基于超声指标的乳腺结节良恶性预测模型的建立与验证  

Establishment and validation of a predictive model for benign and malignant breast nodules based on ultrasound indicators

作　　者：周文秀 阿衣尔红 夏宇 曾维佳 葛斌 ZHOU Wenxiu;Ayierhong;XIA Yu;ZENG Weijia;GE Bin(Department of Ultrasound,Pidu District People’s Hospital Affiliated to Chengdu Medical College,Chengdu 611730,China)

机构地区：[1]成都医学院第三附属医院、成都市郫都区人民医院超声科,成都611730 [2]成都医学院第三附属医院、成都市郫都区人民医院检验科,611730

出　　处：《临床肿瘤学杂志》2024年第9期854-858,共5页Chinese Clinical Oncology

基　　金：四川省卫健委科研课题资助项目(19PJ181)。

摘　　要：目的建立基于超声指标的乳腺结节良恶性预测模型,并进行验证。方法收集2021年1月至2023年7月288例BI-RADS 4类及以上乳腺结节患者,以7∶3的比例分为模型组201例,验证组87例。患者均行常规多普勒超声、超声造影及弹性成像检查。以病理结果为金标准,将模型组患者分为良性组与恶性组,比较两组各项超声检查指标。采用LASSO回归筛选潜在变量后行多因素分析,以此建立列线图模型并验证。结果模型组201例患者中有66例被组织病理学证实为恶性结节。在LASSO回归基础上行多因素Logisitc回归分析,结果显示,形态、钙化、纵横比、腋下肿大淋巴结、达峰时间、消退时间和平均弹性值为影响预测乳腺结节良恶性的独立因素(P<0.05)。模型验证结果显示,模型组受试者工作特征(ROC)曲线下面积为0.924(95%CI:0.883~0.964),灵敏度为79.3%,特异度为93.9%;验证组ROC曲线下面积为0.917(95%CI:0.875~0.959),灵敏度为80.7%,特异度为89.4%。校准曲线斜率为1,截距为0.000,模型曲线与理想模型基本拟合成对角线。有效性分析结果显示,当预测概率阈值为0.10~0.90时,使用本研究模型预测乳腺结节良恶性的净获益最高。结论本研究建立的模型可用于预测乳腺结节良恶性,可为此类患者良恶性筛查提供参考。Objective To cunstruct and validate a predictive model to differentiate benign and malignant breast nodule based on ultrasound indicators.Methods A total of 288 patients with BI-RADS grade 4 and above from January 2021 to July 2023 were selected and divided into model group(n=201)and validation group(n=87)at a 7∶3 ratio.Patients underwent conventional Doppler ultrasound,ultrasonography and elastography.Patients in the model group were divided into benign and malignant groups using pathological examination as the gold standard,and the ultrasound indicators of the two groups were compared.LASSO regression was used to screen for latent variables and then a multivariate analysis was performed,which was used to develop and validate the column-line graphical model.Results Among the 201 patients in the model group,66 cases were confirmed as malignant nodules by pathological examination.The results of Logisitc regression analysis performed on the basis of LASSO regression showed that morphology,calcification,aspect ratio,axillary enlarged lymph nodes,time to peak,time to regress,and mean elasticity value were independent influencing factors for predicting benign or malignant breast nodules(P<0.05).The results of model validation showed that the area under the receiver operating characteristic(ROC)curve in the model group was 0.924(95%CI:0.883-0.964),with a sensitivity of 79.3%and a specificity of 93.9%;and in the validation group,the area under the ROC curve was 0.917(95%CI:0.875-0.959),with a sensitivity of 80.7% and a specificity of 89.4%.The calibration curve had a slope of 1 and an intercept of 0.000,and the model curve was essentially fitted diagonally to the ideal model.The effectiveness analysis results showed that when the prediction probability threshold was between 0.10 and 0.90,the net benefit of using this model to predict the benign and malignant breast nodules was the highest.Conclusion The model constructed in this study can be used to predict the benign and malignant breast nodules and provide a referen

关 键 词：乳腺结节 超声 多因素分析 预测模型 

分 类 号：R737.9[医药卫生—肿瘤]