基于SEER数据库使用支持向量机和列线图对骨肉瘤诊断时肺转移风险预测模型的建立与验证  

Development and validation of a predictive model of pulmonary metastasis risk in diagnosis of osteosarcoma by support vector machine and nomogram based on SEER database

作　　者：林跃玮 刘文渤 胡志锋 梁超伦 高明 黄永明 陈耿鑫 Lin Yuewei;Liu Wenbo;Hu Zhifeng;Liang Chaolun;Gao Ming;Huang Yongming;Chen Gengxin(The Second Clinical Medical College,Guangzhou University of Chinese Medicine,Guangzhou 510006,China;Department of Orthopaedics,the Second Affiliated Hospital of Guangzhou University of Chinese Medicine,Guangzhou 510006,China)

机构地区：[1]广州中医药大学第二临床医学院,广东广州510006 [2]广州中医药大学第二附属医院骨伤科,广东广州510006

出　　处：《实用肿瘤杂志》2023年第3期251-257,共7页Journal of Practical Oncology

基　　金：广东省自然科学基金面上项目(2018A030313643)。

摘　　要：目的开发列线图和支持向量机模型预测骨肉瘤患者的肺转移风险。方法从美国“监测,流行病学和结果”(Surveillance,Epidemiology and End Results,SEER)数据库获得2010年至2016年的原发性骨肉瘤患者数据,通过单因素和多因素Logistic回归进行筛选,确定肺转移的危险因素。基于Logistic回归结果,构建列线图,使用10-fold cross validation对其进行验证。按照8∶2的比例,将数据拆分为训练集和验证集,使用支持向量机构建预测模型。绘制受试者工作特征(receiver operating characteristic,ROC)曲线比较列线图和支持向量机模型的预测能力。结果共纳入1038例骨肉瘤患者。Logistic回归分析显示,男性患者、更大的肿瘤直径、N1或NX分期、伴有骨转移和原发部位为四肢骨骼为骨肉瘤患者出现肺转移的独立危险因素(均P<0.05)。列线图预测骨肉瘤肺转移的最佳ROC曲线下面积(area under the curve,AUC)为0.769,平均AUC为0.695,支持向量机模型的AUC为0.792,预测准确率为0.861。支持向量机模型具有更好的预测能力。结论男性患者、更大的肿瘤体积、N1或者NX分期、伴有骨转移和原发部位为四肢骨骼为骨肉瘤患者的肺转移独立危险因素。支持向量机模型和列线图均有较好的预测准确率。支持向量机模型比传统列线图有更好的预测能力,列线图则有更好的临床实用性。Objective To develop nomogram and support vector machine model to predict the risk of pulmonary metastasis in patients with osteosarcoma.Methods Data of patients with osteosarcoma were obtained from the Surveillance,Epidemiology,and End R esults(SEER)database from 2010 to 2016.Univariate and multivariate logistic regression was used to identify the risk factors of pulmonary me-tastasis.Based on the logistic regression results,a nomogram was constructed and validated using a 10-fold cross validation.The data were divided into training and validation sets according to the ratio of 8∶2,and the prediction model was constructed using the support vector machine.A receiver operating characteristic(ROC)curve was plotted to compare the predictive ability of the nomogram and the support vector machine model.Results A total of 1038 patients with osteosarcoma were included.Logistic regression analysis showed that male patients,larger tumor diameter,N1 or NX staging,concomitant bone metastasis and primary site of extremity bone were independent risk factors for the development of pulmonary metastasis in patients with osteosarcoma(all P<0.05).The optimal area under the curve(AUC)for the nomogram predicting pulmonary metastasis of osteosarcoma was 0.769,the average AUC was 0.695,and the AUC of the support vector machine model was 0.792.The prediction accuracy was 0.861.The support vector machine model had better predictive power.Conclu-sions Male patients,larger tumor diameter,N1 or NX staging,presence of bone metastasis and primary site of bone in the extremities are independent risk factors for pulmonary metastasis in patients with osteosarcoma.Both the support vector machine model and the nomogram have good prediction accuracy.The support vector machine model has better predictive ability than the nomogram,while the nomogram has better clinical utility.

关 键 词：骨肉瘤 肺转移 列线图 支持向量机 SEER数据库 

分 类 号：R738.1[医药卫生—肿瘤]