Auxiliary diagnostic model of proliferative lupus nephritis based on machine learning algorithm
10.3760/cma.j.cn141217-20240425-00141
- VernacularTitle:基于机器学习算法建立增殖型狼疮肾炎辅助诊断模型
- Author:
Yaning WANG
1
;
Yang DONG
;
Na LI
;
Linlin LI
;
Lina ZHANG
;
Huixia CAO
;
Lei YAN
;
Fengmin SHAO
Author Information
1. 郑州大学人民医院（河南省人民医院）肾内科 河南省肾脏病免疫重点实验室 河南省肾病临床医学研究中心，郑州 450003
- Publication Type:Journal Article
- Keywords:
Machine learning;
Lupus nephritis;
Diagnosis;
Prediction model
- From:
Chinese Journal of Rheumatology
2025;29(1):31-37
- CountryChina
- Language:Chinese
-
Abstract:
Objective:This study aimed to construct a prediction model for diagnosis of proliferative lupus nephritis based on a machine learning algorithm. Additionally, a user-friendly platform was developed to propose a non-invasive method to assist the pathologic classification of lupus nephritis.Methods:A retrospective analysis was conducted on clinical and pathological data of lupus nephritis patients confirmed by renal biopsy at Zhengzhou University People′s Hospital from January 2017 to August 2023. The study population was randomly divided into training and testing sets in a 7∶3 ratio. Utilizing six machine learning algorithms, classification models were developed. The predictive performance of each model was assessed using metrics such as accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve (AUC). The optimal model, once identified, was deployed as a web-based calculator for convenient model application. SPSS 25.0 and R 4.2.2 were used to analyze the data.Results:The study included a total of 212 patients, with 138 cases with proliferative lupus nephritis and 74 cases with non-proliferative lupus nephritis. The AUC values for the six models, namely logistic regression, decision tree, random forest, support vector machine, extreme gradient boosting, and light gradient boosting machine, were 0.79, 0.62, 0.79, 0.88, 0.81, and 0.77, respectively; the accuracy rates were 82.54%, 65.08%, 74.60%, 85.71%, 69.84%, 71.43%, respectively. Among them, the support vector machine model demonstrated the optimal performance. This model had deployed as a web-based calculator. Based on feature importance scores, the top 10 influencing factors were identified, including anti URNP antibody, immunoglobulin G, serum globulin, estimated glomerular filtration rate, anti Smith antibody, BMI index, anti dsDNA antibody, uric acid, anti-Rib.p antibody, and gender.Conclusion:A prediction model based on machine learning algorithms was successfully established, and a web calculator was developed to offer a simple and non-invasive method for diagnosing proliferative lupus nephritis. This can assist clinicians in evaluating the risk-benefit ratio of kidney biopsy in patients with lupus nephritis.
