医学信息2025，Vol.38Issue(14)：13-19,24,8.DOI:10.3969/j.issn.1006-1959.2025.14.002
基于Super Learner的前列腺癌风险预测模型构建与验证
Construction and Validation of Prostate Cancer Risk Prediction Model Based on Super Learner
摘要
Abstract
Objective To construct a risk prediction model of prostate cancer(PCa)by using super ensemble learning algorithm Super Learner,and to provide reference for early screening and early diagnosis and treatment of PCa by using machine learning algorithm.Methods Based on the prostate cancer early warning data set(n=1679)provided by the National Population Health Science Data Center,seven machine learning algorithms such as Super Learner were used to construct a PCa risk prediction model.According to the ratio of 7∶3,the data set was randomly divided into training set and verification set,and the model was constructed and verified respectively.Results A total of 7 machine learning PCa risk prediction models were constructed.In the validation set,the comprehensive performance of the Super Learner model was the best(AUC:0.762,sensitivity:0.752,specificity:0.696,positive predictive value:0.746,negative predictive value:0.702,Brier score:0.197),which was superior to the other six models.Total PSA,free PSA,inorganic phosphorus,creatinine and low-density lipoprotein cholesterol are the five most important variables for predicting PCa.Conclusion The PCa risk prediction model was successfully constructed,which can provide a scientific basis for the application of Super Learner algorithm in early screening and early diagnosis and treatment of PCa.关键词
前列腺癌/风险预测/机器学习/集成学习/可解释性Key words
Prostate cancer/Risk prediction/Machine learning/Integrated learning/Interpretability分类
医药卫生引用本文复制引用
靳帅,李静,路潜,肖倩,刘均娥..基于Super Learner的前列腺癌风险预测模型构建与验证[J].医学信息,2025,38(14):13-19,24,8.基金项目
1.北京市自然科学基金资助项目(编号:7244285) （编号:7244285）
2.首都医科大学校级自然科学基金项目(编号:PYZ23028) （编号:PYZ23028）
