A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesis
Abstract Deep learning has shown tremendous potential in the task of object detection in images. However, a common challenge with this task is when only a limited number of images containing the object of interest are available. This is a particular issue in cancer screening, such as digital breast...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2021-05-01
|
| Series: | Scientific Reports |
| Online Access: | https://doi.org/10.1038/s41598-021-89626-1 |
| _version_ | 1818858465060192256 |
|---|---|
| author | Albert Swiecicki Nicholas Konz Mateusz Buda Maciej A. Mazurowski |
| author_facet | Albert Swiecicki Nicholas Konz Mateusz Buda Maciej A. Mazurowski |
| author_sort | Albert Swiecicki |
| collection | DOAJ |
| description | Abstract Deep learning has shown tremendous potential in the task of object detection in images. However, a common challenge with this task is when only a limited number of images containing the object of interest are available. This is a particular issue in cancer screening, such as digital breast tomosynthesis (DBT), where less than 1% of cases contain cancer. In this study, we propose a method to train an inpainting generative adversarial network to be used for cancer detection using only images that do not contain cancer. During inference, we removed a part of the image and used the network to complete the removed part. A significant error in completing an image part was considered an indication that such location is unexpected and thus abnormal. A large dataset of DBT images used in this study was collected at Duke University. It consisted of 19,230 reconstructed volumes from 4348 patients. Cancerous masses and architectural distortions were marked with bounding boxes by radiologists. Our experiments showed that the locations containing cancer were associated with a notably higher completion error than the non-cancer locations (mean error ratio of 2.77). All data used in this study has been made publicly available by the authors. |
| first_indexed | 2024-12-19T08:56:43Z |
| format | Article |
| id | doaj.art-508eb076572e45cab8b174526afb95f2 |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-12-19T08:56:43Z |
| publishDate | 2021-05-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-508eb076572e45cab8b174526afb95f22022-12-21T20:28:35ZengNature PortfolioScientific Reports2045-23222021-05-0111111310.1038/s41598-021-89626-1A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesisAlbert Swiecicki0Nicholas Konz1Mateusz Buda2Maciej A. Mazurowski3Department of Electrical and Computer Engineering, Duke UniversityDepartment of Electrical and Computer Engineering, Duke UniversityDepartment of Radiology, Duke UniversityDepartment of Electrical and Computer Engineering, Duke UniversityAbstract Deep learning has shown tremendous potential in the task of object detection in images. However, a common challenge with this task is when only a limited number of images containing the object of interest are available. This is a particular issue in cancer screening, such as digital breast tomosynthesis (DBT), where less than 1% of cases contain cancer. In this study, we propose a method to train an inpainting generative adversarial network to be used for cancer detection using only images that do not contain cancer. During inference, we removed a part of the image and used the network to complete the removed part. A significant error in completing an image part was considered an indication that such location is unexpected and thus abnormal. A large dataset of DBT images used in this study was collected at Duke University. It consisted of 19,230 reconstructed volumes from 4348 patients. Cancerous masses and architectural distortions were marked with bounding boxes by radiologists. Our experiments showed that the locations containing cancer were associated with a notably higher completion error than the non-cancer locations (mean error ratio of 2.77). All data used in this study has been made publicly available by the authors.https://doi.org/10.1038/s41598-021-89626-1 |
| spellingShingle | Albert Swiecicki Nicholas Konz Mateusz Buda Maciej A. Mazurowski A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesis Scientific Reports |
| title | A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesis |
| title_full | A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesis |
| title_fullStr | A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesis |
| title_full_unstemmed | A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesis |
| title_short | A generative adversarial network-based abnormality detection using only normal images for model training with application to digital breast tomosynthesis |
| title_sort | generative adversarial network based abnormality detection using only normal images for model training with application to digital breast tomosynthesis |
| url | https://doi.org/10.1038/s41598-021-89626-1 |
| work_keys_str_mv | AT albertswiecicki agenerativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis AT nicholaskonz agenerativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis AT mateuszbuda agenerativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis AT maciejamazurowski agenerativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis AT albertswiecicki generativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis AT nicholaskonz generativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis AT mateuszbuda generativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis AT maciejamazurowski generativeadversarialnetworkbasedabnormalitydetectionusingonlynormalimagesformodeltrainingwithapplicationtodigitalbreasttomosynthesis |