Ultrasound-based deep learning in the establishment of a breast lesion risk stratification system: a multicenter study

作者全名："Gu, Yang; Xu, Wen; Liu, Ting; An, Xing; Tian, Jiawei; Ran, Haitao; Ren, Weidong; Chang, Cai; Yuan, Jianjun; Kang, Chunsong; Deng, Youbin; Wang, Hui; Luo, Baoming; Guo, Shenglan; Zhou, Qi; Xue, Ensheng; Zhan, Weiwei; Zhou, Qing; Li, Jie; Zhou, Ping; Chen, Man; Gu, Ying; Chen, Wu; Zhang, Yuhong; Li, Jianchu; Cong, Longfei; Zhu, Lei; Wang, Hongyan; Jiang, Yuxin"

作者地址："[Gu, Yang; Xu, Wen; Li, Jianchu; Wang, Hongyan; Jiang, Yuxin] Chinese Acad Med Sci & Peking Union Med Coll, Peking Union Med Coll Hosp, Dept Ultrasound, 1 Shuai Fu Yuan, Beijing 100730, Peoples R China; [Liu, Ting; An, Xing; Cong, Longfei] Shenzhen Mindray Biomed Elect Co Ltd, Dept Med Imaging Adv Res, Beijing, Peoples R China; [Tian, Jiawei] Harbin Med Univ, Dept Ultrasound, Affiliated Hosp 2, Harbin, Peoples R China; [Ran, Haitao] Chongqing Med Univ, Dept Ultrasound, Affiliated Hosp 2, Chongqing, Peoples R China; [Ran, Haitao] Chongqing Key Lab Ultrasound Mol Imaging, Chongqing, Peoples R China; [Ren, Weidong] China Med Univ, Dept Ultrasound, Shengjing Hosp, Shenyang, Peoples R China; [Chang, Cai] Fudan Univ, Shanghai Canc Ctr, Dept Med Ultrasound, Shanghai, Peoples R China; [Yuan, Jianjun] Henan Prov People S Hosp, Dept Ultrasonog, Zhengzhou, Peoples R China; [Kang, Chunsong] Shanxi Bethune Hosp, Shanxi Acad Med Sci, Dept Ultrasound, Taiyuan, Peoples R China; [Deng, Youbin] Huazhong Univ Sci & Technol, Tongji Hosp, Dept Med Ultrasound, Tongji Med Coll, Wuhan, Peoples R China; [Wang, Hui] Jilin Univ, Dept Ultrasound, China Japan Union Hosp, Changchun, Peoples R China; [Luo, Baoming] Sun Yat Sen Univ, Sun Yat Sen Mem Hosp, Dept Ultrasound, Guangzhou, Peoples R China; [Guo, Shenglan] Guangxi Med Univ, Dept Ultrasonog, Affiliated Hosp 1, Nanning, Peoples R China; [Zhou, Qi] Xi An Jiao Tong Univ, Affiliated Hosp 2, Sch Med, Dept Med Ultrasound, Xian, Peoples R China; [Xue, Ensheng] Fujian Med Univ, Fujian Inst Ultrasound Med, Dept Ultrasound, Union Hosp, Fuzhou, Peoples R China; [Zhan, Weiwei] Shanghai Jiao Tong Univ, Ruijin Hosp, Sch Med, Dept Ultrasound, Shanghai, Peoples R China; [Zhou, Qing] Wuhan Univ, Dept Ultrasonog, Renmin Hosp, Wuhan, Peoples R China; [Li, Jie] Shandong Univ, Qilu Hosp, Dept Ultrasound, Jinan, Peoples R China; [Zhou, Ping] Cent South Univ, Dept Ultrasound, Xiangya Hosp 3, Changsha, Peoples R China; [Chen, Man] Shanghai Jiao Tong Univ, Tongren Hosp, Dept Ultrasound Med, Sch Med, Shanghai, Peoples R China; [Gu, Ying] Guizhou Med Univ, Dept Ultrasonog, Affiliated Hosp, Guiyang, Peoples R China; [Chen, Wu] Shanxi Med Univ, Dept Ultrasound, Hosp 1, Taiyuan, Peoples R China; [Zhang, Yuhong] Dalian Med Univ, Dept Ultrasound, Hosp 2, Dalian, Peoples R China; [Zhu, Lei] Shenzhen Mindray Biomed Elect Co Ltd, Dept Med Imaging Adv Res, Shenzhen, Peoples R China"

通信作者："Wang, HY; Jiang, YX (通讯作者)，Chinese Acad Med Sci & Peking Union Med Coll, Peking Union Med Coll Hosp, Dept Ultrasound, 1 Shuai Fu Yuan, Beijing 100730, Peoples R China."

来源：EUROPEAN RADIOLOGY

ESI学科分类：CLINICAL MEDICINE

WOS号：WOS:000886855200002

JCR分区：Q1

影响因子：5.9

年份：2022

卷号：　

期号：　

开始页：　

结束页：　

文献类型：Article; Early Access

关键词：Artificial intelligence; Deep learning; Ultrasonography; Breast neoplasms; Diagnosis

摘要："Objectives To establish a breast lesion risk stratification system using ultrasound images to predict breast malignancy and assess Breast Imaging Reporting and Data System (BI-RADS) categories simultaneously. Methods This multicenter study prospectively collected a dataset of ultrasound images for 5012 patients at thirty-two hospitals from December 2018 to December 2020. A deep learning (DL) model was developed to conduct binary categorization (benign and malignant) and BI-RADS categories (2, 3, 4a, 4b, 4c, and 5) simultaneously. The training set of 4212 patients and the internal test set of 416 patients were from thirty hospitals. The remaining two hospitals with 384 patients were used as an external test set. Three experienced radiologists performed a reader study on 324 patients randomly selected from the test sets. We compared the performance of the DL model with that of three radiologists and the consensus of the three radiologists. Results In the external test set, the DL model achieved areas under the receiver operating characteristic curve (AUCs) of 0.980 and 0.945 for the binary categorization and six-way categorizations, respectively. In the reader study set, the DL BI-RADS categories achieved a similar AUC (0.901 vs. 0.933, p = 0.0632), sensitivity (90.98% vs. 95.90%, p = 0.1094), and accuracy (83.33% vs. 79.01%, p = 0.0541), but higher specificity (78.71% vs. 68.81%, p = 0.0012) than those of the consensus of the three radiologists. Conclusions The DL model performed well in distinguishing benign from malignant breast lesions and yielded outcomes similar to experienced radiologists. This indicates the potential applicability of the DL model in clinical diagnosis."

基金机构：Beijing Natural Science Foundation [7202156]; Foundation of International Health Exchange and Cooperation Center NHC PRC [ihecc2018C0032-2]

基金资助正文："This work is supported by the Beijing Natural Science Foundation (7202156), and the Foundation of International Health Exchange and Cooperation Center NHC PRC (ihecc2018C0032-2)."