URCA: Uncertainty-based region clipping algorithm for semi-supervised medical image segmentation

•Uncertainty-based region clipping algorithm is proposed for semi-supervised medical image segmentation.•More reliable targets can be learned by the model using uncertainty information.•The ablation experiments demonstrate the validity of the proposed model. Training convolutional neural networks ba...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Computer methods and programs in biomedicine 2024-09, Vol.254, p.108278, Article 108278
Hauptverfasser: Qin, Chendong, Wang, Yongxiong, Zhang, Jiapeng
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page 108278
container_title Computer methods and programs in biomedicine
container_volume 254
creator Qin, Chendong
Wang, Yongxiong
Zhang, Jiapeng
description •Uncertainty-based region clipping algorithm is proposed for semi-supervised medical image segmentation.•More reliable targets can be learned by the model using uncertainty information.•The ablation experiments demonstrate the validity of the proposed model. Training convolutional neural networks based on large amount of labeled data has made great progress in the field of image segmentation. However, in medical image segmentation tasks, annotating the data is expensive and time-consuming because pixel-level annotation requires experts in the relevant field. Currently, the combination of consistent regularization and pseudo labeling-based semi-supervised methods has shown good performance in image segmentation. However, in the training process, a portion of low-confidence pseudo labels are generated by the model. And the semi-supervised segmentation method still has the problem of distribution bias between labeled and unlabeled data. The objective of this study is to address the challenges of semi-supervised learning and improve the segmentation accuracy of semi-supervised models on medical images. To address these issues, we propose an Uncertainty-based Region Clipping Algorithm for semi-supervised medical image segmentation, which consists of two main modules. A module is introduced to compute the uncertainty of two sub-networks predictions with diversity using Monte Carlo Dropout, allowing the model to gradually learn from more reliable targets. To retain model diversity, we use different loss functions for different branches and use Non-Maximum Suppression in one of the branches. The other module is proposed to generate new samples by masking the low-confidence pixels in the original image based on uncertainty information. New samples are fed into the model to facilitate the model to generate pseudo labels with high confidence and enlarge the training data distribution. Comprehensive experiments on the combination of two benchmarks ACDC and BraTS2019 show that our proposed model outperforms state-of-the-art methods in terms of Dice, HD95 and ASD. The results reach an average Dice score of 87.86 % and a HD95 score of 4.214 mm on ACDC dataset. For the brain tumor segmentation, the results reach an average Dice score of 84.79 % and a HD score of 10.13 mm. Our proposed method improves the accuracy of semi-supervised medical image segmentation. Extensive experiments on two public medical image datasets including 2D and 3D modalities demonstrate the superior
doi_str_mv 10.1016/j.cmpb.2024.108278
format Article
fullrecord <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_miscellaneous_3068752476</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><els_id>S0169260724002736</els_id><sourcerecordid>3068752476</sourcerecordid><originalsourceid>FETCH-LOGICAL-c237t-ef5ac8763f5aad6df68f972cb5e9fbc3178415bb460c23e489fa9a8d489544383</originalsourceid><addsrcrecordid>eNp9kMtOwzAQRS0EoqXwAyxQlmxSHCd-BLGpKl5SJSRE18ZxJsFVXthJpf49jlJYsprR-N47noPQdYSXEY7Y3W6p6y5bEkwSPxCEixM0jwQnIaeMnqK5F6UhYZjP0IVzO4wxoZSdo1ksBBcxw3P0uX1fr-6DbaPB9so0_SHMlIM8sFCatgl0ZbrONGWgqrK1pv-qg6K1gYPahG7owO7NqK4hN1pVgalVCf61rKHpVe8TLtFZoSoHV8e6QNunx4_1S7h5e35drzahJjHvQyio0oKz2FeVs7xgokg50RmFtMh0HHGRRDTLEoa9ARKRFipVIvcNTZJYxAt0O-V2tv0ewPWyNk5DVakG2sHJGDPBKUn8igUik1Tb1jkLheys_7g9yAjLkazcyZGsHMnKiaw33Rzzh8xf-2f5RekFD5MA_JV7A1Y6bcBzzY0F3cu8Nf_l_wA_a4sn</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3068752476</pqid></control><display><type>article</type><title>URCA: Uncertainty-based region clipping algorithm for semi-supervised medical image segmentation</title><source>MEDLINE</source><source>Access via ScienceDirect (Elsevier)</source><creator>Qin, Chendong ; Wang, Yongxiong ; Zhang, Jiapeng</creator><creatorcontrib>Qin, Chendong ; Wang, Yongxiong ; Zhang, Jiapeng</creatorcontrib><description>•Uncertainty-based region clipping algorithm is proposed for semi-supervised medical image segmentation.•More reliable targets can be learned by the model using uncertainty information.•The ablation experiments demonstrate the validity of the proposed model. Training convolutional neural networks based on large amount of labeled data has made great progress in the field of image segmentation. However, in medical image segmentation tasks, annotating the data is expensive and time-consuming because pixel-level annotation requires experts in the relevant field. Currently, the combination of consistent regularization and pseudo labeling-based semi-supervised methods has shown good performance in image segmentation. However, in the training process, a portion of low-confidence pseudo labels are generated by the model. And the semi-supervised segmentation method still has the problem of distribution bias between labeled and unlabeled data. The objective of this study is to address the challenges of semi-supervised learning and improve the segmentation accuracy of semi-supervised models on medical images. To address these issues, we propose an Uncertainty-based Region Clipping Algorithm for semi-supervised medical image segmentation, which consists of two main modules. A module is introduced to compute the uncertainty of two sub-networks predictions with diversity using Monte Carlo Dropout, allowing the model to gradually learn from more reliable targets. To retain model diversity, we use different loss functions for different branches and use Non-Maximum Suppression in one of the branches. The other module is proposed to generate new samples by masking the low-confidence pixels in the original image based on uncertainty information. New samples are fed into the model to facilitate the model to generate pseudo labels with high confidence and enlarge the training data distribution. Comprehensive experiments on the combination of two benchmarks ACDC and BraTS2019 show that our proposed model outperforms state-of-the-art methods in terms of Dice, HD95 and ASD. The results reach an average Dice score of 87.86 % and a HD95 score of 4.214 mm on ACDC dataset. For the brain tumor segmentation, the results reach an average Dice score of 84.79 % and a HD score of 10.13 mm. Our proposed method improves the accuracy of semi-supervised medical image segmentation. Extensive experiments on two public medical image datasets including 2D and 3D modalities demonstrate the superiority of our model. The code is available at: https://github.com/QuintinDong/URCA.</description><identifier>ISSN: 0169-2607</identifier><identifier>ISSN: 1872-7565</identifier><identifier>EISSN: 1872-7565</identifier><identifier>DOI: 10.1016/j.cmpb.2024.108278</identifier><identifier>PMID: 38878360</identifier><language>eng</language><publisher>Ireland: Elsevier B.V</publisher><subject>Algorithms ; Brain Neoplasms - diagnostic imaging ; Distribution bias ; Humans ; Image Processing, Computer-Assisted - methods ; Magnetic Resonance Imaging - methods ; Medical image segmentation ; Monte Carlo Method ; Neural Networks, Computer ; Non-maximum suppression ; Semi-supervised ; Supervised Machine Learning ; Uncertainty ; Uncertainty-aware</subject><ispartof>Computer methods and programs in biomedicine, 2024-09, Vol.254, p.108278, Article 108278</ispartof><rights>2024 Elsevier B.V.</rights><rights>Copyright © 2024 Elsevier B.V. All rights reserved.</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c237t-ef5ac8763f5aad6df68f972cb5e9fbc3178415bb460c23e489fa9a8d489544383</cites><orcidid>0000-0003-0680-781X ; 0000-0002-3242-0857</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://dx.doi.org/10.1016/j.cmpb.2024.108278$$EHTML$$P50$$Gelsevier$$H</linktohtml><link.rule.ids>315,782,786,3552,27931,27932,46002</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/38878360$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Qin, Chendong</creatorcontrib><creatorcontrib>Wang, Yongxiong</creatorcontrib><creatorcontrib>Zhang, Jiapeng</creatorcontrib><title>URCA: Uncertainty-based region clipping algorithm for semi-supervised medical image segmentation</title><title>Computer methods and programs in biomedicine</title><addtitle>Comput Methods Programs Biomed</addtitle><description>•Uncertainty-based region clipping algorithm is proposed for semi-supervised medical image segmentation.•More reliable targets can be learned by the model using uncertainty information.•The ablation experiments demonstrate the validity of the proposed model. Training convolutional neural networks based on large amount of labeled data has made great progress in the field of image segmentation. However, in medical image segmentation tasks, annotating the data is expensive and time-consuming because pixel-level annotation requires experts in the relevant field. Currently, the combination of consistent regularization and pseudo labeling-based semi-supervised methods has shown good performance in image segmentation. However, in the training process, a portion of low-confidence pseudo labels are generated by the model. And the semi-supervised segmentation method still has the problem of distribution bias between labeled and unlabeled data. The objective of this study is to address the challenges of semi-supervised learning and improve the segmentation accuracy of semi-supervised models on medical images. To address these issues, we propose an Uncertainty-based Region Clipping Algorithm for semi-supervised medical image segmentation, which consists of two main modules. A module is introduced to compute the uncertainty of two sub-networks predictions with diversity using Monte Carlo Dropout, allowing the model to gradually learn from more reliable targets. To retain model diversity, we use different loss functions for different branches and use Non-Maximum Suppression in one of the branches. The other module is proposed to generate new samples by masking the low-confidence pixels in the original image based on uncertainty information. New samples are fed into the model to facilitate the model to generate pseudo labels with high confidence and enlarge the training data distribution. Comprehensive experiments on the combination of two benchmarks ACDC and BraTS2019 show that our proposed model outperforms state-of-the-art methods in terms of Dice, HD95 and ASD. The results reach an average Dice score of 87.86 % and a HD95 score of 4.214 mm on ACDC dataset. For the brain tumor segmentation, the results reach an average Dice score of 84.79 % and a HD score of 10.13 mm. Our proposed method improves the accuracy of semi-supervised medical image segmentation. Extensive experiments on two public medical image datasets including 2D and 3D modalities demonstrate the superiority of our model. The code is available at: https://github.com/QuintinDong/URCA.</description><subject>Algorithms</subject><subject>Brain Neoplasms - diagnostic imaging</subject><subject>Distribution bias</subject><subject>Humans</subject><subject>Image Processing, Computer-Assisted - methods</subject><subject>Magnetic Resonance Imaging - methods</subject><subject>Medical image segmentation</subject><subject>Monte Carlo Method</subject><subject>Neural Networks, Computer</subject><subject>Non-maximum suppression</subject><subject>Semi-supervised</subject><subject>Supervised Machine Learning</subject><subject>Uncertainty</subject><subject>Uncertainty-aware</subject><issn>0169-2607</issn><issn>1872-7565</issn><issn>1872-7565</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>EIF</sourceid><recordid>eNp9kMtOwzAQRS0EoqXwAyxQlmxSHCd-BLGpKl5SJSRE18ZxJsFVXthJpf49jlJYsprR-N47noPQdYSXEY7Y3W6p6y5bEkwSPxCEixM0jwQnIaeMnqK5F6UhYZjP0IVzO4wxoZSdo1ksBBcxw3P0uX1fr-6DbaPB9so0_SHMlIM8sFCatgl0ZbrONGWgqrK1pv-qg6K1gYPahG7owO7NqK4hN1pVgalVCf61rKHpVe8TLtFZoSoHV8e6QNunx4_1S7h5e35drzahJjHvQyio0oKz2FeVs7xgokg50RmFtMh0HHGRRDTLEoa9ARKRFipVIvcNTZJYxAt0O-V2tv0ewPWyNk5DVakG2sHJGDPBKUn8igUik1Tb1jkLheys_7g9yAjLkazcyZGsHMnKiaw33Rzzh8xf-2f5RekFD5MA_JV7A1Y6bcBzzY0F3cu8Nf_l_wA_a4sn</recordid><startdate>202409</startdate><enddate>202409</enddate><creator>Qin, Chendong</creator><creator>Wang, Yongxiong</creator><creator>Zhang, Jiapeng</creator><general>Elsevier B.V</general><scope>CGR</scope><scope>CUY</scope><scope>CVF</scope><scope>ECM</scope><scope>EIF</scope><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7X8</scope><orcidid>https://orcid.org/0000-0003-0680-781X</orcidid><orcidid>https://orcid.org/0000-0002-3242-0857</orcidid></search><sort><creationdate>202409</creationdate><title>URCA: Uncertainty-based region clipping algorithm for semi-supervised medical image segmentation</title><author>Qin, Chendong ; Wang, Yongxiong ; Zhang, Jiapeng</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c237t-ef5ac8763f5aad6df68f972cb5e9fbc3178415bb460c23e489fa9a8d489544383</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Algorithms</topic><topic>Brain Neoplasms - diagnostic imaging</topic><topic>Distribution bias</topic><topic>Humans</topic><topic>Image Processing, Computer-Assisted - methods</topic><topic>Magnetic Resonance Imaging - methods</topic><topic>Medical image segmentation</topic><topic>Monte Carlo Method</topic><topic>Neural Networks, Computer</topic><topic>Non-maximum suppression</topic><topic>Semi-supervised</topic><topic>Supervised Machine Learning</topic><topic>Uncertainty</topic><topic>Uncertainty-aware</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Qin, Chendong</creatorcontrib><creatorcontrib>Wang, Yongxiong</creatorcontrib><creatorcontrib>Zhang, Jiapeng</creatorcontrib><collection>Medline</collection><collection>MEDLINE</collection><collection>MEDLINE (Ovid)</collection><collection>MEDLINE</collection><collection>MEDLINE</collection><collection>PubMed</collection><collection>CrossRef</collection><collection>MEDLINE - Academic</collection><jtitle>Computer methods and programs in biomedicine</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Qin, Chendong</au><au>Wang, Yongxiong</au><au>Zhang, Jiapeng</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>URCA: Uncertainty-based region clipping algorithm for semi-supervised medical image segmentation</atitle><jtitle>Computer methods and programs in biomedicine</jtitle><addtitle>Comput Methods Programs Biomed</addtitle><date>2024-09</date><risdate>2024</risdate><volume>254</volume><spage>108278</spage><pages>108278-</pages><artnum>108278</artnum><issn>0169-2607</issn><issn>1872-7565</issn><eissn>1872-7565</eissn><abstract>•Uncertainty-based region clipping algorithm is proposed for semi-supervised medical image segmentation.•More reliable targets can be learned by the model using uncertainty information.•The ablation experiments demonstrate the validity of the proposed model. Training convolutional neural networks based on large amount of labeled data has made great progress in the field of image segmentation. However, in medical image segmentation tasks, annotating the data is expensive and time-consuming because pixel-level annotation requires experts in the relevant field. Currently, the combination of consistent regularization and pseudo labeling-based semi-supervised methods has shown good performance in image segmentation. However, in the training process, a portion of low-confidence pseudo labels are generated by the model. And the semi-supervised segmentation method still has the problem of distribution bias between labeled and unlabeled data. The objective of this study is to address the challenges of semi-supervised learning and improve the segmentation accuracy of semi-supervised models on medical images. To address these issues, we propose an Uncertainty-based Region Clipping Algorithm for semi-supervised medical image segmentation, which consists of two main modules. A module is introduced to compute the uncertainty of two sub-networks predictions with diversity using Monte Carlo Dropout, allowing the model to gradually learn from more reliable targets. To retain model diversity, we use different loss functions for different branches and use Non-Maximum Suppression in one of the branches. The other module is proposed to generate new samples by masking the low-confidence pixels in the original image based on uncertainty information. New samples are fed into the model to facilitate the model to generate pseudo labels with high confidence and enlarge the training data distribution. Comprehensive experiments on the combination of two benchmarks ACDC and BraTS2019 show that our proposed model outperforms state-of-the-art methods in terms of Dice, HD95 and ASD. The results reach an average Dice score of 87.86 % and a HD95 score of 4.214 mm on ACDC dataset. For the brain tumor segmentation, the results reach an average Dice score of 84.79 % and a HD score of 10.13 mm. Our proposed method improves the accuracy of semi-supervised medical image segmentation. Extensive experiments on two public medical image datasets including 2D and 3D modalities demonstrate the superiority of our model. The code is available at: https://github.com/QuintinDong/URCA.</abstract><cop>Ireland</cop><pub>Elsevier B.V</pub><pmid>38878360</pmid><doi>10.1016/j.cmpb.2024.108278</doi><orcidid>https://orcid.org/0000-0003-0680-781X</orcidid><orcidid>https://orcid.org/0000-0002-3242-0857</orcidid></addata></record>
fulltext fulltext
identifier ISSN: 0169-2607
ispartof Computer methods and programs in biomedicine, 2024-09, Vol.254, p.108278, Article 108278
issn 0169-2607
1872-7565
1872-7565
language eng
recordid cdi_proquest_miscellaneous_3068752476
source MEDLINE; Access via ScienceDirect (Elsevier)
subjects Algorithms
Brain Neoplasms - diagnostic imaging
Distribution bias
Humans
Image Processing, Computer-Assisted - methods
Magnetic Resonance Imaging - methods
Medical image segmentation
Monte Carlo Method
Neural Networks, Computer
Non-maximum suppression
Semi-supervised
Supervised Machine Learning
Uncertainty
Uncertainty-aware
title URCA: Uncertainty-based region clipping algorithm for semi-supervised medical image segmentation
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-05T20%3A11%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=URCA:%20Uncertainty-based%20region%20clipping%20algorithm%20for%20semi-supervised%20medical%20image%20segmentation&rft.jtitle=Computer%20methods%20and%20programs%20in%20biomedicine&rft.au=Qin,%20Chendong&rft.date=2024-09&rft.volume=254&rft.spage=108278&rft.pages=108278-&rft.artnum=108278&rft.issn=0169-2607&rft.eissn=1872-7565&rft_id=info:doi/10.1016/j.cmpb.2024.108278&rft_dat=%3Cproquest_cross%3E3068752476%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=3068752476&rft_id=info:pmid/38878360&rft_els_id=S0169260724002736&rfr_iscdi=true