基于机器学习对亚实性IA期肺腺癌气道内播散的预测价值  

Applications of machine learning models in the prediction of tumor spread through air spaces in patients with stage IA part-solid lung adenocarcinoma

作　　者：刘欢欢[1] 陈业 周露 齐宏亮 Liu Huanhuan;Chen Ye;Zhou Lu;Qi Hongliang(Department of Radiotherapy,Nanfang Hospital,Southern Medical University,Guangzhou 510515,China;Department of Medical Engineering,Nanfang Hospital,Southern Medical University,Guangzhou 510515,China;Department of Radiology,Xiangtan Central Hospital,Xiangtan 411100,China)

机构地区：[1]南方医科大学南方医院放疗中心,广州510515 [2]南方医科大学南方医院医学工程科,广州510515 [3]湘潭市中心医院放射科,湘潭411100

出　　处：《现代仪器与医疗》2025年第1期37-42,共6页Modern Instruments & Medical Treatment

基　　金：南方医科大学南方医院院长基金项目(2022B030)。

摘　　要：目的探讨IA期亚实性肺腺癌中气道内播散(Spread Through Air Spaces,STAS)的临床特征及CT影像表现,并基于机器学习算法构建预测模型。方法本研究对2021年1月—2023年6月期间在湘潭市中心医院收治的IA期亚实性肺腺癌患者进行了回顾性分析。根据术后的病理结果,将患者分为STAS(+)组和STAS(-)组。使用包括Logistic回归、支持向量机(Support Vector Machine,SVM)、K近邻(K-Nearest Neighbour,KNN)、极端梯度提升(Extreme Gradient Boosting,Xgboost)、最小绝对收缩和选择算子(Least Absolute Shrinkage and Selection Operator,LASSO)、随机森林(Random Forest,RF)在内的五种机器学习算法来构建IA期亚实性肺腺癌的STAS预测模型。结果在106例IA期亚实性肺腺癌患者中,有16例(15%)被诊断为STAS。在对五种机器学习算法进行比较后,Xgboost在诊断准确性方面表现最佳,曲线下面积:0.969(95%CI:0.915~0.993)。Xgboost算法确定了IA期亚实性肺腺癌中STAS的以下关键预测因子,包括实性成分比例更高、实性部分的CT值、病灶大小、周围磨玻璃影(Ground-Glass Opacity,GGO)边界的模糊程度以及实性成分体积。结论通过分析患者的临床及影像学数据,Xgboost能够有效预测IA期亚实性肺腺癌的STAS,有利于临床决策。Objective This study aims to explore the clinical features and CT imaging manifestations of spread through air spaces(STAS)in stage IA subsolid lung adenocarcinoma and to construct a predictive model based on machine learning algorithms.Methods We retrospectively analyzed patients with stage IA subsolid lung adenocarcinoma admitted to Xiangtan Central Hospital from January 2021 to June 2023.Patients were classified into STAS(+)and STAS(-)groups based on postoperative pathological results.We employed five machine learning algorithms—Logistic Regression,Support Vector Machine(SVM),K-Nearest Neighbour(KNN),Extreme Gradient Boosting(Xgboost),Least Absolute Shrinkage and Selection Operator(LASSO),and Random Forest(RF)—to develop a predictive model for STAS in stage IA subsolid lung adenocarcinoma.Results Among the 106 patients with stage IA subsolid lung adenocarcinoma,16(15%)were diagnosed with STAS.After comparing the five machine learning algorithms,Xgboost demonstrated the highest diagnostic accuracy,achieving an area under the curve(AUC)of 0.969(95%CI:0.915-0.993).The Xgboost algorithm identified several key predictors of STAS in stage IA subsolid lung adenocarcinoma:higher proportion of solid component,CT value of the solid component,lesion size,degree of blurring in the surrounding ground-glass opacity(GGO)border,and volume of the solid component.Conclusion Through the analysis of clinical and imaging data,the Xgboost algorithm can effectively predict STAS in stage IA subsolid lung adenocarcinoma,providing valuable insights for clinical decision-making.

关 键 词：IA期亚实性肺腺癌 气道内播散 机器学习 预测模型1 

分 类 号：TH77[机械工程—仪器科学与技术] R737.9[机械工程—精密仪器及机械]