Artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy

Abstract Background Advancements in biomedical optical imaging have enabled researchers to achieve cellular-level imaging in the living human body. However, research-grade technology is not always widely available in routine clinical practice. In this paper, we incorporated artificial intelligence (...

Full description

Saved in:
Bibliographic Details
Main Authors: Joanne Li, Jianfei Liu, Vineeta Das, Hong Le, Nancy Aguilera, Andrew J. Bower, John P. Giannini, Rongwen Lu, Sarah Abouassali, Emily Y. Chew, Brian P. Brooks, Wadih M. Zein, Laryssa A. Huryn, Andrei Volkov, Tao Liu, Johnny Tam
Format: Article
Language:English
Published: Nature Portfolio 2025-04-01
Series:Communications Medicine
Online Access:https://doi.org/10.1038/s43856-025-00803-z
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850143408460398592
author Joanne Li
Jianfei Liu
Vineeta Das
Hong Le
Nancy Aguilera
Andrew J. Bower
John P. Giannini
Rongwen Lu
Sarah Abouassali
Emily Y. Chew
Brian P. Brooks
Wadih M. Zein
Laryssa A. Huryn
Andrei Volkov
Tao Liu
Johnny Tam
author_facet Joanne Li
Jianfei Liu
Vineeta Das
Hong Le
Nancy Aguilera
Andrew J. Bower
John P. Giannini
Rongwen Lu
Sarah Abouassali
Emily Y. Chew
Brian P. Brooks
Wadih M. Zein
Laryssa A. Huryn
Andrei Volkov
Tao Liu
Johnny Tam
author_sort Joanne Li
collection DOAJ
description Abstract Background Advancements in biomedical optical imaging have enabled researchers to achieve cellular-level imaging in the living human body. However, research-grade technology is not always widely available in routine clinical practice. In this paper, we incorporated artificial intelligence (AI) with standard clinical imaging to successfully obtain images of the retinal pigment epithelial (RPE) cells in living human eyes. Methods Following intravenous injection of indocyanine green (ICG) dye, subjects were imaged by both conventional instruments and adaptive optics (AO) ophthalmoscopy. To improve the visibility of RPE cells in conventional ICG images, we demonstrate both a hardware approach using a custom lens add-on and an AI-based approach using a stratified cycleGAN network. Results We observe similar fluorescent mosaic patterns arising from labeled RPE cells on both conventional and AO images, suggesting that cellular-level imaging of RPE may be obtainable using conventional imaging, albeit at lower resolution. Results show that higher resolution ICG RPE images of both healthy and diseased eyes can be obtained from conventional images using AI with a potential 220-fold improvement in time. Conclusions The application of using AI as an add-on module for existing instrumentation is an important step towards routine screening and detection of disease at earlier stages.
format Article
id doaj-art-ac4bceab78e644c6bd8912bbe0db9ac8
institution OA Journals
issn 2730-664X
language English
publishDate 2025-04-01
publisher Nature Portfolio
record_format Article
series Communications Medicine
spelling doaj-art-ac4bceab78e644c6bd8912bbe0db9ac82025-08-20T02:28:42ZengNature PortfolioCommunications Medicine2730-664X2025-04-015111110.1038/s43856-025-00803-zArtificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopyJoanne Li0Jianfei Liu1Vineeta Das2Hong Le3Nancy Aguilera4Andrew J. Bower5John P. Giannini6Rongwen Lu7Sarah Abouassali8Emily Y. Chew9Brian P. Brooks10Wadih M. Zein11Laryssa A. Huryn12Andrei Volkov13Tao Liu14Johnny Tam15National Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthNational Eye Institute, National Institutes of HealthAbstract Background Advancements in biomedical optical imaging have enabled researchers to achieve cellular-level imaging in the living human body. However, research-grade technology is not always widely available in routine clinical practice. In this paper, we incorporated artificial intelligence (AI) with standard clinical imaging to successfully obtain images of the retinal pigment epithelial (RPE) cells in living human eyes. Methods Following intravenous injection of indocyanine green (ICG) dye, subjects were imaged by both conventional instruments and adaptive optics (AO) ophthalmoscopy. To improve the visibility of RPE cells in conventional ICG images, we demonstrate both a hardware approach using a custom lens add-on and an AI-based approach using a stratified cycleGAN network. Results We observe similar fluorescent mosaic patterns arising from labeled RPE cells on both conventional and AO images, suggesting that cellular-level imaging of RPE may be obtainable using conventional imaging, albeit at lower resolution. Results show that higher resolution ICG RPE images of both healthy and diseased eyes can be obtained from conventional images using AI with a potential 220-fold improvement in time. Conclusions The application of using AI as an add-on module for existing instrumentation is an important step towards routine screening and detection of disease at earlier stages.https://doi.org/10.1038/s43856-025-00803-z
spellingShingle Joanne Li
Jianfei Liu
Vineeta Das
Hong Le
Nancy Aguilera
Andrew J. Bower
John P. Giannini
Rongwen Lu
Sarah Abouassali
Emily Y. Chew
Brian P. Brooks
Wadih M. Zein
Laryssa A. Huryn
Andrei Volkov
Tao Liu
Johnny Tam
Artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy
Communications Medicine
title Artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy
title_full Artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy
title_fullStr Artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy
title_full_unstemmed Artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy
title_short Artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy
title_sort artificial intelligence assisted clinical fluorescence imaging achieves in vivo cellular resolution comparable to adaptive optics ophthalmoscopy
url https://doi.org/10.1038/s43856-025-00803-z
work_keys_str_mv AT joanneli artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT jianfeiliu artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT vineetadas artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT hongle artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT nancyaguilera artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT andrewjbower artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT johnpgiannini artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT rongwenlu artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT sarahabouassali artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT emilyychew artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT brianpbrooks artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT wadihmzein artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT laryssaahuryn artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT andreivolkov artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT taoliu artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy
AT johnnytam artificialintelligenceassistedclinicalfluorescenceimagingachievesinvivocellularresolutioncomparabletoadaptiveopticsophthalmoscopy