Self-supervised speckle noise reduction of optical coherence tomography without clean data

Optical coherence tomography (OCT) is widely used in clinical diagnosis due to its non-invasive, real-time, and high-resolution characteristics. However, the inherent speckle noise seriously degrades the image quality, which might damage the fine structures in OCT, thus affecting the diagnosis resul...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Biomedical optics express 2022-12, Vol.13 (12), p.6357-6372
Hauptverfasser: Li, Yangxi, Fan, Yingwei, Liao, Hongen
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 6372
container_issue 12
container_start_page 6357
container_title Biomedical optics express
container_volume 13
creator Li, Yangxi
Fan, Yingwei
Liao, Hongen
description Optical coherence tomography (OCT) is widely used in clinical diagnosis due to its non-invasive, real-time, and high-resolution characteristics. However, the inherent speckle noise seriously degrades the image quality, which might damage the fine structures in OCT, thus affecting the diagnosis results. In recent years, supervised deep learning-based denoising methods have shown excellent denoising ability. To train a deep denoiser, a large number of paired noisy-clean images are required, which is difficult to achieve in clinical practice, since acquiring a speckle-free OCT image requires dozens of repeated scans and image registration. In this research, we propose a self-supervised strategy that helps build a despeckling model by training it to map neighboring pixels in a single noisy OCT image. Adjacent pixel patches are randomly selected from the original OCT image to generate two similar undersampled images, which are respectively used as the input and target images for training a deep neural network. To ensure both the despeckling and the structure-preserving effects, a multi-scale pixel patch sampler and corresponding loss functions are adopted in our practice. Through quantitative evaluation and qualitative visual comparison, we found that the proposed method performs better than state-of-the-art methods regarding despeckling effects and structure preservation. Besides, the proposed method is much easier to train and deploy without the need for clean OCT images, which has great significance in clinical practice.
doi_str_mv 10.1364/BOE.471497
format Article
fullrecord <record><control><sourceid>proquest_pubme</sourceid><recordid>TN_cdi_pubmedcentral_primary_oai_pubmedcentral_nih_gov_9774848</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2760168861</sourcerecordid><originalsourceid>FETCH-LOGICAL-c308t-234f850324d172690719cff3a849dfcd9b67abad6ca19d6dff52b6846cf25f9e3</originalsourceid><addsrcrecordid>eNpVkctKBTEMhosoKurGB5AuRRhtp51eNoKKNxBcqBs3padNPaNzpmPbUXx7R46KZpOEfPwJ-RHapeSQMsGPTm_PD7mkXMsVtFnTRlSSqGb1T72BdnJ-JlNwLglT62iDiUbpRvNN9HgHXajyOEB6azN4nAdwLx3gPk4tTuBHV9rY4xhwHErrbIddnEOC3gEucRGfkh3mH_i9LfM4Fuw6sD32tthttBZsl2HnO2-hh4vz-7Or6ub28vrs5KZyjKhS1YwH1RBWc09lLTSRVLsQmFVc--C8nglpZ9YLZ6n2wofQ1DOhuHChboIGtoWOl7rDOFuAd9CXZDszpHZh04eJtjX_J307N0_xzWgpueJqEtj_FkjxdYRczKLNDrrO9hDHbGopCBVKCTqhB0vUpZhzgvC7hhLz5YeZ_DBLPyZ47-9hv-jP99knobKISw</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2760168861</pqid></control><display><type>article</type><title>Self-supervised speckle noise reduction of optical coherence tomography without clean data</title><source>DOAJ Directory of Open Access Journals</source><source>Elektronische Zeitschriftenbibliothek - Frei zugängliche E-Journals</source><source>PubMed Central</source><creator>Li, Yangxi ; Fan, Yingwei ; Liao, Hongen</creator><creatorcontrib>Li, Yangxi ; Fan, Yingwei ; Liao, Hongen</creatorcontrib><description>Optical coherence tomography (OCT) is widely used in clinical diagnosis due to its non-invasive, real-time, and high-resolution characteristics. However, the inherent speckle noise seriously degrades the image quality, which might damage the fine structures in OCT, thus affecting the diagnosis results. In recent years, supervised deep learning-based denoising methods have shown excellent denoising ability. To train a deep denoiser, a large number of paired noisy-clean images are required, which is difficult to achieve in clinical practice, since acquiring a speckle-free OCT image requires dozens of repeated scans and image registration. In this research, we propose a self-supervised strategy that helps build a despeckling model by training it to map neighboring pixels in a single noisy OCT image. Adjacent pixel patches are randomly selected from the original OCT image to generate two similar undersampled images, which are respectively used as the input and target images for training a deep neural network. To ensure both the despeckling and the structure-preserving effects, a multi-scale pixel patch sampler and corresponding loss functions are adopted in our practice. Through quantitative evaluation and qualitative visual comparison, we found that the proposed method performs better than state-of-the-art methods regarding despeckling effects and structure preservation. Besides, the proposed method is much easier to train and deploy without the need for clean OCT images, which has great significance in clinical practice.</description><identifier>ISSN: 2156-7085</identifier><identifier>EISSN: 2156-7085</identifier><identifier>DOI: 10.1364/BOE.471497</identifier><identifier>PMID: 36589594</identifier><language>eng</language><publisher>United States: Optica Publishing Group</publisher><ispartof>Biomedical optics express, 2022-12, Vol.13 (12), p.6357-6372</ispartof><rights>2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement.</rights><rights>2022 Optica Publishing Group under the terms of the 2022 Optica Publishing Group</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c308t-234f850324d172690719cff3a849dfcd9b67abad6ca19d6dff52b6846cf25f9e3</citedby><cites>FETCH-LOGICAL-c308t-234f850324d172690719cff3a849dfcd9b67abad6ca19d6dff52b6846cf25f9e3</cites><orcidid>0000-0003-3847-9347 ; 0000-0002-1401-4054 ; 0000-0003-4535-3451</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC9774848/pdf/$$EPDF$$P50$$Gpubmedcentral$$H</linktopdf><linktohtml>$$Uhttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC9774848/$$EHTML$$P50$$Gpubmedcentral$$H</linktohtml><link.rule.ids>230,314,723,776,780,860,881,27901,27902,53766,53768</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/36589594$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Li, Yangxi</creatorcontrib><creatorcontrib>Fan, Yingwei</creatorcontrib><creatorcontrib>Liao, Hongen</creatorcontrib><title>Self-supervised speckle noise reduction of optical coherence tomography without clean data</title><title>Biomedical optics express</title><addtitle>Biomed Opt Express</addtitle><description>Optical coherence tomography (OCT) is widely used in clinical diagnosis due to its non-invasive, real-time, and high-resolution characteristics. However, the inherent speckle noise seriously degrades the image quality, which might damage the fine structures in OCT, thus affecting the diagnosis results. In recent years, supervised deep learning-based denoising methods have shown excellent denoising ability. To train a deep denoiser, a large number of paired noisy-clean images are required, which is difficult to achieve in clinical practice, since acquiring a speckle-free OCT image requires dozens of repeated scans and image registration. In this research, we propose a self-supervised strategy that helps build a despeckling model by training it to map neighboring pixels in a single noisy OCT image. Adjacent pixel patches are randomly selected from the original OCT image to generate two similar undersampled images, which are respectively used as the input and target images for training a deep neural network. To ensure both the despeckling and the structure-preserving effects, a multi-scale pixel patch sampler and corresponding loss functions are adopted in our practice. Through quantitative evaluation and qualitative visual comparison, we found that the proposed method performs better than state-of-the-art methods regarding despeckling effects and structure preservation. Besides, the proposed method is much easier to train and deploy without the need for clean OCT images, which has great significance in clinical practice.</description><issn>2156-7085</issn><issn>2156-7085</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><recordid>eNpVkctKBTEMhosoKurGB5AuRRhtp51eNoKKNxBcqBs3padNPaNzpmPbUXx7R46KZpOEfPwJ-RHapeSQMsGPTm_PD7mkXMsVtFnTRlSSqGb1T72BdnJ-JlNwLglT62iDiUbpRvNN9HgHXajyOEB6azN4nAdwLx3gPk4tTuBHV9rY4xhwHErrbIddnEOC3gEucRGfkh3mH_i9LfM4Fuw6sD32tthttBZsl2HnO2-hh4vz-7Or6ub28vrs5KZyjKhS1YwH1RBWc09lLTSRVLsQmFVc--C8nglpZ9YLZ6n2wofQ1DOhuHChboIGtoWOl7rDOFuAd9CXZDszpHZh04eJtjX_J307N0_xzWgpueJqEtj_FkjxdYRczKLNDrrO9hDHbGopCBVKCTqhB0vUpZhzgvC7hhLz5YeZ_DBLPyZ47-9hv-jP99knobKISw</recordid><startdate>20221201</startdate><enddate>20221201</enddate><creator>Li, Yangxi</creator><creator>Fan, Yingwei</creator><creator>Liao, Hongen</creator><general>Optica Publishing Group</general><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7X8</scope><scope>5PM</scope><orcidid>https://orcid.org/0000-0003-3847-9347</orcidid><orcidid>https://orcid.org/0000-0002-1401-4054</orcidid><orcidid>https://orcid.org/0000-0003-4535-3451</orcidid></search><sort><creationdate>20221201</creationdate><title>Self-supervised speckle noise reduction of optical coherence tomography without clean data</title><author>Li, Yangxi ; Fan, Yingwei ; Liao, Hongen</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c308t-234f850324d172690719cff3a849dfcd9b67abad6ca19d6dff52b6846cf25f9e3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Li, Yangxi</creatorcontrib><creatorcontrib>Fan, Yingwei</creatorcontrib><creatorcontrib>Liao, Hongen</creatorcontrib><collection>PubMed</collection><collection>CrossRef</collection><collection>MEDLINE - Academic</collection><collection>PubMed Central (Full Participant titles)</collection><jtitle>Biomedical optics express</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Li, Yangxi</au><au>Fan, Yingwei</au><au>Liao, Hongen</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Self-supervised speckle noise reduction of optical coherence tomography without clean data</atitle><jtitle>Biomedical optics express</jtitle><addtitle>Biomed Opt Express</addtitle><date>2022-12-01</date><risdate>2022</risdate><volume>13</volume><issue>12</issue><spage>6357</spage><epage>6372</epage><pages>6357-6372</pages><issn>2156-7085</issn><eissn>2156-7085</eissn><abstract>Optical coherence tomography (OCT) is widely used in clinical diagnosis due to its non-invasive, real-time, and high-resolution characteristics. However, the inherent speckle noise seriously degrades the image quality, which might damage the fine structures in OCT, thus affecting the diagnosis results. In recent years, supervised deep learning-based denoising methods have shown excellent denoising ability. To train a deep denoiser, a large number of paired noisy-clean images are required, which is difficult to achieve in clinical practice, since acquiring a speckle-free OCT image requires dozens of repeated scans and image registration. In this research, we propose a self-supervised strategy that helps build a despeckling model by training it to map neighboring pixels in a single noisy OCT image. Adjacent pixel patches are randomly selected from the original OCT image to generate two similar undersampled images, which are respectively used as the input and target images for training a deep neural network. To ensure both the despeckling and the structure-preserving effects, a multi-scale pixel patch sampler and corresponding loss functions are adopted in our practice. Through quantitative evaluation and qualitative visual comparison, we found that the proposed method performs better than state-of-the-art methods regarding despeckling effects and structure preservation. Besides, the proposed method is much easier to train and deploy without the need for clean OCT images, which has great significance in clinical practice.</abstract><cop>United States</cop><pub>Optica Publishing Group</pub><pmid>36589594</pmid><doi>10.1364/BOE.471497</doi><tpages>16</tpages><orcidid>https://orcid.org/0000-0003-3847-9347</orcidid><orcidid>https://orcid.org/0000-0002-1401-4054</orcidid><orcidid>https://orcid.org/0000-0003-4535-3451</orcidid><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 2156-7085
ispartof Biomedical optics express, 2022-12, Vol.13 (12), p.6357-6372
issn 2156-7085
2156-7085
language eng
recordid cdi_pubmedcentral_primary_oai_pubmedcentral_nih_gov_9774848
source DOAJ Directory of Open Access Journals; Elektronische Zeitschriftenbibliothek - Frei zugängliche E-Journals; PubMed Central
title Self-supervised speckle noise reduction of optical coherence tomography without clean data
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-09T04%3A53%3A55IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_pubme&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Self-supervised%20speckle%20noise%20reduction%20of%20optical%20coherence%20tomography%20without%20clean%20data&rft.jtitle=Biomedical%20optics%20express&rft.au=Li,%20Yangxi&rft.date=2022-12-01&rft.volume=13&rft.issue=12&rft.spage=6357&rft.epage=6372&rft.pages=6357-6372&rft.issn=2156-7085&rft.eissn=2156-7085&rft_id=info:doi/10.1364/BOE.471497&rft_dat=%3Cproquest_pubme%3E2760168861%3C/proquest_pubme%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2760168861&rft_id=info:pmid/36589594&rfr_iscdi=true