Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention Mechanism
<b>Background:</b> The accurate classification between malignant and benign breast lesions detected on mammograms is a crucial but difficult challenge for reducing false-positive recall rates and improving the efficacy of breast cancer screening. <b>Objective:</b> This study...
| Những tác giả chính: | , , , , , |
|---|---|
| Định dạng: | Bài viết |
| Ngôn ngữ: | English |
| Được phát hành: |
MDPI AG
2022-09-01
|
| Loạt: | Tomography |
| Những chủ đề: | |
| Truy cập trực tuyến: | https://www.mdpi.com/2379-139X/8/5/200 |
| _version_ | 1827647916664160256 |
|---|---|
| author | Warid Islam Meredith Jones Rowzat Faiz Negar Sadeghipour Yuchen Qiu Bin Zheng |
| author_facet | Warid Islam Meredith Jones Rowzat Faiz Negar Sadeghipour Yuchen Qiu Bin Zheng |
| author_sort | Warid Islam |
| collection | DOAJ |
| description | <b>Background:</b> The accurate classification between malignant and benign breast lesions detected on mammograms is a crucial but difficult challenge for reducing false-positive recall rates and improving the efficacy of breast cancer screening. <b>Objective:</b> This study aims to optimize a new deep transfer learning model by implementing a novel attention mechanism in order to improve the accuracy of breast lesion classification. <b>Methods:</b> ResNet50 is selected as the base model to develop a new deep transfer learning model. To enhance the accuracy of breast lesion classification, we propose adding a convolutional block attention module (CBAM) to the standard ResNet50 model and optimizing a new model for this task. We assembled a large dataset with 4280 mammograms depicting suspicious soft-tissue mass-type lesions. A region of interest (ROI) is extracted from each image based on lesion center. Among them, 2480 and 1800 ROIs depict verified benign and malignant lesions, respectively. The image dataset is randomly split into two subsets with a ratio of 9:1 five times to train and test two ResNet50 models with and without using CBAM. <b>Results:</b> Using the area under ROC curve (AUC) as an evaluation index, the new CBAM-based ResNet50 model yields AUC = 0.866 ± 0.015, which is significantly higher than that obtained by the standard ResNet50 model (AUC = 0.772 ± 0.008) (<i>p</i> < 0.01). <b>Conclusion:</b> This study demonstrates that although deep transfer learning technology attracted broad research interest in medical-imaging informatic fields, adding a new attention mechanism to optimize deep transfer learning models for specific application tasks can play an important role in further improving model performances. |
| first_indexed | 2024-03-09T19:26:13Z |
| format | Article |
| id | doaj.art-c6d81fbe5cc44e08940f512b20b51c44 |
| institution | Directory Open Access Journal |
| issn | 2379-1381 2379-139X |
| language | English |
| last_indexed | 2024-03-09T19:26:13Z |
| publishDate | 2022-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Tomography |
| spelling | doaj.art-c6d81fbe5cc44e08940f512b20b51c442023-11-24T02:56:29ZengMDPI AGTomography2379-13812379-139X2022-09-01852411242510.3390/tomography8050200Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention MechanismWarid Islam0Meredith Jones1Rowzat Faiz2Negar Sadeghipour3Yuchen Qiu4Bin Zheng5School of Electrical & Computer Engineering, University of Oklahoma, Norman, OK 73019, USAStephenson School of Biomedical Engineering, University of Oklahoma, Norman, OK 73019, USASchool of Electrical & Computer Engineering, University of Oklahoma, Norman, OK 73019, USASchool of Electrical & Computer Engineering, University of Oklahoma, Norman, OK 73019, USASchool of Electrical & Computer Engineering, University of Oklahoma, Norman, OK 73019, USASchool of Electrical & Computer Engineering, University of Oklahoma, Norman, OK 73019, USA<b>Background:</b> The accurate classification between malignant and benign breast lesions detected on mammograms is a crucial but difficult challenge for reducing false-positive recall rates and improving the efficacy of breast cancer screening. <b>Objective:</b> This study aims to optimize a new deep transfer learning model by implementing a novel attention mechanism in order to improve the accuracy of breast lesion classification. <b>Methods:</b> ResNet50 is selected as the base model to develop a new deep transfer learning model. To enhance the accuracy of breast lesion classification, we propose adding a convolutional block attention module (CBAM) to the standard ResNet50 model and optimizing a new model for this task. We assembled a large dataset with 4280 mammograms depicting suspicious soft-tissue mass-type lesions. A region of interest (ROI) is extracted from each image based on lesion center. Among them, 2480 and 1800 ROIs depict verified benign and malignant lesions, respectively. The image dataset is randomly split into two subsets with a ratio of 9:1 five times to train and test two ResNet50 models with and without using CBAM. <b>Results:</b> Using the area under ROC curve (AUC) as an evaluation index, the new CBAM-based ResNet50 model yields AUC = 0.866 ± 0.015, which is significantly higher than that obtained by the standard ResNet50 model (AUC = 0.772 ± 0.008) (<i>p</i> < 0.01). <b>Conclusion:</b> This study demonstrates that although deep transfer learning technology attracted broad research interest in medical-imaging informatic fields, adding a new attention mechanism to optimize deep transfer learning models for specific application tasks can play an important role in further improving model performances.https://www.mdpi.com/2379-139X/8/5/200computer-aided diagnosis (CAD) scheme of mammogramsconvolutional block attention module (CBAM)residual network (ResNet)breast lesion classification |
| spellingShingle | Warid Islam Meredith Jones Rowzat Faiz Negar Sadeghipour Yuchen Qiu Bin Zheng Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention Mechanism Tomography computer-aided diagnosis (CAD) scheme of mammograms convolutional block attention module (CBAM) residual network (ResNet) breast lesion classification |
| title | Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention Mechanism |
| title_full | Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention Mechanism |
| title_fullStr | Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention Mechanism |
| title_full_unstemmed | Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention Mechanism |
| title_short | Improving Performance of Breast Lesion Classification Using a ResNet50 Model Optimized with a Novel Attention Mechanism |
| title_sort | improving performance of breast lesion classification using a resnet50 model optimized with a novel attention mechanism |
| topic | computer-aided diagnosis (CAD) scheme of mammograms convolutional block attention module (CBAM) residual network (ResNet) breast lesion classification |
| url | https://www.mdpi.com/2379-139X/8/5/200 |
| work_keys_str_mv | AT waridislam improvingperformanceofbreastlesionclassificationusingaresnet50modeloptimizedwithanovelattentionmechanism AT meredithjones improvingperformanceofbreastlesionclassificationusingaresnet50modeloptimizedwithanovelattentionmechanism AT rowzatfaiz improvingperformanceofbreastlesionclassificationusingaresnet50modeloptimizedwithanovelattentionmechanism AT negarsadeghipour improvingperformanceofbreastlesionclassificationusingaresnet50modeloptimizedwithanovelattentionmechanism AT yuchenqiu improvingperformanceofbreastlesionclassificationusingaresnet50modeloptimizedwithanovelattentionmechanism AT binzheng improvingperformanceofbreastlesionclassificationusingaresnet50modeloptimizedwithanovelattentionmechanism |