Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network
Abstract Background The classification of benign and malignant microcalcification clusters (MCs) is an important task for computer-aided diagnosis (CAD) of digital breast tomosynthesis (DBT) images. Influenced by imaging method, DBT has the characteristic of anisotropic resolution, in which the reso...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2021-07-01
|
| Series: | BioMedical Engineering OnLine |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12938-021-00908-1 |
| _version_ | 1818368057148440576 |
|---|---|
| author | Bingbing Xiao Haotian Sun You Meng Yunsong Peng Xiaodong Yang Shuangqing Chen Zhuangzhi Yan Jian Zheng |
| author_facet | Bingbing Xiao Haotian Sun You Meng Yunsong Peng Xiaodong Yang Shuangqing Chen Zhuangzhi Yan Jian Zheng |
| author_sort | Bingbing Xiao |
| collection | DOAJ |
| description | Abstract Background The classification of benign and malignant microcalcification clusters (MCs) is an important task for computer-aided diagnosis (CAD) of digital breast tomosynthesis (DBT) images. Influenced by imaging method, DBT has the characteristic of anisotropic resolution, in which the resolution of intra-slice and inter-slice is quite different. In addition, the sharpness of MCs in different slices of DBT is quite different, among which the clearest slice is called focus slice. These characteristics limit the performance of CAD algorithms based on standard 3D convolution neural network (CNN). Methods To make full use of the characteristics of the DBT, we proposed a new ensemble CNN, which consists of the 2D ResNet34 and the anisotropic 3D ResNet to extract the 2D focus slice features and 3D contextual features of MCs, respectively. Moreover, the anisotropic 3D convolution is used to build 3D ResNet to avoid the influence of DBT anisotropy. Results The proposed method was evaluated on 495 MCs in DBT images of 275 patients, which are collected from our collaborative hospital. The area under the curve (AUC) of receiver operating characteristic (ROC) and accuracy of classifying benign and malignant MCs using decision-level ensemble strategy were 0.8837 and 82.00%, which were significantly higher than the experimental results of 2D ResNet34 (AUC: 0.8264, ACC: 76.00%) and anisotropic 3D ResNet (AUC: 0.8455, ACC: 76.00%). Compared with the results of 3D features classification in the radiomics, the AUC of the deep learning method with decision-level ensemble strategy was improved by 0.0435, and the F1 score was improved from 79.37 to 85.71%. More importantly, the sensitivity increased from 78.13 to 84.38%, and the specificity increased from 66.67 to 77.78%, which effectively reduced the false positives of diagnosis Conclusion The results fully prove that the ensemble CNN can effectively integrate 2D features and 3D features, improve the classification performance of benign and malignant MCs in DBT, and reduce the false positives. |
| first_indexed | 2024-12-13T23:01:53Z |
| format | Article |
| id | doaj.art-6807f91ee8a64c38928ecf364ae22274 |
| institution | Directory Open Access Journal |
| issn | 1475-925X |
| language | English |
| last_indexed | 2024-12-13T23:01:53Z |
| publishDate | 2021-07-01 |
| publisher | BMC |
| record_format | Article |
| series | BioMedical Engineering OnLine |
| spelling | doaj.art-6807f91ee8a64c38928ecf364ae222742022-12-21T23:28:22ZengBMCBioMedical Engineering OnLine1475-925X2021-07-0120112010.1186/s12938-021-00908-1Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural networkBingbing Xiao0Haotian Sun1You Meng2Yunsong Peng3Xiaodong Yang4Shuangqing Chen5Zhuangzhi Yan6Jian Zheng7Institute of Biomedical Engineering, School of Communication and Information Engineering, Shanghai UniversityUniversity of Science and Technology of ChinaDepartment of Breast Surgery, The Affiliated Suzhou Hospital of Nanjing Medical UniversityUniversity of Science and Technology of ChinaDepartment of Medical Imaging, Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of SciencesGusu School, Nanjing Medical UniversityInstitute of Biomedical Engineering, School of Communication and Information Engineering, Shanghai UniversityDepartment of Medical Imaging, Suzhou Institute of Biomedical Engineering and Technology, Chinese Academy of SciencesAbstract Background The classification of benign and malignant microcalcification clusters (MCs) is an important task for computer-aided diagnosis (CAD) of digital breast tomosynthesis (DBT) images. Influenced by imaging method, DBT has the characteristic of anisotropic resolution, in which the resolution of intra-slice and inter-slice is quite different. In addition, the sharpness of MCs in different slices of DBT is quite different, among which the clearest slice is called focus slice. These characteristics limit the performance of CAD algorithms based on standard 3D convolution neural network (CNN). Methods To make full use of the characteristics of the DBT, we proposed a new ensemble CNN, which consists of the 2D ResNet34 and the anisotropic 3D ResNet to extract the 2D focus slice features and 3D contextual features of MCs, respectively. Moreover, the anisotropic 3D convolution is used to build 3D ResNet to avoid the influence of DBT anisotropy. Results The proposed method was evaluated on 495 MCs in DBT images of 275 patients, which are collected from our collaborative hospital. The area under the curve (AUC) of receiver operating characteristic (ROC) and accuracy of classifying benign and malignant MCs using decision-level ensemble strategy were 0.8837 and 82.00%, which were significantly higher than the experimental results of 2D ResNet34 (AUC: 0.8264, ACC: 76.00%) and anisotropic 3D ResNet (AUC: 0.8455, ACC: 76.00%). Compared with the results of 3D features classification in the radiomics, the AUC of the deep learning method with decision-level ensemble strategy was improved by 0.0435, and the F1 score was improved from 79.37 to 85.71%. More importantly, the sensitivity increased from 78.13 to 84.38%, and the specificity increased from 66.67 to 77.78%, which effectively reduced the false positives of diagnosis Conclusion The results fully prove that the ensemble CNN can effectively integrate 2D features and 3D features, improve the classification performance of benign and malignant MCs in DBT, and reduce the false positives.https://doi.org/10.1186/s12938-021-00908-1Microcalcification clusterDigital breast tomosynthesisConvolution neural networkEnsemble learningClassification |
| spellingShingle | Bingbing Xiao Haotian Sun You Meng Yunsong Peng Xiaodong Yang Shuangqing Chen Zhuangzhi Yan Jian Zheng Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network BioMedical Engineering OnLine Microcalcification cluster Digital breast tomosynthesis Convolution neural network Ensemble learning Classification |
| title | Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network |
| title_full | Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network |
| title_fullStr | Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network |
| title_full_unstemmed | Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network |
| title_short | Classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network |
| title_sort | classification of microcalcification clusters in digital breast tomosynthesis using ensemble convolutional neural network |
| topic | Microcalcification cluster Digital breast tomosynthesis Convolution neural network Ensemble learning Classification |
| url | https://doi.org/10.1186/s12938-021-00908-1 |
| work_keys_str_mv | AT bingbingxiao classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork AT haotiansun classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork AT youmeng classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork AT yunsongpeng classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork AT xiaodongyang classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork AT shuangqingchen classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork AT zhuangzhiyan classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork AT jianzheng classificationofmicrocalcificationclustersindigitalbreasttomosynthesisusingensembleconvolutionalneuralnetwork |