超声影像组学预测雌激素受体及孕激素受体双阴性乳腺癌  

Ultrasound radiomics for predicting estrogen receptor and progesterone receptor double-negative breast cancer

作　　者：许荣[1] 欧阳秋芳[1] 林晴[1] 郭鹊晖[1] 刘磊磊[1] 肖凡 游涛[1] XU Rong;OUYANG Qiufang;LIN Qing;GUO Quehui;LIU Leilei;XIAO Fan;YOU Tao(Department of Ultrasound,the Second Affiliated Hospital of Fujian University of Traditional Chinese Medicine,Fuzhou 350003,China)

机构地区：[1]福建中医药大学附属第二人民医院超声科,福建福州350003

出　　处：《中国医学影像技术》2023年第9期1346-1349,共4页Chinese Journal of Medical Imaging Technology

基　　金：国家自然科学基金(82174469);福建省科协科技创新智库课题研究项目(FJKX-2022XKB037)。

摘　　要：目的观察超声影像组学预测雌激素受体(ER)及孕激素受体(PR)双阴性乳腺癌的价值。方法回顾性分析经病理确诊的342例乳腺癌359个病灶,326例可见单发、16例见多发病灶;其中119例见127个ER(-)PR(-)病灶、223例见232个其他病灶[36例共36个ER(+)PR(-)、2例共2个ER(-)PR(+)、185例共194个ER(+)PR(+)病灶];按照7∶3比例将病灶分为训练集(n=251)和测试集(n=108)。基于术前超声资料提取1314个病灶影像组学特征,经预处理后获得1205个特征;采用最小绝对收缩和选择算子(LASSO)算法筛选最佳影像组学特征,并利用支持向量机以训练集数据进行训练,构建预测ER及PR双阴性乳腺癌的影像组学模型;绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC),评估模型的诊断效能。结果共筛选出37个最佳影像组学特征,以之构建的影像组学模型预测训练集和测试集ER及PR双阴性乳腺癌的AUC分别为0.872[95%CI(0.820,0.924)]和0.867[95%CI(0.798,0.936)]。结论超声影像组学可有效预测ER及PR双阴性乳腺癌。Objective To observe the value of ultrasound radiomics for predicting estrogen receptor(ER)and progesterone receptor(PR)double-regative breast cancer.Methods Data of 359 lesions of breast cancer confirmed pathologically in 342 patients were retrospectively analyzed.There were 127 ER(-)and PR(-)lesions in 119 cases,232 other lesions in 223 cases,including 36 ER(+)and PR(-)lesions in 36 cases,2 ER(-)and PR(+)lesions in 2 cases with and 194 ER(+)and PR(+)lesions in 185 cases.Then the lesions were divided into training set(n=251)and validation set(n=108)at the ratio of 7∶3.Totally 1314 radiomics features were extracted from preoperative ultrasonograms of the lesions,and 1205 features were obtained after pre-processing.The least absolute shrinkage and selection operator(LASSO)algorithm was used to select the optimal radiomic features,then support vector machine was used for training using data in the training set,so as to construct the radiomic model for predicting ER and PR double-regative breast cancer.The receiver operating characteristic(ROC)curve was drawn,and area under the curve(AUC)was calculated to evaluate the diagnostic efficacy of this model.Results Totally 37 optimal radiomic features were screened out using LASSO algorithm.AUC of the constructed radiomic model for predicting ER and PR double-regative breast cancer in training set and validation set was 0.872(95%CI[0.820,0.924])and 0.867(95%CI[0.798,0.936]),respectively.Conclusion Ultrasound radiomics could effectively predict ER and PR double-regative breast cancer.

关 键 词：乳腺肿瘤 超声检查 受体 雌激素 受体 孕酮 影像组学 

分 类 号：R737.9[医药卫生—肿瘤] R445.1[医药卫生—临床医学]