Synthetic artificial intelligence using generative adversarial network for retinal imaging in detection of age-related macular degeneration.

Introduction: Age-related macular degeneration (AMD) is one of the leading causes of vision impairment globally and early detection is crucial to prevent vision loss. However, the screening of AMD is resource dependent and demands experienced healthcare providers. Recently, deep learning (DL) systems have shown the potential for effective detection of various eye diseases from retinal fundus images, but the development of such robust systems requires a large amount of datasets, which could be limited by prevalence of the disease and privacy of patient. As in the case of AMD, the advanced phenotype is often scarce for conducting DL analysis, which may be tackled via generating synthetic images using Generative Adversarial Networks (GANs). This study aims to develop GAN-synthesized fundus photos with AMD lesions, and to assess the realness of these images with an objective scale. Methods: To build our GAN models, a total of 125,012 fundus photos were used from a real-world non-AMD phenotypical dataset. StyleGAN2 and human-in-the-loop (HITL) method were then applied to synthesize fundus images with AMD features. To objectively assess the quality of the synthesized images, we proposed a novel realness scale based on the frequency of the broken vessels observed in the fundus photos. Four residents conducted two rounds of gradings on 300 images to distinguish real from synthetic images, based on their subjective impression and the objective scale respectively. Results and discussion: The introduction of HITL training increased the percentage of synthetic images with AMD lesions, despite the limited number of AMD images in the initial training dataset. Qualitatively, the synthesized images have been proven to be robust in that our residents had limited ability to distinguish real from synthetic ones, as evidenced by an overall accuracy of 0.66 (95% CI: 0.61-0.66) and Cohen's kappa of 0.320. For the non-referable AMD classes (no or early AMD), the accuracy was only 0.51. With the objective scale, the overall accuracy improved to 0.72. In conclusion, GAN models built with HITL training are capable of producing realistic-looking fundus images that could fool human experts, while our objective realness scale based on broken vessels can help identifying the synthetic fundus photos.

Early detection of visual impairment in young children using a smartphone-based deep learning system.

Early detection of visual impairment is crucial but is frequently missed in young children, who are capable of only limited cooperation with standard vision tests. Although certain features of visually impaired children, such as facial appearance and ocular movements, can assist ophthalmic practice, applying these features to real-world screening remains challenging. Here, we present a mobile health (mHealth) system, the smartphone-based Apollo Infant Sight (AIS), which identifies visually impaired children with any of 16 ophthalmic disorders by recording and analyzing their gazing behaviors and facial features under visual stimuli. Videos from 3,652 children (≤48 months in age; 54.5% boys) were prospectively collected to develop and validate this system. For detecting visual impairment, AIS achieved an area under the receiver operating curve (AUC) of 0.940 in an internal validation set and an AUC of 0.843 in an external validation set collected in multiple ophthalmology clinics across China. In a further test of AIS for at-home implementation by untrained parents or caregivers using their smartphones, the system was able to adapt to different testing conditions and achieved an AUC of 0.859. This mHealth system has the potential to be used by healthcare professionals, parents and caregivers for identifying young children with visual impairment across a wide range of ophthalmic disorders.

Acceptance and Perception of Artificial Intelligence Usability in Eye Care (APPRAISE) for Ophthalmologists: A Multinational Perspective.

Background: Many artificial intelligence (AI) studies have focused on development of AI models, novel techniques, and reporting guidelines. However, little is understood about clinicians' perspectives of AI applications in medical fields including ophthalmology, particularly in light of recent regulatory guidelines. The aim for this study was to evaluate the perspectives of ophthalmologists regarding AI in 4 major eye conditions: diabetic retinopathy (DR), glaucoma, age-related macular degeneration (AMD) and cataract. Methods: This was a multi-national survey of ophthalmologists between March 1st, 2020 to February 29th, 2021 disseminated via the major global ophthalmology societies. The survey was designed based on microsystem, mesosystem and macrosystem questions, and the software as a medical device (SaMD) regulatory framework chaired by the Food and Drug Administration (FDA). Factors associated with AI adoption for ophthalmology analyzed with multivariable logistic regression random forest machine learning. Results: One thousand one hundred seventy-six ophthalmologists from 70 countries participated with a response rate ranging from 78.8 to 85.8% per question. Ophthalmologists were more willing to use AI as clinical assistive tools (88.1%, n = 890/1,010) especially those with over 20 years' experience (OR 3.70, 95% CI: 1.10-12.5, p = 0.035), as compared to clinical decision support tools (78.8%, n = 796/1,010) or diagnostic tools (64.5%, n = 651). A majority of Ophthalmologists felt that AI is most relevant to DR (78.2%), followed by glaucoma (70.7%), AMD (66.8%), and cataract (51.4%) detection. Many participants were confident their roles will not be replaced (68.2%, n = 632/927), and felt COVID-19 catalyzed willingness to adopt AI (80.9%, n = 750/927). Common barriers to implementation include medical liability from errors (72.5%, n = 672/927) whereas enablers include improving access (94.5%, n = 876/927). Machine learning modeling predicted acceptance from participant demographics with moderate to high accuracy, and area under the receiver operating curves of 0.63-0.83. Conclusion: Ophthalmologists are receptive to adopting AI as assistive tools for DR, glaucoma, and AMD. Furthermore, ML is a useful method that can be applied to evaluate predictive factors on clinical qualitative questionnaires. This study outlines actionable insights for future research and facilitation interventions to drive adoption and operationalization of AI tools for Ophthalmology.

Advances in OCT Imaging in Myopia and Pathologic Myopia.

Advances in imaging with optical coherence tomography (OCT) and optical coherence tomography angiography (OCTA) technology, including the development of swept source OCT/OCTA, widefield or ultra-widefield systems, have greatly improved the understanding, diagnosis, and treatment of myopia and myopia-related complications. Anterior segment OCT is useful for imaging the anterior segment of myopes, providing the basis for implantable collamer lens optimization, or detecting intraocular lens decentration in high myopic patients. OCT has enhanced imaging of vitreous properties, and measurement of choroidal thickness in myopic eyes. Widefield OCT systems have greatly improved the visualization of peripheral retinal lesions and have enabled the evaluation of wide staphyloma and ocular curvature. Based on OCT imaging, a new classification system and guidelines for the management of myopic traction maculopathy have been proposed; different dome-shaped macula morphologies have been described; and myopia-related abnormalities in the optic nerve and peripapillary region have been demonstrated. OCTA can quantitatively evaluate the retinal microvasculature and choriocapillaris, which is useful for the early detection of myopic choroidal neovascularization and the evaluation of anti-vascular endothelial growth factor therapy in these patients. In addition, the application of artificial intelligence in OCT/OCTA imaging in myopia has achieved promising results.

MACULAR SENSITIVITY AND CAPILLARY PERFUSION IN HIGHLY MYOPIC EYES WITH MYOPIC MACULAR DEGENERATION.

PURPOSE: To evaluate the interrelationship between macular sensitivity and retinal perfusion density (PD) in eyes with myopic macular degeneration (MMD). METHODS: One hundred and thirty-eight highly myopic eyes from 82 adult participants were recruited. Macular sensitivity was evaluated using the Microperimeter MP-3. Retinal PD was measured using the PLEX Elite 9000 swept source optical coherence tomography angiography. Macular sensitivity values between different categories of MMD and its relationship with optical coherence tomography angiography measurements were evaluated using multivariable linear mixed models, adjusting for age and axial length. RESULTS: Macular sensitivity reduced with increasing severity of MMD (ß ≤ -0.95, P < 0.001), whereas the best-corrected visual acuity was not associated with MMD severity (P > 0.04). Persons who were older (ß = -0.08, P < 0.001), with longer axial length (ß = -0.32, P = 0.005), presence of macular diffuse choroidal atrophy (ß = -2.16, P < 0.001) or worse MMD (ß = -5.70, P < 0.001), and presence of macular posterior staphyloma (ß ≤ -2.98, P < 0.001) or Fuchs spot (ß = -1.58, P = 0.04) were associated with reduced macular sensitivity. Macular sensitivity was significantly associated with deep retinal PD in MMD (ß = 0.15, P = 0.004) but not with superficial retinal PD (P = 0.62). CONCLUSION: There was a strong correlation between reduced macular sensitivity and increasing MMD severity, even in mild MMD independent of the best-corrected visual acuity. Furthermore, macular sensitivity was correlated with deep retinal PD, suggesting a vasculature-function relationship in MMD.

Quantitative OCT angiography of the retinal microvasculature and choriocapillaris in highly myopic eyes with myopic macular degeneration.

PURPOSE: To quantify retinal and choriocapillaris (CC) microvasculature in highly myopic (HM) eyes with myopic macular degeneration (MMD) using swept-source optical coherence tomography angiography (SS-OCTA). METHODS: 162 HM eyes (spherical equivalent ≤ -6.0 dioptres or axial length (AL) ≥26.5 mm) from 98 participants were enrolled, including 60 eyes (37.0%) with tessellated fundus, 54 eyes (33.3%) with peripapillary diffuse chorioretinal atrophy (PDCA), 27 eyes (16.7%) with macular diffuse chorioretinal atrophy (MDCA) and 21 eyes (13.0%) with patchy or macular atrophy. PLEX Elite 9000 SS-OCTA was performed to obtain perfusion densities (PD) of the superficial and deep retinal capillary plexus, and CC signal voids (number, area and density). RESULTS: Retinal PD decreased with increasing severity of MMD. Multivariable analysis showed that after adjustment of age and other factors, retinal PD decreased significantly in eyes with longer AL (ß≤-0.51, p<0.001) and with an MMD severity of MDCA or worse (ß≤-1.63, p<0.001). Reduced retinal PD were significantly associated with worse vision (ß≤-0.01, p≤0.04). In terms of CC signal voids, multivariable analysis showed that longer AL (p<0.001), but not MMD severity (p≥0.12) was significantly associated with CC signal void changes in the earliest stage of MMD. CONCLUSION: We demonstrate significant OCTA alterations in the retina and CC in HM eyes with varying severities of MMD. In eyes with early-stage PDCA, lower retinal PD and more extensive CC signal voids are predominantly associated with increasing AL. In contrast, in eyes with MDCA or worse, MMD itself was associated with sparser retinal and CC circulation.

Updates in deep learning research in ophthalmology.

Ophthalmology has been one of the early adopters of artificial intelligence (AI) within the medical field. Deep learning (DL), in particular, has garnered significant attention due to the availability of large amounts of data and digitized ocular images. Currently, AI in Ophthalmology is mainly focused on improving disease classification and supporting decision-making when treating ophthalmic diseases such as diabetic retinopathy, age-related macular degeneration (AMD), glaucoma and retinopathy of prematurity (ROP). However, most of the DL systems (DLSs) developed thus far remain in the research stage and only a handful are able to achieve clinical translation. This phenomenon is due to a combination of factors including concerns over security and privacy, poor generalizability, trust and explainability issues, unfavorable end-user perceptions and uncertain economic value. Overcoming this challenge would require a combination approach. Firstly, emerging techniques such as federated learning (FL), generative adversarial networks (GANs), autonomous AI and blockchain will be playing an increasingly critical role to enhance privacy, collaboration and DLS performance. Next, compliance to reporting and regulatory guidelines, such as CONSORT-AI and STARD-AI, will be required to in order to improve transparency, minimize abuse and ensure reproducibility. Thirdly, frameworks will be required to obtain patient consent, perform ethical assessment and evaluate end-user perception. Lastly, proper health economic assessment (HEA) must be performed to provide financial visibility during the early phases of DLS development. This is necessary to manage resources prudently and guide the development of DLS.

Ultrawide field, distortion-corrected ocular shape estimation with MHz optical coherence tomography (OCT).

Ocular deformation may be associated with biomechanical alterations in the structures of the eye, especially the cornea and sclera in conditions such as keratoconus, congenital glaucoma, and pathological myopia. Here, we propose a method to estimate ocular shape using an ultra-wide field MHz swept-source optical coherence tomography (SS-OCT) with a Fourier Domain Mode-Locked (FDML) laser and distortion correction of the images. The ocular biometrics for distortion correction was collected by an IOLMaster 700, and localized Gaussian curvature was proposed to quantify the ocular curvature covering a field-of-view up to 65°×62°. We achieved repeatable curvature shape measurements (intraclass coefficient = 0.88 ± 0.06) and demonstrated its applicability in a pilot study with individuals (N = 11) with various degrees of myopia.

Prevalence, risk factors and impact of posterior staphyloma diagnosed from wide-field optical coherence tomography in Singapore adults with high myopia.

PURPOSE: To investigate the prevalence and risk factors of posterior staphyloma using wide-field optical coherence tomography (WF-OCT) in adults with high myopia in Singapore. DESIGN: Population-based cross-sectional study. METHODS: Adults with spherical equivalent (SE) ≤ -5D in either eye at the first visit of Singapore Epidemiology of Eye Diseases study and Singapore Prospective Study Program study were recruited. Posterior staphyloma was diagnosed using WF-OCT (PLEX® Elite9000, Carl Zeiss Meditec). Myopic macular degeneration (MMD), myopic traction maculopathy (MTM) and vision-related quality of life (VRQoL) were assessed using fundus photographs, DRI-Triton OCT (Topcon) and the Impact of Vision Impairment (IVI) questionnaire, respectively. Factors associated with posterior staphyloma were identified with multilevel, multivariable logistic regression. Impact of posterior staphyloma on MMD, MTM and visual function was analysed with multilevel, multivariable logistic regression and linear mixed model, respectively. RESULTS: Among the 225 eyes [mean SE = -6.5 ± 2.2 D, mean axial length (AL) = 26.2 ± 1.5 mm] of 117 participants (mean age = 60.3 ± 7.1 years), posterior staphyloma was detected in 47 (20.9%) eyes of 38 (32.5%) participants. Older age [odds ratio (OR), 1.18; 95% confidence interval (CI), 1.10-1.26], more myopic SE (0.63; 0.51-0.77) and increased AL (2.51; 1.69-3.73) were associated with higher prevalence of posterior staphyloma (all p < 0.001). Adults with posterior staphyloma had higher odds of MMD (2.67; 1.23-5.82; p = 0.013), MTM (3.79; 1.13-12.68; p = 0.031) and worse IVI Reading (ß = -1.44; -2.31 to 0.58; p = 0.001) scores. CONCLUSIONS: About one in three adults with high myopia had posterior staphyloma, which was associated with increased odds of having myopic maculopathy and a detrimental impact on VRQoL.

Diagnostic criteria for detection of retinal nerve fibre layer thickness and neuroretinal rim width abnormalities in glaucoma.

BACKGROUND/AIMS: Although measurements of the Bruch's membrane opening minimum rim width (BMO-MRW) and retinal nerve fibre layer thickness (RNFLT) with optical coherence tomography (OCT) have been widely adopted in the diagnostic evaluation of glaucoma, there is no consensus on the diagnostic criteria to define BMO-MRW and RNFLT abnormalities. This study investigated the sensitivities and specificities of different diagnostic criteria based on the OCT classification reports for detection of glaucoma. METHODS: 340 eyes of 137 patients with glaucoma and 87 healthy individuals, all with axial length ≤26mm, had global and sectoral BMO-MRW and RNFLT measured with Spectralis OCT (Heidelberg Engineering). Six diagnostic criteria were examined: global measurement below the fifth or the first percentile; ≥1 sector measurement below the fifth or the first percentile; superotemporal and/or inferotemporal measurement below the fifth or the first percentile. The sensitivities and specificities of BMO-MRW/RNFLT assessment for detection of glaucoma (eyes with visual field (VF) defects) were compared. RESULTS: Among the six criteria examined, superotemporal and/or inferotemporal measurement below the fifth percentile showed the highest sensitivities and specificities for glaucoma detection. Abnormal superotemporal and/or inferotemporal RNFLT attained a higher sensitivity than abnormal superotemporal and/or inferotemporal BMO-MRW to detect mild glaucoma (mean VF MD: -3.32±1.59 dB) (97.9% and 88.4%, respectively, p=0.006), and glaucoma (mean VF MD: -9.36±8.31 dB) (98.4% and 93.6%, respectively, p=0.006), at the same specificity (96.1%). CONCLUSIONS: Superotemporal and/or inferotemporal RNFLT/MRW below the fifth percentile yield the best diagnostic performance for glaucoma detection with RNFLT attains higher sensitivities than MRW at the same specificity in eyes without high myopia.

Detection of Bruch's Membrane Opening in Healthy Individuals and Glaucoma Patients with and without High Myopia.

PURPOSE: To determine the proportions of eyes with indiscernible Bruch's membrane opening (BMO) in glaucoma patients and healthy individuals with and without high myopia and to investigate factors contributing to indiscernible BMO. DESIGN: Cross-sectional study. PARTICIPANTS: Five hundred eyes from 315 participants, including 212 high myopic eyes (axial length [AL], ≥26 mm) from 80 glaucoma patients and 60 healthy individuals and 288 non-high myopic eyes (AL, <26 mm) from 96 glaucoma patients and 88 healthy individuals. METHODS: The optic disc was imaged by the Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany) using 24 equally spaced radial B-scans. The BMO was located independently by 2 trained observers and was recorded as discernible or indiscernible over 48 meridians in each eye. The BMO of a meridian was determined to be indiscernible when both observers failed to identify its location. The agreement between the observers was evaluated with κ statistics. Factors associated with indiscernible BMO were identified with multivariable, multilevel logistic regression modeling. MAIN OUTCOME MEASURES: Proportions of eyes with indiscernible BMO and hazard ratios (HRs) of factors associated with indiscernible BMO. RESULTS: The agreement for assessment of BMO visibility between the observers was good (κ, 0.631; 95% confidence interval, 0.602-0.661). In the groups with and without high myopia, 32.1% and 8.2% of glaucomatous eyes had 1 or more meridians with indiscernible BMO, respectively. In the healthy eyes, the respective proportions were 28.0% and 3.9%. The proportions were significantly higher in eyes with high myopia compared to eyes without high myopia in the glaucoma group (P < 0.001) and the healthy group (P < 0.001). The temporal meridian, followed by the inferotemporal and superotemporal meridians, of the optic disc were the most frequent locations with indiscernible BMO. Increased AL, advanced glaucoma, ß parapapillary atrophy, and young age were associated with an increased HR of indiscernible BMO (P ≤ 0.032). CONCLUSIONS: A significant proportion of high myopic eyes have indiscernible BMO at the temporal, superotemporal, and inferotemporal meridians of the optic disc, which may compromise the measurement of neuroretinal rim in the diagnostic evaluation of glaucoma.

Up-Regulation of ENO1 by HIF-1α in Retinal Pigment Epithelial Cells after Hypoxic Challenge Is Not Involved in the Regulation of VEGF Secretion.

PURPOSE: Alpha-enolase (ENO1), a major glycolytic enzyme, is reported to be over-expressed in various cancer tissues. It has been demonstrated to be regulated by the Hypoxia-inducible factor 1-α (HIF-1α), a crucial transcriptional factor implicated in tumor progression and cancer angiogenesis. Choroidal neovascularization (CNV), which is a leading cause of severe vision loss caused by newly formed blood vessels in the choroid, is also engendered by hypoxic stress. In this report, we investigated the expression of ENO1 and the effects of its down-regulation upon cobalt (II) chloride-induced hypoxia in retinal pigment epithelial cells, identified as the primary source of ocular angiogenic factors. METHODS: HIF-1α-diminished retinal pigment epithelial cells were generated by small interfering RNA (siRNA) technology in ARPE-19 cells, a human retinal pigment epithelial cell line. Both normal and HIF-1α-diminished ARPE-19 cells were then subjected to hypoxic challenge using cobalt (II) chloride (CoCl2) or anaerobic chamber. The relation between ENO1 expression and vascular endothelial growth factor (VEGF) secretion by retinal pigment epithelial cells were examined. Protein levels of HIF-1α and ENO1 were analyzed using Western Blot, while VEGF secretion was essayed by enzyme-linked immunosorbent assay (ELISA). Cytotoxicity after hypoxia was detected by Lactate Dehydrogenase (LDH) Assay. RESULTS: Upon 24 hr of CoCl2-induced hypoxia, the expression levels of ENO1 and VEGF were increased along with HIF-1α in ARPE-19 cells, both of which can in turn be down-regulated by HIF-1α siRNA application. However, knockdown of ENO1 alone or together with HIF-1α did not help suppress VEGF secretion in hypoxic ARPE-19 cells. CONCLUSION: ENO1 was demonstrated to be up-regulated by HIF-1α in retinal pigment epithelial cells in response to hypoxia, without influencing VEGF secretion.