机电工程技术2026，Vol.55Issue(1)：29-34,6.DOI:10.3969/j.issn.1009-9492.2026.01.006
基于机器学习的使用PD-1抑制剂后患者出现甲状腺障碍风险预测
Machine Learning Based Risk Prediction for Thyroid Disorders in Patients Using PD-1 Inhibitors
摘要
Abstract
A risk prediction model is constructed for thyroid dysfunction in cancer patients using PD-1 inhibitors,analysis is carried out on the risk factors related to thyroid dysfunction caused by the use of PD-1 tumor inhibitors,and a monitoring and early warning system is designed.The clinical data of 1225 cancer patients using PD-1 inhibitors in the Affiliated Tumor Hospital of Guangxi Medical University from 2020 to 2023 are selected,including 63 variables such as demographic characteristics,medical history,and laboratory tests.Four traditional machine learning models with the top 10/20/30/40/50/60 variables are selected for performance comparison.The performance of the above prediction models is evaluated by F1 score,sensitivity,accuracy,precision,and specificity area under the curve(AUC),and SHAP(Shapley additive explanation)is used to visualize the machine learning model.The top 10 variables correlated with thyroid stimulating hormone are:hydroxybutyrate dehydrogenase,lactate dehydrogenase,lymphocyte absolute value,aspartate transferase,calcium ion,alkaline phosphatase,glutamyl transpeptidase,monocyte absolute value,red blood cell distribution width SD,and cholinesterase.A risk prediction model is established for thyroid dysfunction in cancer patients using PD-1 inhibitors,and the influence of variables on the model prediction results is explained.关键词
PD1/甲状腺功能障碍/机器学习/Shapley加性解释(SHAP)Key words
PD1/thyroid dysfunction/machine learning/Shapley additive explanation(SHAP)分类
医药卫生引用本文复制引用
钟灿晖,赖信君,陈文戈,林璐,詹陆川..基于机器学习的使用PD-1抑制剂后患者出现甲状腺障碍风险预测[J].机电工程技术,2026,55(1):29-34,6.基金项目
广州市科技计划项目(2024B03J1293) （2024B03J1293）
茂名市科技计划项目(2022DZXHT035) （2022DZXHT035）
