Rethinking the Transfer Learning for FCN Based Polyp Segmentation in Colonoscopy
Besides the complex nature of colonoscopy frames with intrinsic frame formation artefacts such as light reflections and the diversity of polyp types/shapes, the publicly available polyp segmentation training datasets are limited, small and imbalanced. In this case, the automated polyp segmentation u...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10044660/ |
| _version_ | 1797900226840756224 |
|---|---|
| author | Yan Wen Lei Zhang Xiangli Meng Xujiong Ye |
| author_facet | Yan Wen Lei Zhang Xiangli Meng Xujiong Ye |
| author_sort | Yan Wen |
| collection | DOAJ |
| description | Besides the complex nature of colonoscopy frames with intrinsic frame formation artefacts such as light reflections and the diversity of polyp types/shapes, the publicly available polyp segmentation training datasets are limited, small and imbalanced. In this case, the automated polyp segmentation using a deep neural network remains an open challenge due to the overfitting of training on small datasets. We proposed a simple yet effective polyp segmentation pipeline that couples the segmentation (FCN) and classification (CNN) tasks. We find the effectiveness of interactive weight transfer between dense and coarse vision tasks that mitigates the overfitting in learning. This motivates us to design a new training scheme within our segmentation pipeline. Our method is evaluated on CVC-EndoSceneStill and Kvasir-SEG datasets. It achieves 4.34% and 5.70% Polyp-IoU improvements compared to the state-of-the-art methods on the EndoSceneStill and Kvasir-SEG datasets, respectively and achieves real-time performance in inference. The model and code are available at <uri>https://github.com/MELSunny/Keras-FCN</uri> |
| first_indexed | 2024-04-10T08:42:36Z |
| format | Article |
| id | doaj.art-742bc8590c3a461f82f7370a282ec160 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-04-10T08:42:36Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-742bc8590c3a461f82f7370a282ec1602023-02-23T00:00:30ZengIEEEIEEE Access2169-35362023-01-0111161831619310.1109/ACCESS.2023.324551910044660Rethinking the Transfer Learning for FCN Based Polyp Segmentation in ColonoscopyYan Wen0Lei Zhang1https://orcid.org/0000-0002-9236-6004Xiangli Meng2Xujiong Ye3https://orcid.org/0000-0003-0115-0724Laboratory of Vision Engineering (LoVE), School of Computer Science, University of Lincoln, Lincoln, U.KLaboratory of Vision Engineering (LoVE), School of Computer Science, University of Lincoln, Lincoln, U.KSchool of Electronic and Electrical Engineering, Lingnan Normal University, Zhanjiang, ChinaLaboratory of Vision Engineering (LoVE), School of Computer Science, University of Lincoln, Lincoln, U.KBesides the complex nature of colonoscopy frames with intrinsic frame formation artefacts such as light reflections and the diversity of polyp types/shapes, the publicly available polyp segmentation training datasets are limited, small and imbalanced. In this case, the automated polyp segmentation using a deep neural network remains an open challenge due to the overfitting of training on small datasets. We proposed a simple yet effective polyp segmentation pipeline that couples the segmentation (FCN) and classification (CNN) tasks. We find the effectiveness of interactive weight transfer between dense and coarse vision tasks that mitigates the overfitting in learning. This motivates us to design a new training scheme within our segmentation pipeline. Our method is evaluated on CVC-EndoSceneStill and Kvasir-SEG datasets. It achieves 4.34% and 5.70% Polyp-IoU improvements compared to the state-of-the-art methods on the EndoSceneStill and Kvasir-SEG datasets, respectively and achieves real-time performance in inference. The model and code are available at <uri>https://github.com/MELSunny/Keras-FCN</uri>https://ieeexplore.ieee.org/document/10044660/Colonoscopyreal-time polyp segmentationtransfer learningconvolutional neural network |
| spellingShingle | Yan Wen Lei Zhang Xiangli Meng Xujiong Ye Rethinking the Transfer Learning for FCN Based Polyp Segmentation in Colonoscopy IEEE Access Colonoscopy real-time polyp segmentation transfer learning convolutional neural network |
| title | Rethinking the Transfer Learning for FCN Based Polyp Segmentation in Colonoscopy |
| title_full | Rethinking the Transfer Learning for FCN Based Polyp Segmentation in Colonoscopy |
| title_fullStr | Rethinking the Transfer Learning for FCN Based Polyp Segmentation in Colonoscopy |
| title_full_unstemmed | Rethinking the Transfer Learning for FCN Based Polyp Segmentation in Colonoscopy |
| title_short | Rethinking the Transfer Learning for FCN Based Polyp Segmentation in Colonoscopy |
| title_sort | rethinking the transfer learning for fcn based polyp segmentation in colonoscopy |
| topic | Colonoscopy real-time polyp segmentation transfer learning convolutional neural network |
| url | https://ieeexplore.ieee.org/document/10044660/ |
| work_keys_str_mv | AT yanwen rethinkingthetransferlearningforfcnbasedpolypsegmentationincolonoscopy AT leizhang rethinkingthetransferlearningforfcnbasedpolypsegmentationincolonoscopy AT xianglimeng rethinkingthetransferlearningforfcnbasedpolypsegmentationincolonoscopy AT xujiongye rethinkingthetransferlearningforfcnbasedpolypsegmentationincolonoscopy |