-α Levels in Patients with Diabetic Retinopathy[J].Journal of Medical Information,2020,33(22):100.[doi:10.3969/j.issn.1006-1959.2020.05.030]
[5]李 娟,冯双双,康京京,等.尿β2微球蛋白水平与2型糖尿病患者视网膜病变和肾脏病变的相关性[J].医学信息,2020,33(19):99.[doi:10.3969/j.issn.1006-1959.2020.19.029]
　LI Juan,FENG Shuang-shuang,KANG Jing-jing,et al.Correlation Between Urine β2 Microglobulin Level and Retinopathy and Nephropathy in Type 2 Diabetic Patients[J].Journal of Medical Information,2020,33(22):99.[doi:10.3969/j.issn.1006-1959.2020.19.029]
[6]孙天慧,谭晓霞,连晓芬,等.2型糖尿病合并视网膜病变的影响因素分析[J].医学信息,2020,33(21):94.[doi:10.3969/j.issn.1006-1959.2020.21.027]
　SUN Tian-hui,TAN Xiao-xia,LIAN Xiao-fen,et al.Analysis of Influencing Factors of Type 2 Diabetes with Retinopathy[J].Journal of Medical Information,2020,33(22):94.[doi:10.3969/j.issn.1006-1959.2020.21.027]
[7]彭 越,高 宇,王云飞,等.VEGF在糖尿病视网膜病变发病机制中的作用[J].医学信息,2021,34(05):36.[doi:10.3969/j.issn.1006-1959.2021.05.011]
　PENG Yue,GAO Yu,WANG Yun-fei,et al.The Role of VEGF in the Pathogenesis of Diabetic Retinopathy[J].Journal of Medical Information,2021,34(22):36.[doi:10.3969/j.issn.1006-1959.2021.05.011]
[8]马 斌,王彩星,王金虹.基于文献计量学的我国2型糖尿病护理研究现状及热点分析[J].医学信息,2021,34(22):12.[doi:10.3969/j.issn.1006-1959.2021.22.004]
　MA Bin,WANG Cai-xing,WANG Jin-hong.Analysis of Current Status and Hot Spots in Nursing Research of Type 2 Diabetes Mellitus in China Based on Bibliometrics[J].Journal of Medical Information,2021,34(22):12.[doi:10.3969/j.issn.1006-1959.2021.22.004]
[9]宋晨音.他汀类药物在糖尿病治疗中的研究[J].医学信息,2022,35(17):166.[doi:10.3969/j.issn.1006-1959.2022.17.045]
　SONG Chen-yin.Study of Statins in the Treatment of Diabetes Mellitus[J].Journal of Medical Information,2022,35(22):166.[doi:10.3969/j.issn.1006-1959.2022.17.045]
[10]苟青青,李 鸿.糖尿病相关眼部改变及并发症的研究[J].医学信息,2022,35(18):153.[doi:10.3969/j.issn.1006-1959.2022.18.043]
　GOU Qing-qing,LI Hong.Study on Diabetes-related Ocular Changes and Complications[J].Journal of Medical Information,2022,35(22):153.[doi:10.3969/j.issn.1006-1959.2022.18.043]

更新日期/Last Update: 1900-01-01