Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study
Abstract Background Studies on deep learning (DL)-based models in breast ultrasound (US) remain at the early stage due to a lack of large datasets for training and independent test sets for verification. We aimed to develop a DL model for differentiating benign from malignant breast lesions on US us...
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2022-07-01
|
| Series: | Insights into Imaging |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s13244-022-01259-8 |
| _version_ | 1828778244306894848 |
|---|---|
| author | Yang Gu Wen Xu Bin Lin Xing An Jiawei Tian Haitao Ran Weidong Ren Cai Chang Jianjun Yuan Chunsong Kang Youbin Deng Hui Wang Baoming Luo Shenglan Guo Qi Zhou Ensheng Xue Weiwei Zhan Qing Zhou Jie Li Ping Zhou Man Chen Ying Gu Wu Chen Yuhong Zhang Jianchu Li Longfei Cong Lei Zhu Hongyan Wang Yuxin Jiang |
| author_facet | Yang Gu Wen Xu Bin Lin Xing An Jiawei Tian Haitao Ran Weidong Ren Cai Chang Jianjun Yuan Chunsong Kang Youbin Deng Hui Wang Baoming Luo Shenglan Guo Qi Zhou Ensheng Xue Weiwei Zhan Qing Zhou Jie Li Ping Zhou Man Chen Ying Gu Wu Chen Yuhong Zhang Jianchu Li Longfei Cong Lei Zhu Hongyan Wang Yuxin Jiang |
| author_sort | Yang Gu |
| collection | DOAJ |
| description | Abstract Background Studies on deep learning (DL)-based models in breast ultrasound (US) remain at the early stage due to a lack of large datasets for training and independent test sets for verification. We aimed to develop a DL model for differentiating benign from malignant breast lesions on US using a large multicenter dataset and explore the model’s ability to assist the radiologists. Methods A total of 14,043 US images from 5012 women were prospectively collected from 32 hospitals. To develop the DL model, the patients from 30 hospitals were randomly divided into a training cohort (n = 4149) and an internal test cohort (n = 466). The remaining 2 hospitals (n = 397) were used as the external test cohorts (ETC). We compared the model with the prospective Breast Imaging Reporting and Data System assessment and five radiologists. We also explored the model’s ability to assist the radiologists using two different methods. Results The model demonstrated excellent diagnostic performance with the ETC, with a high area under the receiver operating characteristic curve (AUC, 0.913), sensitivity (88.84%), specificity (83.77%), and accuracy (86.40%). In the comparison set, the AUC was similar to that of the expert (p = 0.5629) and one experienced radiologist (p = 0.2112) and significantly higher than that of three inexperienced radiologists (p < 0.01). After model assistance, the accuracies and specificities of the radiologists were substantially improved without loss in sensitivities. Conclusions The DL model yielded satisfactory predictions in distinguishing benign from malignant breast lesions. The model showed the potential value in improving the diagnosis of breast lesions by radiologists. |
| first_indexed | 2024-12-11T16:38:21Z |
| format | Article |
| id | doaj.art-cabbdfcb2d8e411097d0d7dfa83805b5 |
| institution | Directory Open Access Journal |
| issn | 1869-4101 |
| language | English |
| last_indexed | 2024-12-11T16:38:21Z |
| publishDate | 2022-07-01 |
| publisher | SpringerOpen |
| record_format | Article |
| series | Insights into Imaging |
| spelling | doaj.art-cabbdfcb2d8e411097d0d7dfa83805b52022-12-22T00:58:24ZengSpringerOpenInsights into Imaging1869-41012022-07-0113111410.1186/s13244-022-01259-8Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic studyYang Gu0Wen Xu1Bin Lin2Xing An3Jiawei Tian4Haitao Ran5Weidong Ren6Cai Chang7Jianjun Yuan8Chunsong Kang9Youbin Deng10Hui Wang11Baoming Luo12Shenglan Guo13Qi Zhou14Ensheng Xue15Weiwei Zhan16Qing Zhou17Jie Li18Ping Zhou19Man Chen20Ying Gu21Wu Chen22Yuhong Zhang23Jianchu Li24Longfei Cong25Lei Zhu26Hongyan Wang27Yuxin Jiang28Department of Ultrasound, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical CollegeDepartment of Ultrasound, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical CollegeDepartment of Medical Imaging Advanced Research, Beijing Research Institute, Shenzhen Mindray Bio-Medical Electronics Co., Ltd.Department of Medical Imaging Advanced Research, Beijing Research Institute, Shenzhen Mindray Bio-Medical Electronics Co., Ltd.Department of Ultrasound, The Second Affiliated Hospital of Harbin Medical UniversityDepartment of Ultrasound, The Second Affiliated Hospital of Chongqing Medical University and Chongqing Key Laboratory of Ultrasound Molecular ImagingDepartment of Ultrasound, Shengjing Hospital of China Medical UniversityDepartment of Medical Ultrasound, Fudan University Shanghai Cancer CenterDepartment of Ultrasonography, Henan Provincial People’s HospitalDepartment of Ultrasound, Shanxi Bethune Hospital, Shanxi Academy of Medical SciencesDepartment of Medical Ultrasound, Tongji Hospital, Tongji Medical College of Huazhong University of Science and TechnologyDepartment of Ultrasound, China-Japan Union Hospital of Jilin UniversityDepartment of Ultrasound, The Sun Yat-Sen Memorial Hospital, Sun Yat-Sen UniversityDepartment of Ultrasonography, First Affiliated Hospital of Guangxi Medical UniversityDepartment of Medical Ultrasound, The Second Affiliated Hospital, School of Medicine, Xi’an Jiaotong UniversityDepartment of Ultrasound, Union Hospital of Fujian Medical University, Fujian Institute of Ultrasound MedicineDepartment of Ultrasound, Ruijin Hospital, Shanghai Jiaotong University, School of MedicineDepartment of Ultrasonography, Renmin Hospital of Wuhan UniversityDepartment of Ultrasound, Qilu Hospital, Shandong UniversityDepartment of Ultrasound, The Third Xiangya Hospital of Central South UniversityDepartment of Ultrasound Medicine, Tongren Hospital, Shanghai Jiao Tong University School of MedicineDepartment of Ultrasonography, The Affiliated Hospital of Guizhou Medical UniversityDepartment of Ultrasound, The First Hospital of Shanxi Medical UniversityDepartment of Ultrasound, The Second Hospital of Dalian Medical UniversityDepartment of Ultrasound, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical CollegeDepartment of Medical Imaging Advanced Research, Beijing Research Institute, Shenzhen Mindray Bio-Medical Electronics Co., Ltd.Department of Medical Imaging Advanced Research, Shenzhen Mindray Bio-Medical Electronics Co., Ltd.Department of Ultrasound, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical CollegeDepartment of Ultrasound, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences and Peking Union Medical CollegeAbstract Background Studies on deep learning (DL)-based models in breast ultrasound (US) remain at the early stage due to a lack of large datasets for training and independent test sets for verification. We aimed to develop a DL model for differentiating benign from malignant breast lesions on US using a large multicenter dataset and explore the model’s ability to assist the radiologists. Methods A total of 14,043 US images from 5012 women were prospectively collected from 32 hospitals. To develop the DL model, the patients from 30 hospitals were randomly divided into a training cohort (n = 4149) and an internal test cohort (n = 466). The remaining 2 hospitals (n = 397) were used as the external test cohorts (ETC). We compared the model with the prospective Breast Imaging Reporting and Data System assessment and five radiologists. We also explored the model’s ability to assist the radiologists using two different methods. Results The model demonstrated excellent diagnostic performance with the ETC, with a high area under the receiver operating characteristic curve (AUC, 0.913), sensitivity (88.84%), specificity (83.77%), and accuracy (86.40%). In the comparison set, the AUC was similar to that of the expert (p = 0.5629) and one experienced radiologist (p = 0.2112) and significantly higher than that of three inexperienced radiologists (p < 0.01). After model assistance, the accuracies and specificities of the radiologists were substantially improved without loss in sensitivities. Conclusions The DL model yielded satisfactory predictions in distinguishing benign from malignant breast lesions. The model showed the potential value in improving the diagnosis of breast lesions by radiologists.https://doi.org/10.1186/s13244-022-01259-8Deep learningUltrasonographyBreast neoplasmsDiagnosisArtificial intelligence |
| spellingShingle | Yang Gu Wen Xu Bin Lin Xing An Jiawei Tian Haitao Ran Weidong Ren Cai Chang Jianjun Yuan Chunsong Kang Youbin Deng Hui Wang Baoming Luo Shenglan Guo Qi Zhou Ensheng Xue Weiwei Zhan Qing Zhou Jie Li Ping Zhou Man Chen Ying Gu Wu Chen Yuhong Zhang Jianchu Li Longfei Cong Lei Zhu Hongyan Wang Yuxin Jiang Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study Insights into Imaging Deep learning Ultrasonography Breast neoplasms Diagnosis Artificial intelligence |
| title | Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study |
| title_full | Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study |
| title_fullStr | Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study |
| title_full_unstemmed | Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study |
| title_short | Deep learning based on ultrasound images assists breast lesion diagnosis in China: a multicenter diagnostic study |
| title_sort | deep learning based on ultrasound images assists breast lesion diagnosis in china a multicenter diagnostic study |
| topic | Deep learning Ultrasonography Breast neoplasms Diagnosis Artificial intelligence |
| url | https://doi.org/10.1186/s13244-022-01259-8 |
| work_keys_str_mv | AT yanggu deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT wenxu deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT binlin deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT xingan deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT jiaweitian deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT haitaoran deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT weidongren deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT caichang deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT jianjunyuan deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT chunsongkang deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT youbindeng deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT huiwang deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT baomingluo deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT shenglanguo deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT qizhou deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT enshengxue deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT weiweizhan deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT qingzhou deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT jieli deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT pingzhou deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT manchen deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT yinggu deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT wuchen deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT yuhongzhang deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT jianchuli deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT longfeicong deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT leizhu deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT hongyanwang deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy AT yuxinjiang deeplearningbasedonultrasoundimagesassistsbreastlesiondiagnosisinchinaamulticenterdiagnosticstudy |