Loading…

Translation of Color Fundus Photography into Fluorescein Angiography Using Deep Learning for Enhanced Diabetic Retinopathy Screening

To develop and validate a deep learning model that can transform color fundus (CF) photography into corresponding venous and late-phase fundus fluorescein angiography (FFA) images. Cross-sectional study. We included 51 370 CF-venous FFA pairs and 14 644 CF-late FFA pairs from 4438 patients for model...

Full description

Saved in:
Bibliographic Details
Published in:Ophthalmology science (Online) 2023-12, Vol.3 (4), p.100401-100401, Article 100401
Main Authors: Shi, Danli, Zhang, Weiyi, He, Shuang, Chen, Yanxian, Song, Fan, Liu, Shunming, Wang, Ruobing, Zheng, Yingfeng, He, Mingguang
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by cdi_FETCH-LOGICAL-c443t-c00e1ef69587b6d616a8cddbcfa0c1ad987600987d8bfa0790645cc3f22e23673
cites cdi_FETCH-LOGICAL-c443t-c00e1ef69587b6d616a8cddbcfa0c1ad987600987d8bfa0790645cc3f22e23673
container_end_page 100401
container_issue 4
container_start_page 100401
container_title Ophthalmology science (Online)
container_volume 3
creator Shi, Danli
Zhang, Weiyi
He, Shuang
Chen, Yanxian
Song, Fan
Liu, Shunming
Wang, Ruobing
Zheng, Yingfeng
He, Mingguang
description To develop and validate a deep learning model that can transform color fundus (CF) photography into corresponding venous and late-phase fundus fluorescein angiography (FFA) images. Cross-sectional study. We included 51 370 CF-venous FFA pairs and 14 644 CF-late FFA pairs from 4438 patients for model development. External testing involved 50 eyes with CF-FFA pairs and 2 public datasets for diabetic retinopathy (DR) classification, with 86 952 CF from EyePACs, and 1744 CF from MESSIDOR2. We trained a deep-learning model to transform CF into corresponding venous and late-phase FFA images. The translated FFA images’ quality was evaluated quantitatively on the internal test set and subjectively on 100 eyes with CF-FFA paired images (50 from external), based on the realisticity of the global image, anatomical landmarks (macula, optic disc, and vessels), and lesions. Moreover, we validated the clinical utility of the translated FFA for classifying 5-class DR and diabetic macular edema (DME) in the EyePACs and MESSIDOR2 datasets. Image generation was quantitatively assessed by structural similarity measures (SSIM), and subjectively by 2 clinical experts on a 5-point scale (1 refers real FFA); intragrader agreement was assessed by kappa. The DR classification accuracy was assessed by area under the receiver operating characteristic curve. The SSIM of the translated FFA images were > 0.6, and the subjective quality scores ranged from 1.37 to 2.60. Both experts reported similar quality scores with substantial agreement (all kappas > 0.8). Adding the generated FFA on top of CF improved DR classification in the EyePACs and MESSIDOR2 datasets, with the area under the receiver operating characteristic curve increased from 0.912 to 0.939 on the EyePACs dataset and from 0.952 to 0.972 on the MESSIDOR2 dataset. The DME area under the receiver operating characteristic curve also increased from 0.927 to 0.974 in the MESSIDOR2 dataset. Our CF-to-FFA framework produced realistic FFA images. Moreover, adding the translated FFA images on top of CF improved the accuracy of DR screening. These results suggest that CF-to-FFA translation could be used as a surrogate method when FFA examination is not feasible and as a simple add-on to improve DR screening. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
doi_str_mv 10.1016/j.xops.2023.100401
format article
fullrecord <record><control><sourceid>proquest_doaj_</sourceid><recordid>TN_cdi_doaj_primary_oai_doaj_org_article_32d6500ade754903800fcc8b67e33bac</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><els_id>S2666914523001331</els_id><doaj_id>oai_doaj_org_article_32d6500ade754903800fcc8b67e33bac</doaj_id><sourcerecordid>2895699973</sourcerecordid><originalsourceid>FETCH-LOGICAL-c443t-c00e1ef69587b6d616a8cddbcfa0c1ad987600987d8bfa0790645cc3f22e23673</originalsourceid><addsrcrecordid>eNp9UcFu1DAQjRBIVKU_wMlHLruM7cRJJC7VtguVVgJBe7Ym9mTXq9QOdoLonQ_HIYA4cRl7nuc9j94ritccthy4envefg9j2goQMgNQAn9WXAil1KblZfX8n_vL4iqlMwCIiktR8ovix31EnwacXPAs9GwXhhDZfvZ2TuzTKUzhGHE8PTHnp8D2wxwiJUPOs2t_dH8eH5LzR3ZDNLIDYfRL12edW39Cb8iyG4cdTc6wz7n6MOKUWV9MJFpmXxUvehwSXf0-L4uH_e397sPm8PH93e76sDFlKaeNASBOvWqrpu6UVVxhY6ztTI9gONq2qRVArrbpMlS3oMrKGNkLQUKqWl4Wd6uuDXjWY3SPGJ90QKd_ASEeNca85UBaCqsqALRUV2ULsgHojWk6VZOUHZqs9WbVGmP4OlOa9KPLxgwDegpz0qJpK9W2bS3zqFhHTQwpRer_fs1BLwnqs14S1EuCek0wk96tJMqGfHMUdTKOFjNdJDPljd3_6D8BM8ymUQ</addsrcrecordid><sourcetype>Open Website</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2895699973</pqid></control><display><type>article</type><title>Translation of Color Fundus Photography into Fluorescein Angiography Using Deep Learning for Enhanced Diabetic Retinopathy Screening</title><source>ScienceDirect</source><source>PubMed Central</source><creator>Shi, Danli ; Zhang, Weiyi ; He, Shuang ; Chen, Yanxian ; Song, Fan ; Liu, Shunming ; Wang, Ruobing ; Zheng, Yingfeng ; He, Mingguang</creator><creatorcontrib>Shi, Danli ; Zhang, Weiyi ; He, Shuang ; Chen, Yanxian ; Song, Fan ; Liu, Shunming ; Wang, Ruobing ; Zheng, Yingfeng ; He, Mingguang</creatorcontrib><description>To develop and validate a deep learning model that can transform color fundus (CF) photography into corresponding venous and late-phase fundus fluorescein angiography (FFA) images. Cross-sectional study. We included 51 370 CF-venous FFA pairs and 14 644 CF-late FFA pairs from 4438 patients for model development. External testing involved 50 eyes with CF-FFA pairs and 2 public datasets for diabetic retinopathy (DR) classification, with 86 952 CF from EyePACs, and 1744 CF from MESSIDOR2. We trained a deep-learning model to transform CF into corresponding venous and late-phase FFA images. The translated FFA images’ quality was evaluated quantitatively on the internal test set and subjectively on 100 eyes with CF-FFA paired images (50 from external), based on the realisticity of the global image, anatomical landmarks (macula, optic disc, and vessels), and lesions. Moreover, we validated the clinical utility of the translated FFA for classifying 5-class DR and diabetic macular edema (DME) in the EyePACs and MESSIDOR2 datasets. Image generation was quantitatively assessed by structural similarity measures (SSIM), and subjectively by 2 clinical experts on a 5-point scale (1 refers real FFA); intragrader agreement was assessed by kappa. The DR classification accuracy was assessed by area under the receiver operating characteristic curve. The SSIM of the translated FFA images were &gt; 0.6, and the subjective quality scores ranged from 1.37 to 2.60. Both experts reported similar quality scores with substantial agreement (all kappas &gt; 0.8). Adding the generated FFA on top of CF improved DR classification in the EyePACs and MESSIDOR2 datasets, with the area under the receiver operating characteristic curve increased from 0.912 to 0.939 on the EyePACs dataset and from 0.952 to 0.972 on the MESSIDOR2 dataset. The DME area under the receiver operating characteristic curve also increased from 0.927 to 0.974 in the MESSIDOR2 dataset. Our CF-to-FFA framework produced realistic FFA images. Moreover, adding the translated FFA images on top of CF improved the accuracy of DR screening. These results suggest that CF-to-FFA translation could be used as a surrogate method when FFA examination is not feasible and as a simple add-on to improve DR screening. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</description><identifier>ISSN: 2666-9145</identifier><identifier>EISSN: 2666-9145</identifier><identifier>DOI: 10.1016/j.xops.2023.100401</identifier><language>eng</language><publisher>Elsevier Inc</publisher><subject>Diabetic retinopathy ; FFA generation ; Generative Adversarial Networks ; Image-to-image translation</subject><ispartof>Ophthalmology science (Online), 2023-12, Vol.3 (4), p.100401-100401, Article 100401</ispartof><rights>2023 American Academy of Ophthalmology</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c443t-c00e1ef69587b6d616a8cddbcfa0c1ad987600987d8bfa0790645cc3f22e23673</citedby><cites>FETCH-LOGICAL-c443t-c00e1ef69587b6d616a8cddbcfa0c1ad987600987d8bfa0790645cc3f22e23673</cites><orcidid>0000-0002-6912-2810</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://www.sciencedirect.com/science/article/pii/S2666914523001331$$EHTML$$P50$$Gelsevier$$Hfree_for_read</linktohtml><link.rule.ids>314,780,784,3549,27924,27925,45780</link.rule.ids></links><search><creatorcontrib>Shi, Danli</creatorcontrib><creatorcontrib>Zhang, Weiyi</creatorcontrib><creatorcontrib>He, Shuang</creatorcontrib><creatorcontrib>Chen, Yanxian</creatorcontrib><creatorcontrib>Song, Fan</creatorcontrib><creatorcontrib>Liu, Shunming</creatorcontrib><creatorcontrib>Wang, Ruobing</creatorcontrib><creatorcontrib>Zheng, Yingfeng</creatorcontrib><creatorcontrib>He, Mingguang</creatorcontrib><title>Translation of Color Fundus Photography into Fluorescein Angiography Using Deep Learning for Enhanced Diabetic Retinopathy Screening</title><title>Ophthalmology science (Online)</title><description>To develop and validate a deep learning model that can transform color fundus (CF) photography into corresponding venous and late-phase fundus fluorescein angiography (FFA) images. Cross-sectional study. We included 51 370 CF-venous FFA pairs and 14 644 CF-late FFA pairs from 4438 patients for model development. External testing involved 50 eyes with CF-FFA pairs and 2 public datasets for diabetic retinopathy (DR) classification, with 86 952 CF from EyePACs, and 1744 CF from MESSIDOR2. We trained a deep-learning model to transform CF into corresponding venous and late-phase FFA images. The translated FFA images’ quality was evaluated quantitatively on the internal test set and subjectively on 100 eyes with CF-FFA paired images (50 from external), based on the realisticity of the global image, anatomical landmarks (macula, optic disc, and vessels), and lesions. Moreover, we validated the clinical utility of the translated FFA for classifying 5-class DR and diabetic macular edema (DME) in the EyePACs and MESSIDOR2 datasets. Image generation was quantitatively assessed by structural similarity measures (SSIM), and subjectively by 2 clinical experts on a 5-point scale (1 refers real FFA); intragrader agreement was assessed by kappa. The DR classification accuracy was assessed by area under the receiver operating characteristic curve. The SSIM of the translated FFA images were &gt; 0.6, and the subjective quality scores ranged from 1.37 to 2.60. Both experts reported similar quality scores with substantial agreement (all kappas &gt; 0.8). Adding the generated FFA on top of CF improved DR classification in the EyePACs and MESSIDOR2 datasets, with the area under the receiver operating characteristic curve increased from 0.912 to 0.939 on the EyePACs dataset and from 0.952 to 0.972 on the MESSIDOR2 dataset. The DME area under the receiver operating characteristic curve also increased from 0.927 to 0.974 in the MESSIDOR2 dataset. Our CF-to-FFA framework produced realistic FFA images. Moreover, adding the translated FFA images on top of CF improved the accuracy of DR screening. These results suggest that CF-to-FFA translation could be used as a surrogate method when FFA examination is not feasible and as a simple add-on to improve DR screening. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</description><subject>Diabetic retinopathy</subject><subject>FFA generation</subject><subject>Generative Adversarial Networks</subject><subject>Image-to-image translation</subject><issn>2666-9145</issn><issn>2666-9145</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>DOA</sourceid><recordid>eNp9UcFu1DAQjRBIVKU_wMlHLruM7cRJJC7VtguVVgJBe7Ym9mTXq9QOdoLonQ_HIYA4cRl7nuc9j94ritccthy4envefg9j2goQMgNQAn9WXAil1KblZfX8n_vL4iqlMwCIiktR8ovix31EnwacXPAs9GwXhhDZfvZ2TuzTKUzhGHE8PTHnp8D2wxwiJUPOs2t_dH8eH5LzR3ZDNLIDYfRL12edW39Cb8iyG4cdTc6wz7n6MOKUWV9MJFpmXxUvehwSXf0-L4uH_e397sPm8PH93e76sDFlKaeNASBOvWqrpu6UVVxhY6ztTI9gONq2qRVArrbpMlS3oMrKGNkLQUKqWl4Wd6uuDXjWY3SPGJ90QKd_ASEeNca85UBaCqsqALRUV2ULsgHojWk6VZOUHZqs9WbVGmP4OlOa9KPLxgwDegpz0qJpK9W2bS3zqFhHTQwpRer_fs1BLwnqs14S1EuCek0wk96tJMqGfHMUdTKOFjNdJDPljd3_6D8BM8ymUQ</recordid><startdate>202312</startdate><enddate>202312</enddate><creator>Shi, Danli</creator><creator>Zhang, Weiyi</creator><creator>He, Shuang</creator><creator>Chen, Yanxian</creator><creator>Song, Fan</creator><creator>Liu, Shunming</creator><creator>Wang, Ruobing</creator><creator>Zheng, Yingfeng</creator><creator>He, Mingguang</creator><general>Elsevier Inc</general><general>Elsevier</general><scope>6I.</scope><scope>AAFTH</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7X8</scope><scope>DOA</scope><orcidid>https://orcid.org/0000-0002-6912-2810</orcidid></search><sort><creationdate>202312</creationdate><title>Translation of Color Fundus Photography into Fluorescein Angiography Using Deep Learning for Enhanced Diabetic Retinopathy Screening</title><author>Shi, Danli ; Zhang, Weiyi ; He, Shuang ; Chen, Yanxian ; Song, Fan ; Liu, Shunming ; Wang, Ruobing ; Zheng, Yingfeng ; He, Mingguang</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c443t-c00e1ef69587b6d616a8cddbcfa0c1ad987600987d8bfa0790645cc3f22e23673</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Diabetic retinopathy</topic><topic>FFA generation</topic><topic>Generative Adversarial Networks</topic><topic>Image-to-image translation</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Shi, Danli</creatorcontrib><creatorcontrib>Zhang, Weiyi</creatorcontrib><creatorcontrib>He, Shuang</creatorcontrib><creatorcontrib>Chen, Yanxian</creatorcontrib><creatorcontrib>Song, Fan</creatorcontrib><creatorcontrib>Liu, Shunming</creatorcontrib><creatorcontrib>Wang, Ruobing</creatorcontrib><creatorcontrib>Zheng, Yingfeng</creatorcontrib><creatorcontrib>He, Mingguang</creatorcontrib><collection>ScienceDirect Open Access Titles</collection><collection>Elsevier:ScienceDirect:Open Access</collection><collection>CrossRef</collection><collection>MEDLINE - Academic</collection><collection>Directory of Open Access Journals</collection><jtitle>Ophthalmology science (Online)</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Shi, Danli</au><au>Zhang, Weiyi</au><au>He, Shuang</au><au>Chen, Yanxian</au><au>Song, Fan</au><au>Liu, Shunming</au><au>Wang, Ruobing</au><au>Zheng, Yingfeng</au><au>He, Mingguang</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Translation of Color Fundus Photography into Fluorescein Angiography Using Deep Learning for Enhanced Diabetic Retinopathy Screening</atitle><jtitle>Ophthalmology science (Online)</jtitle><date>2023-12</date><risdate>2023</risdate><volume>3</volume><issue>4</issue><spage>100401</spage><epage>100401</epage><pages>100401-100401</pages><artnum>100401</artnum><issn>2666-9145</issn><eissn>2666-9145</eissn><abstract>To develop and validate a deep learning model that can transform color fundus (CF) photography into corresponding venous and late-phase fundus fluorescein angiography (FFA) images. Cross-sectional study. We included 51 370 CF-venous FFA pairs and 14 644 CF-late FFA pairs from 4438 patients for model development. External testing involved 50 eyes with CF-FFA pairs and 2 public datasets for diabetic retinopathy (DR) classification, with 86 952 CF from EyePACs, and 1744 CF from MESSIDOR2. We trained a deep-learning model to transform CF into corresponding venous and late-phase FFA images. The translated FFA images’ quality was evaluated quantitatively on the internal test set and subjectively on 100 eyes with CF-FFA paired images (50 from external), based on the realisticity of the global image, anatomical landmarks (macula, optic disc, and vessels), and lesions. Moreover, we validated the clinical utility of the translated FFA for classifying 5-class DR and diabetic macular edema (DME) in the EyePACs and MESSIDOR2 datasets. Image generation was quantitatively assessed by structural similarity measures (SSIM), and subjectively by 2 clinical experts on a 5-point scale (1 refers real FFA); intragrader agreement was assessed by kappa. The DR classification accuracy was assessed by area under the receiver operating characteristic curve. The SSIM of the translated FFA images were &gt; 0.6, and the subjective quality scores ranged from 1.37 to 2.60. Both experts reported similar quality scores with substantial agreement (all kappas &gt; 0.8). Adding the generated FFA on top of CF improved DR classification in the EyePACs and MESSIDOR2 datasets, with the area under the receiver operating characteristic curve increased from 0.912 to 0.939 on the EyePACs dataset and from 0.952 to 0.972 on the MESSIDOR2 dataset. The DME area under the receiver operating characteristic curve also increased from 0.927 to 0.974 in the MESSIDOR2 dataset. Our CF-to-FFA framework produced realistic FFA images. Moreover, adding the translated FFA images on top of CF improved the accuracy of DR screening. These results suggest that CF-to-FFA translation could be used as a surrogate method when FFA examination is not feasible and as a simple add-on to improve DR screening. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.</abstract><pub>Elsevier Inc</pub><doi>10.1016/j.xops.2023.100401</doi><tpages>1</tpages><orcidid>https://orcid.org/0000-0002-6912-2810</orcidid><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 2666-9145
ispartof Ophthalmology science (Online), 2023-12, Vol.3 (4), p.100401-100401, Article 100401
issn 2666-9145
2666-9145
language eng
recordid cdi_doaj_primary_oai_doaj_org_article_32d6500ade754903800fcc8b67e33bac
source ScienceDirect; PubMed Central
subjects Diabetic retinopathy
FFA generation
Generative Adversarial Networks
Image-to-image translation
title Translation of Color Fundus Photography into Fluorescein Angiography Using Deep Learning for Enhanced Diabetic Retinopathy Screening
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-24T15%3A39%3A29IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_doaj_&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Translation%20of%20Color%20Fundus%20Photography%20into%20Fluorescein%20Angiography%20Using%20Deep%20Learning%20for%20Enhanced%20Diabetic%20Retinopathy%20Screening&rft.jtitle=Ophthalmology%20science%20(Online)&rft.au=Shi,%20Danli&rft.date=2023-12&rft.volume=3&rft.issue=4&rft.spage=100401&rft.epage=100401&rft.pages=100401-100401&rft.artnum=100401&rft.issn=2666-9145&rft.eissn=2666-9145&rft_id=info:doi/10.1016/j.xops.2023.100401&rft_dat=%3Cproquest_doaj_%3E2895699973%3C/proquest_doaj_%3E%3Cgrp_id%3Ecdi_FETCH-LOGICAL-c443t-c00e1ef69587b6d616a8cddbcfa0c1ad987600987d8bfa0790645cc3f22e23673%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=2895699973&rft_id=info:pmid/&rfr_iscdi=true