DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentation

Abstract Sputum smear tests are critical for the diagnosis of respiratory diseases. Automatic segmentation of bacteria from sputum smear images is important for improving diagnostic efficiency. However, this remains a challenging task owing to the high interclass similarity among different categorie...

Full description

Bibliographic Details
Main Authors: Jingkun Wang, Xinyu Ma, Long Cao, Yilin Leng, Zeyi Li, Zihan Cheng, Yuzhu Cao, Xiaoping Huang, Jian Zheng
Format: Article
Language:English
Published: SpringerOpen 2023-07-01
Series:Visual Computing for Industry, Biomedicine, and Art
Subjects:
Online Access:https://doi.org/10.1186/s42492-023-00141-8
_version_ 1797784746149806080
author Jingkun Wang
Xinyu Ma
Long Cao
Yilin Leng
Zeyi Li
Zihan Cheng
Yuzhu Cao
Xiaoping Huang
Jian Zheng
author_facet Jingkun Wang
Xinyu Ma
Long Cao
Yilin Leng
Zeyi Li
Zihan Cheng
Yuzhu Cao
Xiaoping Huang
Jian Zheng
author_sort Jingkun Wang
collection DOAJ
description Abstract Sputum smear tests are critical for the diagnosis of respiratory diseases. Automatic segmentation of bacteria from sputum smear images is important for improving diagnostic efficiency. However, this remains a challenging task owing to the high interclass similarity among different categories of bacteria and the low contrast of the bacterial edges. To explore more levels of global pattern features to promote the distinguishing ability of bacterial categories and maintain sufficient local fine-grained features to ensure accurate localization of ambiguous bacteria simultaneously, we propose a novel dual-branch deformable cross-attention fusion network (DB-DCAFN) for accurate bacterial segmentation. Specifically, we first designed a dual-branch encoder consisting of multiple convolution and transformer blocks in parallel to simultaneously extract multilevel local and global features. We then designed a sparse and deformable cross-attention module to capture the semantic dependencies between local and global features, which can bridge the semantic gap and fuse features effectively. Furthermore, we designed a feature assignment fusion module to enhance meaningful features using an adaptive feature weighting strategy to obtain more accurate segmentation. We conducted extensive experiments to evaluate the effectiveness of DB-DCAFN on a clinical dataset comprising three bacterial categories: Acinetobacter baumannii, Klebsiella pneumoniae, and Pseudomonas aeruginosa. The experimental results demonstrate that the proposed DB-DCAFN outperforms other state-of-the-art methods and is effective at segmenting bacteria from sputum smear images.
first_indexed 2024-03-13T00:44:17Z
format Article
id doaj.art-2d90b9714b7642969d0eaf9ce03edaef
institution Directory Open Access Journal
issn 2524-4442
language English
last_indexed 2024-03-13T00:44:17Z
publishDate 2023-07-01
publisher SpringerOpen
record_format Article
series Visual Computing for Industry, Biomedicine, and Art
spelling doaj.art-2d90b9714b7642969d0eaf9ce03edaef2023-07-09T11:05:22ZengSpringerOpenVisual Computing for Industry, Biomedicine, and Art2524-44422023-07-016111610.1186/s42492-023-00141-8DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentationJingkun Wang0Xinyu Ma1Long Cao2Yilin Leng3Zeyi Li4Zihan Cheng5Yuzhu Cao6Xiaoping Huang7Jian Zheng8School of Biomedical Engineering (Suzhou), Division of Life Sciences and Medicine, University of Science and Technology of ChinaSchool of Biomedical Engineering (Suzhou), Division of Life Sciences and Medicine, University of Science and Technology of ChinaDepartment of Infectious Diseases, the First Affiliated Hospital of Soochow UniversityInstitute of Biomedical Engineering, School of Communication and Information Engineering, Shanghai UniversityCollege of Computer and Information, Hohai UniversitySchool of Electronic and Information Engineering, Changchun University of Science and TechnologySchool of Biomedical Engineering (Suzhou), Division of Life Sciences and Medicine, University of Science and Technology of ChinaDepartment of Infectious Diseases, the First Affiliated Hospital of Soochow UniversitySchool of Biomedical Engineering (Suzhou), Division of Life Sciences and Medicine, University of Science and Technology of ChinaAbstract Sputum smear tests are critical for the diagnosis of respiratory diseases. Automatic segmentation of bacteria from sputum smear images is important for improving diagnostic efficiency. However, this remains a challenging task owing to the high interclass similarity among different categories of bacteria and the low contrast of the bacterial edges. To explore more levels of global pattern features to promote the distinguishing ability of bacterial categories and maintain sufficient local fine-grained features to ensure accurate localization of ambiguous bacteria simultaneously, we propose a novel dual-branch deformable cross-attention fusion network (DB-DCAFN) for accurate bacterial segmentation. Specifically, we first designed a dual-branch encoder consisting of multiple convolution and transformer blocks in parallel to simultaneously extract multilevel local and global features. We then designed a sparse and deformable cross-attention module to capture the semantic dependencies between local and global features, which can bridge the semantic gap and fuse features effectively. Furthermore, we designed a feature assignment fusion module to enhance meaningful features using an adaptive feature weighting strategy to obtain more accurate segmentation. We conducted extensive experiments to evaluate the effectiveness of DB-DCAFN on a clinical dataset comprising three bacterial categories: Acinetobacter baumannii, Klebsiella pneumoniae, and Pseudomonas aeruginosa. The experimental results demonstrate that the proposed DB-DCAFN outperforms other state-of-the-art methods and is effective at segmenting bacteria from sputum smear images.https://doi.org/10.1186/s42492-023-00141-8Bacterial segmentationDual-branch parallel encoderDeformable cross-attention moduleFeature assignment fusion module
spellingShingle Jingkun Wang
Xinyu Ma
Long Cao
Yilin Leng
Zeyi Li
Zihan Cheng
Yuzhu Cao
Xiaoping Huang
Jian Zheng
DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentation
Visual Computing for Industry, Biomedicine, and Art
Bacterial segmentation
Dual-branch parallel encoder
Deformable cross-attention module
Feature assignment fusion module
title DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentation
title_full DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentation
title_fullStr DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentation
title_full_unstemmed DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentation
title_short DB-DCAFN: dual-branch deformable cross-attention fusion network for bacterial segmentation
title_sort db dcafn dual branch deformable cross attention fusion network for bacterial segmentation
topic Bacterial segmentation
Dual-branch parallel encoder
Deformable cross-attention module
Feature assignment fusion module
url https://doi.org/10.1186/s42492-023-00141-8
work_keys_str_mv AT jingkunwang dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT xinyuma dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT longcao dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT yilinleng dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT zeyili dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT zihancheng dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT yuzhucao dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT xiaopinghuang dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation
AT jianzheng dbdcafndualbranchdeformablecrossattentionfusionnetworkforbacterialsegmentation