Application of machine learning in predicting restenosis dysfunction after percutaneous transluminal angioplasty of internal arteriovenous fistula
10.3969/j.issn.1673-9701.2025.24.006
- VernacularTitle:机器学习预测动静脉内瘘经皮腔内血管成形术后再狭窄失功的应用
- Author:
Zemin WANG
1
;
Guojian SHAO
1
;
Yaqian CHENG
1
;
Shuting JIN
1
Author Information
1. 温州市中心医院肾内科,浙江温州 325000
- Publication Type:Journal Article
- Keywords:
Internal arteriovenous fistula;
Machine learning;
Internal fistula failure;
Percutaneous transluminal angioplasty;
Prediction model
- From:
China Modern Doctor
2025;63(24):25-28,79
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the risk factors analysis and prediction model establishment of restenosis dysfunction at 1 year after percutaneous transluminal angioplasty(PTA)of internal arteriovenous fistula based on machine learning.Methods A total of 322 patients who underwent PTA of internal arteriovenous fistula in Wenzhou Central Hospital from June 1,2018 to December 31,2023 were enrolled.The operation-related data were collected.Variables were used to construct prediction models using five machine learning algorithms:Random forest(RF),extreme gradient boosting(XGBoost),support vector machine(SVM),gradient boosting decision tree(GBDT)and Logistic regression(LR).The predictive efficacy was evaluated by area under receiver operating characteristic curve.Results There were 97 cases of restenosis dysfunction and 225 cases of non-dysfunction.The incidence of internal fistula restenosis dysfuction was 30.1％1 year after PTA.The age,diabetes,smoking,calcium-phosphorus product,dilatation pressure ≥20mmHg,and balloon diameter ≥6mnm in dysfunction group were higher than those in non-dysfunction group.The difference was statistically significant(P＜0.05).The area under the curve of RF,XGBoost,SVM,GBDT and LR models based on machine learning was 0.908(95％CI:0.836-0.980),0.809(95％CI:0.696-0.922),0.745(95％CI:0.624-0.867),0.711(95％CI:0.576-0.847)and 0.651(95％CI:0.508-0.795),respectively.The sensitivity was 79.1％,70.8％,83.3％,62.5％and 72.3％,respectively.The specificity was 89.0％,81.2％,57.8％,78.9％and 71.0％,respectively.Conclusion Age,diabetes mellitus,smoking,calcium-phosphorus product,expansion pressure ≥20mnmHg,balloon diameter ≥ 6mm are independent risk factors for restenosis failure after PTA in patients with internal arteriovenous fistula,which can be used as an index to predict restenosis failure 1 year after PTA in internal arteriovenous fistula.The random forest prediction model based on machine learning algorithm has good prediction performance and can better predict restenosis failure 1 year after PTA in internal arteriovenous fistula.
