Chuqian Lei1, Wei Wei2, Zhenyu Liu3, Qianqian Xiong1, Ciqiu Yang4, Mei Yang4, Liulu Zhang4, Teng Zhu4, Xiaosheng Zhuang5, Chunling Liu6, Zaiyi Liu6, Jie Tian7, Kun Wang8. 1. The Second School of Clinical Medicine, Southern Medical University, Guangzhou, 510515, China; Department of Breast Cancer, Cancer Center, Guangdong Provincial People's Hospital & Guangdong Academy of Medical Sciences, 510080, China. 2. School of Electronics and Information, Xi'an Polytechnic University, Xi'an, 710000, China; Engineering Research Center of Molecular and Neuro Imaging of Ministry of Education, School of Life Science and Technology, Xidian University, Xi'an, 710126, China; CAS Key Laboratory of Molecular Imaging, Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China. 3. CAS Key Laboratory of Molecular Imaging, Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China. 4. The Second School of Clinical Medicine, Southern Medical University, Guangzhou, 510515, China. 5. The Second School of Clinical Medicine, Southern Medical University, Guangzhou, 510515, China; Shantou University Medical College, Shantou, 515041, Guangdong, China. 6. Department of Radiology, Guangdong Provincial People's Hospital, Guangzhou, 510080, China. 7. Engineering Research Center of Molecular and Neuro Imaging of Ministry of Education, School of Life Science and Technology, Xidian University, Xi'an, 710126, China; CAS Key Laboratory of Molecular Imaging, Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China; University of Chinese Academy of Sciences, Beijing, 100190, China. Electronic address: jie.tian@ia.ac.cn. 8. Department of Breast Cancer, Cancer Center, Guangdong Provincial People's Hospital & Guangdong Academy of Medical Sciences, 510080, China; The Second School of Clinical Medicine, Southern Medical University, Guangzhou, 510515, China. Electronic address: gzwangkun@126.com.
Abstract
PURPOSE: We developed and validated a radiomic model based on mammography and assessed its value for predicting the pathological diagnosis of Breast Imaging Reporting and Data System (BI-RADS) category 4 calcifications. MATERIALS AND METHODS: Patients with a total of 212 eligible calcifications were recruited (159 cases in the primary cohort and 53 cases in the validation cohort). In total, 8286 radiomic features were extracted from the craniocaudal (CC) and mediolateral oblique (MLO) images. Machine learning was used to select features and build a radiomic signature. The clinical risk factors were selected from the independent clinical factors through logistic regression analyses. The radiomic nomogram incorporated the radiomic signature and an independent clinical risk factor. The diagnostic performance of the radiomic model and the radiologists' empirical prediction model was evaluated by the area under the receiver operating characteristic curve (AUC). The differences between the various AUCs were compared with DeLong's test. RESULTS: Six radiomic features and the menopausal state were included in the radiomic nomogram, which discriminated benign calcifications from malignant calcifications with an AUC of 0.80 in the validation cohort. The difference between the classification results of the radiomic nomogram and that of radiologists was significant (p < 0.05). Particularly for patients with calcifications that are negative on ultrasounds but can be detected by mammography (MG+/US- calcifications), the identification ability of the radiomic nomogram was very strong. CONCLUSIONS: The mammography-based radiomic nomogram is a potential tool to distinguish benign calcifications from malignant calcifications.
PURPOSE: We developed and validated a radiomic model based on mammography and assessed its value for predicting the pathological diagnosis of Breast Imaging Reporting and Data System (BI-RADS) category 4 calcifications. MATERIALS AND METHODS:Patients with a total of 212 eligible calcifications were recruited (159 cases in the primary cohort and 53 cases in the validation cohort). In total, 8286 radiomic features were extracted from the craniocaudal (CC) and mediolateral oblique (MLO) images. Machine learning was used to select features and build a radiomic signature. The clinical risk factors were selected from the independent clinical factors through logistic regression analyses. The radiomic nomogram incorporated the radiomic signature and an independent clinical risk factor. The diagnostic performance of the radiomic model and the radiologists' empirical prediction model was evaluated by the area under the receiver operating characteristic curve (AUC). The differences between the various AUCs were compared with DeLong's test. RESULTS: Six radiomic features and the menopausal state were included in the radiomic nomogram, which discriminated benign calcifications from malignant calcifications with an AUC of 0.80 in the validation cohort. The difference between the classification results of the radiomic nomogram and that of radiologists was significant (p < 0.05). Particularly for patients with calcifications that are negative on ultrasounds but can be detected by mammography (MG+/US- calcifications), the identification ability of the radiomic nomogram was very strong. CONCLUSIONS: The mammography-based radiomic nomogram is a potential tool to distinguish benign calcifications from malignant calcifications.