Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules
Abstract The large variations of polyp sizes and shapes and the close resemblances of polyps to their surroundings call for features with long‐range information in rich scales and strong discrimination. This article proposes two parallel structured modules for building those features. One is the Tra...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2023-06-01
|
| Series: | IET Image Processing |
| Subjects: | |
| Online Access: | https://doi.org/10.1049/ipr2.12813 |
| _version_ | 1827935640109449216 |
|---|---|
| author | Qingqing Guo Xianyong Fang Kaibing Wang Yuqing Shi Linbo Wang Enming Zhang Zhengyi Liu |
| author_facet | Qingqing Guo Xianyong Fang Kaibing Wang Yuqing Shi Linbo Wang Enming Zhang Zhengyi Liu |
| author_sort | Qingqing Guo |
| collection | DOAJ |
| description | Abstract The large variations of polyp sizes and shapes and the close resemblances of polyps to their surroundings call for features with long‐range information in rich scales and strong discrimination. This article proposes two parallel structured modules for building those features. One is the Transformer Inception module (TI) which applies Transformers with different reception fields in parallel to input features and thus enriches them with more long‐range information in more scales. The other is the Local‐Detail Augmentation module (LDA) which applies the spatial and channel attentions in parallel to each block and thus locally augments the features from two complementary dimensions for more object details. Integrating TI and LDA, a new Transformer encoder based framework, Parallel‐Enhanced Network (PENet), is proposed, where LDA is specifically adopted twice in a coarse‐to‐fine way for accurate prediction. PENet is efficient in segmenting polyps with different sizes and shapes without the interference from the background tissues. Experimental comparisons with state‐of‐the‐arts methods show its merits. |
| first_indexed | 2024-03-13T07:57:46Z |
| format | Article |
| id | doaj.art-1509ab8bb19f4a0ba658026e3a7be9c9 |
| institution | Directory Open Access Journal |
| issn | 1751-9659 1751-9667 |
| language | English |
| last_indexed | 2024-03-13T07:57:46Z |
| publishDate | 2023-06-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Image Processing |
| spelling | doaj.art-1509ab8bb19f4a0ba658026e3a7be9c92023-06-02T03:06:38ZengWileyIET Image Processing1751-96591751-96672023-06-011782503251510.1049/ipr2.12813Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modulesQingqing Guo0Xianyong Fang1Kaibing Wang2Yuqing Shi3Linbo Wang4Enming Zhang5Zhengyi Liu6School of Computer Science and Technology Anhui University HefeiChinaSchool of Computer Science and Technology Anhui University HefeiChinaSchool of Computer Science and Technology Anhui University HefeiChinaSchool of Computer Science and Technology Anhui University HefeiChinaSchool of Computer Science and Technology Anhui University HefeiChinaIslet Pathophysiology, Department of Clinical Science Lund University Diabetes Centre Malmö SwedenSchool of Computer Science and Technology Anhui University HefeiChinaAbstract The large variations of polyp sizes and shapes and the close resemblances of polyps to their surroundings call for features with long‐range information in rich scales and strong discrimination. This article proposes two parallel structured modules for building those features. One is the Transformer Inception module (TI) which applies Transformers with different reception fields in parallel to input features and thus enriches them with more long‐range information in more scales. The other is the Local‐Detail Augmentation module (LDA) which applies the spatial and channel attentions in parallel to each block and thus locally augments the features from two complementary dimensions for more object details. Integrating TI and LDA, a new Transformer encoder based framework, Parallel‐Enhanced Network (PENet), is proposed, where LDA is specifically adopted twice in a coarse‐to‐fine way for accurate prediction. PENet is efficient in segmenting polyps with different sizes and shapes without the interference from the background tissues. Experimental comparisons with state‐of‐the‐arts methods show its merits.https://doi.org/10.1049/ipr2.12813biomedical imagingcomputer visionimage segmentation |
| spellingShingle | Qingqing Guo Xianyong Fang Kaibing Wang Yuqing Shi Linbo Wang Enming Zhang Zhengyi Liu Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules IET Image Processing biomedical imaging computer vision image segmentation |
| title | Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules |
| title_full | Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules |
| title_fullStr | Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules |
| title_full_unstemmed | Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules |
| title_short | Parallel matters: Efficient polyp segmentation with parallel structured feature augmentation modules |
| title_sort | parallel matters efficient polyp segmentation with parallel structured feature augmentation modules |
| topic | biomedical imaging computer vision image segmentation |
| url | https://doi.org/10.1049/ipr2.12813 |
| work_keys_str_mv | AT qingqingguo parallelmattersefficientpolypsegmentationwithparallelstructuredfeatureaugmentationmodules AT xianyongfang parallelmattersefficientpolypsegmentationwithparallelstructuredfeatureaugmentationmodules AT kaibingwang parallelmattersefficientpolypsegmentationwithparallelstructuredfeatureaugmentationmodules AT yuqingshi parallelmattersefficientpolypsegmentationwithparallelstructuredfeatureaugmentationmodules AT linbowang parallelmattersefficientpolypsegmentationwithparallelstructuredfeatureaugmentationmodules AT enmingzhang parallelmattersefficientpolypsegmentationwithparallelstructuredfeatureaugmentationmodules AT zhengyiliu parallelmattersefficientpolypsegmentationwithparallelstructuredfeatureaugmentationmodules |