基于超声组学特征构建列线图预测早期乳腺癌前哨淋巴结转移  

Prediction of sentinel lymph node metastasis in early breast cancer by constructing nomogram based on ultrasonic radiomics features

作　　者：李鑫焱 刘菲菲[1] 焦玉婷 郑琪 许永波[1] 孙芳[1] LI Xinyan;LIU Feifei;JIAO Yuting;ZHENG Qi;XU Yongbo;SUN Fang(Department of Ultrasound,Binzhou Medical University Hospital,Shandong Binzhou 256603,China)

机构地区：[1]滨州医学院附属医院超声医学科,山东滨州256603

出　　处：《现代肿瘤医学》2023年第14期2682-2686,共5页Journal of Modern Oncology

基　　金：山东省医药卫生科技发展计划项目(编号:202009020663)。

摘　　要：目的:基于早期乳腺癌原发灶纹理特征建立超声组学模型并构建列线图预测前哨淋巴结转移状态,以为临床诊断及治疗提供指导意义。方法:回顾性分析2017年1月至2020年12月就诊于我院乳腺外科经术后病理证实为早期乳腺癌的222例患者的超声图像,使用开源成像平台FireVoxel进行手动分割,并自动提取超声组学特征。采用最小绝对收缩和选择算子(LASSO)回归算法及Logistic回归分析筛选变量并计算预测概率。基于提取的组学特征及预测概率绘制列线图,给予每个哑变量具体的分值。并绘制校准曲线评价列线图的预测性能,绘制决策曲线评价列线图的临床应用效能。采用1000次Bootstrap方法进行内部验证,计算平均AUC。结果:分割图像后共提取了859个超声组学特征,经过LASSO回归分析及Logistic回归分析筛选出5个超声组学特征。基于上述提取的组学特征绘制列线图,基于列线图建立的预测模型预测概率绘制ROC曲线,AUC为0.808(95%CI,0.751~0.865)。校准曲线显示该列线图预测的SLNM的发生概率与训练队列实际的SLNM发生概率之间有很好的一致性,决策曲线显示该列线图具有良好的临床应用效能。内部验证显示1000次Bootstrap迭代结果一致,平均AUC为0.810。结论:基于早期乳腺癌原发灶纹理特征建立超声组学模型并构建列线图可有效预测前哨淋巴结转移状态,从而为临床诊断及治疗提供指导意义。Objective:To establish the ultrasonic radiomics model and construct the nomogram to predict the status of sentinel lymph node metastasis based on the texture characteristics of the primary site of early breast cancer,so as to provide guidance for clinical diagnosis and treatment.Methods:The ultrasound images of 222 patients with early breast cancer confirmed by postoperative pathology admitted the breast surgery department of our hospital from January 2017 to December 2020 were retrospectively analyzed.FireVoxel,an open source imaging platform,was used for manual segmentation and automatic extraction of ultrasonic omics features.The minimum absolute contraction and selection operator(LASSO)regression algorithm and Logistic regression analysis were used to screen variables and calculate the predictive probability.A nomogram was drawn based on the screened radiomics features and predictive probability,and a specific score was assigned to each dummy variable.The calibration curve was drawn to evaluate the predictive performance of the nomogram,and the decision curve was drawn to evaluate the clinical application efficacy of thenomogram.The 1000 times Bootstrap method was used for internal verification,and the average AUC was calculated.Results:A total of 859 ultrasonic radiomics features were extracted after image segmentation,and 5 ultrasonic radiomics features were selected by LASSO regression analysis and Logistic regression analysis.The nomogram was drawn based on the above screened radiomics features,and the ROC curve was drawn based on the prediction probability of the prediction model established based on the nomogram,with an AUC of 0.808(95%CI,0.751~0.865).The calibration curve showed that the probability of SLNM predicted by the nomogram was in good agreement with the actual probability of SLNM in the training cohort,and the decision curve showed that the nomogram had good clinical application efficacy.Internal verification showed that 1000 Bootstrap iterations produced consistent results,with an average AU

关 键 词：早期乳腺癌 超声组学 前哨淋巴结 列线图 

分 类 号：R737.9[医药卫生—肿瘤]