基于可解释机器学习的前列腺癌风险预测模型构建

Development and Validation of Machine Learning Models for Predicting Prostate Cancer Risk: A Multi-Cohort Study

摘要

目的：通过使用临床医学数据，构建并验证多种机器学习预测模型，用于前列腺癌风险预测，旨在筛选出最佳性能模型，并验证其在国内外独立数据集的泛化能力，为前列腺癌的临床决策提供稳健、可信的工具。方法：选择国家人口健康科学数据中心的前列腺肿瘤患者临床数据，经高级特征筛选移除噪声数据后，按7:3划分训练集和测试集，构建5种机器学习模型，包括逻辑回归（logistic regression，LR）、极端梯度提升（extreme gradient boosting，XGBoost）、随机森林（random forest，RF）、分类特征梯度提升（categorical boosting，CatBoost）、K近邻（K-nearest neighbors，KNN）。在测试集上采用受试者工作特征曲线下面积（area under the receiver operating characteristic curve，AUC）、准确率、召回率及F1分数综合评价模型性能，同时评估模型在国外患者独立测试集的表现，最后利用SHAP对最佳模型进行特征重要性解析。结果：在五种模型中，XGBoost模型在内部测试集上表现最佳，其AUC、准确率、召回率及F1分数分别为0.894、0.852、0.825、0.834。SHAP分析显示，总PSA水平、碱性磷酸酶、年龄、肌酸激酶同工酶、游离总前列腺特异性抗原（prostate specific antigen，PSA）比值为最关键特征。同时，XGBoost模型在独立测试集上同样有着优异的性能（AUC:0.810），证明了其强大的泛化能力。结论：本研究构建了一个基于XGBoost的高性能、可解释的前列腺癌风险预测模型。该模型不仅在中国人群中表现出色，同时在国际数据中也表现出良好的普适性，可以为临床医生提供精准的风险分层，从而优化诊疗策略。

Abstract

Objective: To develop and validate machine learning models for predicting prostate cancer risk using clinical data, aiming to identify the optimal model and verify its generalizability across independent domestic and international cohorts, thereby providing a robust tool for clinical decision support. Methods: Clinical data of patients with prostate tumors were obtained from the National Population Health Science Data Center. Following advanced feature selection to eliminate noise, the data were randomly split into training and testing sets at a 7:3 ratio. Five machine learning models were constructed: logistic regression (LR), extreme gradient boosting (XGBoost), random forest (RF), categorical boosting (CatBoost), and K-nearest neighbors (KNN). Model performance was evaluated on the internal test set using the area under the receiver operating characteristic curve (AUC), accuracy, recall, and F1-score. The top-performing model was further validated on an independent external test set comprising international patient data. Finally, shapley additive exPlanations (SHAP) were employed to interpret the feature importance of the best model. Results: Among the five models, XGBoost demonstrated superior performance on the internal test set, with an AUC of 0.894, accuracy of 0.852, recall of 0.825, and an F1-score of 0.834. SHAP analysis revealed that total PSA, alkaline phosphatase, age, creatine kinase isoenzyme, and the free-to-total PSA ratio were the most critical predictive features. Notably, the XGBoost model maintained excellent performance on the independent external test set (AUC: 0.810), confirming its strong generalizability. Conclusion: This study developed a high-performance and interpretable XGBoost-based model for prostate cancer risk prediction. The model exhibited outstanding discriminative ability within the Chinese population and sustained good generalizability in an international cohort. It holds promise as a practical tool to assist clinicians in precise risk stratification and optimization of diagnosis and treatment strategies.

关键词

前列腺癌 / 机器学习 / SHAP / XGBoost

Key words

prostate cancer / machine learning / SHAP / XGBoost

参考文献

[1] 张美慧, 程华, 宁浩勇. 前列腺癌早期筛查方法研究进展[J]. 中华男科学杂志, 2025, 31(8): 737-741.

[2] 钟鑫威, 陈漪, 岳博文, 等. 双源CT虚拟单能量成像鉴别诊断前列腺癌与前列腺增生[J]. 中国医学影像技术, 2024, 40(11): 1749-1753.

[3] 黄勇, 周福林, 李静, 等. 高危局限性前列腺癌术前最大雄激素阻断治疗获益的临床预测[J]. 重庆医科大学学报, 2025, 50(04): 463-469.

[4] Wang Hua, Li Wenjin, Deng Weiming, et al. Integrated analysis of single‐cell RNA sequencing and machine learning reveals a t cell‐specific panoptosis signature predicting prognosis and immunotherapy in prostate cancer[J]. Human Mutation, 2025, 2025(1): 8889021.

[5] Liu Tongpeng, Yao Yu, Hu Yang, et al. Integrating multi—omics and machine learning to decipher the role of GSTP1 in endocrine—disrupting chemical—induced prostate cancer pathogenesis[J]. Eur J Pharmacol, 2025, 1008: 178335.

[6] Ahmed Al Marouf, Tarek A Bismar, Sunita Ghosh, et al. Leveraging machine learning for severity level—wise biomarker identification in prostate cancer microarray gene expression data[J]. Biomedicines, 2025, 13(10): 2350.

[7] Linkon Ali Hasan Md, Labib Mahir, Hasan Tarik, et al. Deep learning in prostate cancer diagnosis and gleason grading in histopathology images:an extensive study[J]. Informatics Medicine Unlocked, 2021,(prepublish): 100582.

[8] Salman Mehmet Emin, Çakirsoy Çakar Gözde, Azimjonov Jahongir, et al. Automated prostate cancer grading and diagnosis system using deep learning—based Yolo object detection algorithm[J]. Expert Systems With Applications, 2022: 201.

[9] Kang Dae Y, DeYoung Pamela N, Tantiongloc Justin, et al. Statistical uncertainty quantification to augment clinical decision support:a first implementation in sleep medicine[J]. NPJ digital medicine, 2021, 4(1): 142.

[10] Hany F Atlam, Gbenga Ebenezer Aderibigbe, Muhammad Shahroz Nadeem. Effective epileptic seizure detection with hybrid feature selection and SMOTE—based data balancing using SVM classifier[J]. Applied Sciences, 2025, 15(9): 4690.

[11] Stacy Loeb, William J Catalona. The prostate health index:a new test for the detection of prostate cancer[J]. Ther Adv Urol, 2014, 6(2): 74-77.

[12] Andrew W Roddam, Michael J Duffy, Freddie C Hamdy. Use of prostate—specific antigen(PSA)isoforms for the detection of prostate cancer in men with a PSA level of 2—10 ng/ml:systematic review and meta—analysis[J]. Eur Urol, 48(3): 386-399.

[13] Cui Feilun, Zhang Yueshi, Liu Ziyan, et al. Association between overweight and obesity determined by body mass index and overall survival in patients with metastatic prostate cancer:a meta—analysis[J]. Int J Obes, 2025, 49(11): 2131-2139.

[14] Juan Morote, Berta Miró, Patricia Hernando, et al. Developing a predictive model for significant prostate cancer detection in prostatic biopsies from seven clinical variables:is machine learning superior to logistic regression?[J]. Cancers, 2025, 17(7): 1101—1101.

[15] 陈敏, 陈竹碧, 于红梅, 等. 基于多参数MRI的PI—RADS评分与PSA衍生参数构建列线图预测临床显著前列腺癌的诊断价值[J]. 临床放射学杂志, 2025, 44(11): 2126-2132.

[16] Miroslav Stojadinović, Nebojša Jurišević, Milorad Stojadinović, et al. Enhancing the prediction of clinically significant prostate cancer through feature engineering[J]. J Medical Biological Engineering, 2025,(prepublish): 1-9.

[17] 胡尘翰, 乔晓梦, 胡冀苏, 等. 基于双参数MRI的深度学习—临床混合模型对临床显著性前列腺癌诊断价值的研究[J]. 磁共振成像, 2024, 15(2): 90-96.

[18] 阿不都克尤木·麦麦提依明. 前列腺癌数据集的构建及基于机器学习的新模型和基因特征预测研究[D]. 新疆医科大学, 2024.

[19] Shweta Singh, Abhay Kumar Pathak, Sukhad Kural, et al. Integrating miRNA profiling and machine learning for improved prostate cancer diagnosis[J]. Scientific Reports, 2025, 15(1): 30477—30477.

[20] 尹燕伟, 于成龙, 陈广新, 等. 基于机器学习算法的前列腺癌疾病诊断标志物筛选分析[J]. 中国医药科学, 2023, 13(15): 136-140.

[21] Jensen Carina, Carl Jesper, Boesen Lars, et al. Assessment of prostate cancer prognostic Gleason grade group using zonal—specific features extracted from biparametric MRI using a KNN classifier.[J]. J Applied Clinical Medical Physics, 2019, 20(2): 146-153.

基金资助

河北省自然科学基金资助项目(H2021405012)