北京中医药大学管理学院(北京 102488)
唐燕。E-mail:tangyan97_1017@sina.com
收稿:2026-03-09,
修回:2026-05-06,
网络首发:2026-08-17,
纸质出版:2026-08-20
移动端阅览
王家琪,韩爱庆,王苹,等.基于Stacking的前列腺癌预测模型[J].北京生物医学工程,2026,45(4):348-355.
WANG Jiaqi,HAN Aiqing,WANG Ping,et al.A Stacking-based prostate cancer prediction model[J].Beijing Biomedical Engineering,2026,45(4):348-355.
王家琪,韩爱庆,王苹,等.基于Stacking的前列腺癌预测模型[J].北京生物医学工程,2026,45(4):348-355. DOI: 10.3969/j.issn.1002-3208.2026.04.003.
WANG Jiaqi,HAN Aiqing,WANG Ping,et al.A Stacking-based prostate cancer prediction model[J].Beijing Biomedical Engineering,2026,45(4):348-355. DOI: 10.3969/j.issn.1002-3208.2026.04.003.
目的
2
构建基于可解释性机器学习的前列腺癌风险预测模型,筛选核心预警指标,为前列腺癌早期诊断提供客观依据。
方法
2
数据来源于国家人口健康数据中心的前列腺肿瘤预警数据集,共纳入3 000例样本,按病理诊断结果划分为患癌组与非患癌组;结合临床病理意义构建游离PSA/总PSA、载脂蛋白B/载脂蛋白A1两项核心比值特征;采用10轮独立重复实验设计,以随机森林(random forest, RF)、轻量级梯度提升机(light gradient boosting machine, LightGBM)
、
CatBoost和极端梯度提升(extreme gradient boosting,XGBoost)为基模型,通过贝叶斯超参数优化与5折分层交叉验证完成各基模型的参数寻优,在此基础上构建堆叠集成学习(Stacking)模型;采用ROC曲线下面积(area under the curve, AUC)、准确率(accuracy, ACC)、F1分数等多维度指标评估模型性能,同时结合沙普利可加性解释(Shapley additive explanations, SHAP),量化核心特征的重要性及对预测结果的影响程度。
结果
2
构建的Stacking集成模型综合性能表现最优,在独立测试集上AUC达0.758 4±0.023 9,整体性能优于各单一基模型;SHAP分析显示,游离PSA/总PSA比值为最重要的预测特征,对模型输出贡献最大,且在多轮实验中表现稳定。
结论
2
基于Stacking的集成学习模型具有良好的预测性能与稳健性,结合SHAP可解释性分析能够清晰揭示模型决策依据与关键风险指标,可为前列腺癌临床早期筛查及穿刺决策提供重要参考。
Objective
2
We build an interpretable machine learning model to predict prostate cancer risk. The model screens out core warning indicators. It provides an objective basis for early diagnosis of prostate cancer.
Methods
2
Data came from the prostate tumor warning dataset of the National Population Health Data Center. The dataset contained 3 000 samples. Based on pathological diagnosis, we divided the samples into a cancer group and a non-cancer group.From clinical pathological considerations, we constructed two core ratio features: free PSA to total PSA ratio, and apolipoprotein B to apolipoprotein A1 ratio. We then used a 10-round independent repeated experiment design. Four base models entered the study: random forest, LightGBM, CatBoost, and XGBoost. Bayesian hyperparameter optimization together with 5-fold stratified cross-validation tuned each base model. On this basis, we built a Stacking ensemble model.We evaluated model performance using multiple metrics, including AUC, accuracy, and F1 score. We also used the SHAP interpretation framework. SHAP quantified the importance of core features and their influence on prediction results.
Results
2
The Stacking ensemble model showed the best overall performance. On the independent test set, its AUC reached 0.758 4±0.023 9. It outperformed all the single base models. SHAP analysis revealed that the free PSA to total PSA ratio was the most important predictor. It contributed the most to the model output. This finding remained stable across multiple experiments.
Conclusions
2
The Stacking-based ensemble learning model has good predictive performance and robustness. Combined with SHAP interpretability, it clearly reveals the model’s decision basis and key risk indicators. This provides an important reference for early clinical screening and biopsy decisions in prostate cancer.
Han B, Zheng R, Zeng H, et al. Cancer incidence and mortality in China, 2022. Journal of the National Cancer Center, 2024, 4 (1): 47-53.
刘洋,王涌泉. PSA、PSAD、f/tPSA对前列腺癌的诊断及穿刺后Gleason评分的预测[J]. 医学研究杂志,2016,45(2): 116-120.
Liu Y, Wang YQ. PSA, PSAD, f/t PSA for diagnosis of prostate cancer and prediction of post needle biopsy Gleason score[J]. Journal of Medical Research, 2016, 45(2): 116-120.
刘爽,袁仁斌. 前列腺穿刺活检技术的演变过程[J].微创泌尿外科杂志, 2022, 11(2):139-143.
Liu S, Yuan RB. Evolution of prostate biopsy technology[J]. Journal of Minimally Invasive Urology, 2022, 11(2): 139-143.
彭岩,马铃,张文静,等. 基于集成学习的风险预测模型研究与应用[J]. 计算机工程与设计,2022,43(4): 956-961.
Peng Y, Ma L, Zhang WJ, et al. Research and application of risk forecast model based on ensemble learning[J]. Computer Engineering and Design, 2022, 43(4): 956-961.
路帅,李文杰,徐紫薇,等. 前列腺癌风险预测模型的构建与验证[J]. 重庆医科大学学报,2023,48(3):328-334.
Lu S, Li WJ, Xu ZW, et al. Construction and validation of a prostate cancer risk prediction model[J]. Journal of Chongqing Medical University, 2023, 48(3): 328-334.
Wang C, Chang QX, Wang XM, et al. Prostate cancer risk prediction and online calculation based on machine learning algorithm[J]. Chinese Medical Sciences Journal, 2022, 37 (3): 210-217.
Wolpert DH. Stacked generalization[J]. Neural Networks, 1992, 5(2): 241-259.
熊思伟,刘玉琳. 基于Borderline-SMOTE算法与Stacking集成学习的前列腺肿瘤风险预测研究[J]. 现代肿瘤医学,2023,31(16): 3075-3081.
Xiong SW, Liu YL. Risk prediction study of prostate tumors based on Borderline-SMOTE algorithm and Stacking ensemble learning[J]. Modern Oncology, 2023, 31(16): 3075-3081.
刘佳欣. 基于Stacking集成学习的前列腺癌预测模型研究[D]. 温州: 温州医科大学,2024.
Liu JX. Research on prostate cancer prediction model bas ed on Stacking integrated learning [D]. Wenzhou: Wenzhou Medical University, 2024.
孙悦,陈广新,于淼,等. 基于可解释性机器学习的新冠肺炎疾病风险预测研究[J]. 新一代信息技术, 2023, 6(12): 40-44.
Sun Y, Chen GX, Yu M, et al. Disease risk prediction of COVID-19 based on interpretive machine learning[J]. New Generation of Information Technology, 2023, 6(12): 40-44.
徐青,余冰,周佩敏,等. 基于机器学习与SHAP的全髋关节置换术患者下肢深静脉血栓可解释性预测模型构建研究[J]. 中国医院统计, 2024, 31 (1): 11-18, 24.
Xu Q, Yu B, Zhou PM, et al. Construction of interpretable prediction model for lower limb deep vein thrombosis in patients undergoing total hip arthroplasty based on machine learning and SHAP[J]. Chinese Journal of Hospital Statistics, 2024, 31 (1): 11-18.
吴佳琪,方素真,黎楷,等. 基于机器学习与SHAP可解释性住院老年2型糖尿病患者衰弱风险预测模型的构建及验证[J]. 护理学报,2025,32(11):7-12.
Wu JQ, Fang SZ, Li K, et al. Construction and validation of frailty risk prediction model for hospitalized elderly patients with type 2 diabetes based on machine learning and SHAP explainability[J]. Journal of Nursing (China), 2025, 32(11): 7-12.
富丹,孙悦,郭金兴,等. 基于可解释性机器学习的阿尔茨海默症痴呆风险预测[J]. 新一代信息技术, 2023, 6(24): 12-16.
Fu D, Sun Y, Guo JX, et al. Prediction of Alzheimer’s dementia risk based on interpretable machine learning[J]. New Generation of Information Technology, 2023, 6 (24): 12-16.
Lundberg SM, Erion G, Chen H, et al. From local explanations to global understanding with explainable AI for trees[J]. Nature Machine Intelligence, 2020, 2(1): 56-67.
Huang CG, Chen XY, Ma CQ, et al. Evaluation of serum FPSA/TPSA and hormone determination combined with nuclide bone imaging in diagnosis of prostate carcinoma and its recidivation and metastasis [J]. International Journal of Radiation Medicine and Nuclear Medicine, 2010, 34(6): 337-343.
王逸飞, 吴欢, 薛万国, 等. 前列腺癌与前列腺增生的分类预测及癌症风险因素分析[J]. 解放军医学院学报, 2021, 42(3): 277-281, 305.
Wang YF, Wu H, Xue WG, et al. Differentiating prostate cancer from prostate hyperplasia and related factors analysis[J]. Academic Journal of Chinese PLA Medical School, 2021, 42 (3): 277-281, 305.
Gui Q, Wu D, Xu F, et al. Exploring the positive association of blood lipid levels with prostate cancer risk and their relationship to pathological features in the Chinese population[J]. Medicine (Baltimore), 2025, 104(12): e41762.
Forman DT. The significance of creatine kinase (CKBB) in metastatic cancer of the prostate[J]. Annals of Clinical and Laboratory Science, 1979, 9(4): 333-337.
Yan YB. Creatine kinase in cell cycle regulation and cancer[J]. Amino Acids, 2016, 48(8): 1775-1784.
李宏基,赵晓龙,胡伟,等. 前列腺癌早期筛查风险预测模型的建立与验证研究[J].中国全科医学, 2024, 27(20): 2483-2490.
Li HJ, Zhao XL, Hu W, et al. Development and validation of a prediction model for prostate cancer early screening[J]. Chinese General Practice, 2024, 27(20): 2483-2490.
0
浏览量
0
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621