Foundation models in ophthalmology.

Foundation models represent a paradigm shift in artificial intelligence (AI), evolving from narrow models designed for specific tasks to versatile, generalisable models adaptable to a myriad of diverse applications. Ophthalmology as a specialty has the potential to act as an exemplar for other medical specialties, offering a blueprint for integrating foundation models broadly into clinical practice. This review hopes to serve as a roadmap for eyecare professionals seeking to better understand foundation models, while equipping readers with the tools to explore the use of foundation models in their own research and practice. We begin by outlining the key concepts and technological advances which have enabled the development of these models, providing an overview of novel training approaches and modern AI architectures. Next, we summarise existing literature on the topic of foundation models in ophthalmology, encompassing progress in vision foundation models, large language models and large multimodal models. Finally, we outline major challenges relating to privacy, bias and clinical validation, and propose key steps forward to maximise the benefit of this powerful technology.

Ocular biomarkers: useful incidental findings by deep learning algorithms in fundus photographs.

BACKGROUND/OBJECTIVES: Artificial intelligence can assist with ocular image analysis for screening and diagnosis, but it is not yet capable of autonomous full-spectrum screening. Hypothetically, false-positive results may have unrealized screening potential arising from signals persisting despite training and/or ambiguous signals such as from biomarker overlap or high comorbidity. The study aimed to explore the potential to detect clinically useful incidental ocular biomarkers by screening fundus photographs of hypertensive adults using diabetic deep learning algorithms. SUBJECTS/METHODS: Patients referred for treatment-resistant hypertension were imaged at a hospital unit in Perth, Australia, between 2016 and 2022. The same 45° colour fundus photograph selected for each of the 433 participants imaged was processed by three deep learning algorithms. Two expert retinal specialists graded all false-positive results for diabetic retinopathy in non-diabetic participants. RESULTS: Of the 29 non-diabetic participants misclassified as positive for diabetic retinopathy, 28 (97%) had clinically useful retinal biomarkers. The models designed to screen for fewer diseases captured more incidental disease. All three algorithms showed a positive correlation between severity of hypertensive retinopathy and misclassified diabetic retinopathy. CONCLUSIONS: The results suggest that diabetic deep learning models may be responsive to hypertensive and other clinically useful retinal biomarkers within an at-risk, hypertensive cohort. Observing that models trained for fewer diseases captured more incidental pathology increases confidence in signalling hypotheses aligned with using self-supervised learning to develop autonomous comprehensive screening. Meanwhile, non-referable and false-positive outputs of other deep learning screening models could be explored for immediate clinical use in other populations.

Ophthalmic Findings in Aboriginal Children with High Rates of Prenatal Alcohol Exposure and Fetal Alcohol Spectrum Disorder: The Lililwan Project.

PURPOSE: To describe ophthalmic findings in an Indigenous paediatric population and the associations between fetal alcohol spectrum disorder (FASD), prenatal alcohol exposure (PAE), and eye anomalies. METHODS: Medical records were reviewed for eye problems, and eye assessments were conducted by an orthoptist or ophthalmologist in the Lililwan Project cohort, which comprised 108 (81%) of all children born between 2002 and 2003, and residing in the remote Fitzroy Valley, Western Australia in 2010. Values from ophthalmic assessments and prevalence of abnormalities were presented for the total cohort and stratified by group: FASD; PAE (no FASD); and No PAE. RESULTS: Of children, 55% had PAE and 19% FASD. Most (98%) had normal vision; 15.6% had keratometry cylinder values indicating astigmatism and potential for improved vision with glasses. Strabismus (22.3%), short palpebral fissure length (PFL; 21.3%), upslanting palpebral fissures (12.0%), follicular trachomatous inflammation (6.9%), abnormal slit lamp assessments (6.7%), retinal tortuosity (6.7%), and blepharoptosis (5.6%) were identified. Strabismus and trachoma rates were higher than in the general child population. Ophthalmic findings were similar between groups except for prevalence of short PFL (FASD > No PAE; p = 0.003); abnormal keratometry cylinder values (FASD [29.4%] and PAE (no FASD) [20.0%] > No PAE [5.3%]; p = 0.031) and blepharoptosis (FASD [9.5%] > other groups [0%]; p = 0.040). CONCLUSION: Despite the small sample, some eye abnormalities were higher in children with PAE and/or FASD. Access to eye services or assessment of vision and structural eye anomalies is essential for Indigenous children, particularly those with PAE or FASD to allow early effective treatment.

A Systematic Review of Ophthalmology Education in Medical Schools: The Global Decline.

TOPIC: This systematic review examined geographical and temporal trends in medical school ophthalmology education in relationship to course and student outcomes. CLINICAL RELEVANCE: Evidence suggesting a decline in ophthalmology teaching in medical schools is increasing, raising concern for the adequacy of eye knowledge across the rest of the medical profession. METHODS: Systematic review of Embase and SCOPUS, with inclusion of studies containing data on medical school ophthalmic course length; 1 or more outcome measures on student ophthalmology knowledge, skills, self-evaluation of knowledge or skills, or student course appraisal; or both. The systematic review was registered prospectively on the International Prospective Register of Systematic Reviews (identifier, CRD42022323865). Results were aggregated with outcome subgroup analysis and description in relationship to geographical and temporal trends. Descriptive statistics, including nonparametric correlations, were used to analyze data and trends. RESULTS: Systematic review yielded 4596 publication titles, of which 52 were included in the analysis, with data from 19 countries. Average course length ranged from 12.5 to 208.7 hours, with significant continental disparity among mean course lengths. Africa reported the longest average course length at 103.3 hours, and North America reported the shortest at 36.4 hours. On average, course lengths have been declining over the last 2 decades, from an average overall course length of 92.9 hours in the 2000s to 52.9 hours in the 2020s. Mean student self-evaluation of skills was 51.3%, and mean student self-evaluation of knowledge was 55.4%. Objective mean assessment mark of skills was 57.5% and that of knowledge was 71.7%, compared with an average pass mark of 66.7%. On average, 26.4% of students felt confident in their ophthalmology knowledge and 34.5% felt confident in their skills. DISCUSSION: Most evidence describes declining length of courses devoted to ophthalmology in the last 20 years, significant student dissatisfaction with courses and content, and suboptimal knowledge and confidence. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Validation of a deep learning system for the detection of diabetic retinopathy in Indigenous Australians.

BACKGROUND/AIMS: Deep learning systems (DLSs) for diabetic retinopathy (DR) detection show promising results but can underperform in racial and ethnic minority groups, therefore external validation within these populations is critical for health equity. This study evaluates the performance of a DLS for DR detection among Indigenous Australians, an understudied ethnic group who suffer disproportionately from DR-related blindness. METHODS: We performed a retrospective external validation study comparing the performance of a DLS against a retinal specialist for the detection of more-than-mild DR (mtmDR), vision-threatening DR (vtDR) and all-cause referable DR. The validation set consisted of 1682 consecutive, single-field, macula-centred retinal photographs from 864 patients with diabetes (mean age 54.9 years, 52.4% women) at an Indigenous primary care service in Perth, Australia. Three-person adjudication by a panel of specialists served as the reference standard. RESULTS: For mtmDR detection, sensitivity of the DLS was superior to the retina specialist (98.0% (95% CI, 96.5 to 99.4) vs 87.1% (95% CI, 83.6 to 90.6), McNemar's test p<0.001) with a small reduction in specificity (95.1% (95% CI, 93.6 to 96.4) vs 97.0% (95% CI, 95.9 to 98.0), p=0.006). For vtDR, the DLS's sensitivity was again superior to the human grader (96.2% (95% CI, 93.4 to 98.6) vs 84.4% (95% CI, 79.7 to 89.2), p<0.001) with a slight drop in specificity (95.8% (95% CI, 94.6 to 96.9) vs 97.8% (95% CI, 96.9 to 98.6), p=0.002). For all-cause referable DR, there was a substantial increase in sensitivity (93.7% (95% CI, 91.8 to 95.5) vs 74.4% (95% CI, 71.1 to 77.5), p<0.001) and a smaller reduction in specificity (91.7% (95% CI, 90.0 to 93.3) vs 96.3% (95% CI, 95.2 to 97.4), p<0.001). CONCLUSION: The DLS showed improved sensitivity and similar specificity compared with a retina specialist for DR detection. This demonstrates its potential to support DR screening among Indigenous Australians, an underserved population with a high burden of diabetic eye disease.

Comparison of three modalities of teleophthalmology delivery in regional Western Australia during the COVID-19 lock-down.

CLINICAL RELEVANCE: In conjunction with local optometry services, telehealth may be used in to provide specialist care for patients living in rural areas underserved by ophthalmology. BACKGROUND: To combat travel restrictions for specialist outreach to regional areas during the 2020 COVID-19 lockdown, Lions Outback Vision introduced three different modalities of teleophthalmology consultations; home-based telephone, hospital-based video, and optometry-based video. This study evaluated the utility of these in providing specialist care to rural patients during the pandemic. METHODS: Data from patients referred during the COVID-19 lock-down period (23 March 2020 to 5 June 2020) were analysed. If sufficient clinical information and imaging were available then ophthalmologists conducted home-based telephone consultations. If further ocular imaging or examination was required, then optometry-based video or hospital-based video were used. Data were analysed using ANOVA and two-sided t tests for continuous data and Chi Square statistics for categorical data (p < 0.05). RESULTS: Majority of the 431 consultations were conducted via home telephone (38%) or optometry-based video (37%). Indigenous patients (p = 0.014) and patients in very remote communities (p < 0.01) were more likely to receive a home-based telephone consultation. Because sufficient clinical information had already been obtained for home-based consultations, these patients were more likely to be booked for surgery than optometry (p < 0.01).Cataracts were the predominant diagnosis in optometry consults compared to hospital (p < 0.01). CONCLUSION: Primary optometry and home telephone represent a new modality for providing specialist care for patients living in very remote regions and for Indigenous patients. When appropriate clinical testing has been completed, telephone-based ophthalmology may continue to be useful for certain conditions such as waitlisting patients for cataract surgery and should continue to be funded beyond the duration of the pandemic for rural patients.

A systematic review of diabetic retinopathy screening intervals.

The current evidence on whether annual diabetic retinopathy (DR) screening intervals can be extended was reviewed. A systematic review protocol was followed (PROSPERO ID: CRD42022359590). Original longitudinal articles that specifically assessed DR screening intervals were in English and collected data after 2000 were included. Two reviewers independently conducted the search and reviewed the articles for quality and relevant information. The heterogeneity of the data meant that a meta-analysis was not appropriate. Twelve publications were included. Studies were of good quality and many used data from DR screening programs. Studies fit into three categories; those that assessed specific DR screening intervals, those that determined optimal DR screening intervals and those that developed/assessed DR screening risk equations. For those with type 2 diabetes, extending screening intervals to 3- to 4-yearly in those with no baseline DR appeared safe. DR risk equations considered clinical factors and allocated those at lower risk of DR progression screening intervals of up to five years. Those with baseline DR or type 1 diabetes appeared to have a higher risk of progression to STDR and needed more frequent screening. DR screening intervals can be extended to 3-5 yearly in certain circumstances. These include patients with type 2 diabetes and no current DR, and those who have optimal management of other risk factors such as glucose and blood pressure.

Artificial intelligence and digital health in global eye health: opportunities and challenges.

Global eye health is defined as the degree to which vision, ocular health, and function are maximised worldwide, thereby optimising overall wellbeing and quality of life. Improving eye health is a global priority as a key to unlocking human potential by reducing the morbidity burden of disease, increasing productivity, and supporting access to education. Although extraordinary progress fuelled by global eye health initiatives has been made over the last decade, there remain substantial challenges impeding further progress. The accelerated development of digital health and artificial intelligence (AI) applications provides an opportunity to transform eye health, from facilitating and increasing access to eye care to supporting clinical decision making with an objective, data-driven approach. Here, we explore the opportunities and challenges presented by digital health and AI in global eye health and describe how these technologies could be leveraged to improve global eye health. AI, telehealth, and emerging technologies have great potential, but require specific work to overcome barriers to implementation. We suggest that a global digital eye health task force could facilitate coordination of funding, infrastructural development, and democratisation of AI and digital health to drive progress forwards in this domain.

Adult onset xanthogranuloma of the eyelid.

Purpose: To report a rare case of an eyelid lesion in an adult, with histological features of juvenile xanthogranuloma (JXG). Observations: Juvenile xanthogranuloma primarily affects the skin of infants and young children. It infrequently can involve the structures of the eye and orbit and rarely occurs in individuals beyond the second decade of life. We present a case of adult onset xanthogranuloma (AXG) involving the eyelid of a 29-year-old female. This lesion required management with multiple treatment modalities. Conclusions: This is a rare example of an eyelid xanthogranuloma in an adult. As such, JXG-like lesions should be included as a differential diagnosis for lesions of the eye and orbit in adults. Surgical management may be required if there is no response to intralesional steroids.

Retinopathy prevalence, incidence and trajectories in type 2 diabetes: The Fremantle diabetes study phase II.

AIMS: To determine diabetic retinopathy (DR) prevalence, incidence, and whether distinct trajectories are associated with DR-complicating Type 2 diabetes. METHODS: Retinal photographs from Fremantle Diabetes Study Phase II (FDS2) participants with Type 2 diabetes recruited in 2008-2011 and who attended biennial assessments for up to 6 years were graded as no DR, mild non-proliferative DR (NPDR), moderate NPDR or severe NPDR/proliferative DR. Baseline DR prevalence, and the cumulative incidence of moderate NPDR or worse in those without DR at baseline, were calculated. Group-based DR trajectory modelling was performed. Logistic regression determined independent associates of incident moderate NPDR or worse and trajectory group membership. RESULTS: Of 1521 participants (mean age 65.6 years, 52.1% males, median diabetes duration 9.0 years; 98% of all FDS2 participants with Type 2 diabetes) with gradable baseline photographs, 563 (37.0%) had DR. During a median 6.1 years of follow-up, 23 (3.2%) without baseline DR developed at least moderate NPDR (crude incidence 6.1/1000 person-years) with HbA1c the sole independent predictor (odds ratio [95% CI]: 1.62 [1.30-2.02] per 1% [11 mmol/mol] increase). Trajectory analysis showed two distinct groups, those with baseline/persistent DR (20%) and those remaining DR free (80%). Longer diabetes duration, insulin use, higher mean HbA1c , higher mean systolic blood pressure and higher mean urinary albumin: creatinine ratio all increased the odds (p ≤ 0.014) of being in the persistent DR trajectory group. CONCLUSIONS: The low incidence of at least moderate NPDR reflects the trajectory analysis. The currently recommended biennial retinal screening frequency for individuals without DR could potentially be extended.

Prevalence of Diabetic Retinopathy in Indigenous and Non-Indigenous Australians: A Systematic Review and Meta-analysis.

TOPIC: This systematic review and meta-analysis summarizes evidence relating to the prevalence of diabetic retinopathy (DR) among Indigenous and non-Indigenous Australians. CLINICAL RELEVANCE: Indigenous Australians suffer disproportionately from diabetes-related complications. Exploring ethnic variation in disease is important for equitable distribution of resources and may lead to identification of ethnic-specific modifiable risk factors. Existing DR prevalence studies comparing Indigenous and non-Indigenous Australians have shown conflicting results. METHODS: This study was conducted following Joanna Briggs Institute guidance on systematic reviews of prevalence studies (PROSPERO ID: CRD42022259048). We performed searches of Medline (Ovid), EMBASE, and Web of Science until October 2021, using a strategy designed by an information specialist. We included studies reporting DR prevalence among diabetic patients in Indigenous and non-Indigenous Australian populations. Two independent reviewers performed quality assessments using a 9-item appraisal tool. Meta-analysis and meta-regression were performed using double arcsine transformation and a random-effects model comparing Indigenous and non-Indigenous subgroups. RESULTS: Fifteen studies with 8219 participants met criteria for inclusion. The Indigenous subgroup scored lower on the appraisal tool than the non-Indigenous subgroup (mean score 50% vs. 72%, P = 0.04). In the unadjusted meta-analysis, DR prevalence in the Indigenous subgroup (30.2%; 95% confidence interval [CI], 24.9-35.7) did not differ significantly (P = 0.17) from the non-Indigenous subgroup (23.7%; 95% CI, 16.8-31.4). After adjusting for age and quality, DR prevalence was higher in the Indigenous subgroup (P < 0.01), with prevalence ratio point estimates ranging from 1.72 to 2.58, depending on the meta-regression model. For the secondary outcomes, prevalence estimates were higher in the Indigenous subgroup for diabetic macular edema (DME) (8.7% vs. 2.7%, P = 0.02) and vision-threatening DR (VTDR) (8.6% vs. 3.0%, P = 0.03) but not for proliferative DR (2.5% vs. 0.8%, P = 0.07). CONCLUSIONS: Indigenous studies scored lower for methodological quality, raising the possibility that systematic differences in research practices may be leading to underestimation of disease burden. After adjusting for age and quality, we found a higher DR prevalence in the Indigenous subgroup. This contrasts with a previous review that reported the opposite finding of lower DR prevalence using unadjusted pooled estimates. Future epidemiological work exploring DR burden in Indigenous communities should aim to address methodological weaknesses identified by this review.

Telemedicine and delivery of ophthalmic care in rural and remote communities: Drawing from Australian experience.

Rural and remote communities in Australia are characterised by small but widely dispersed populations. This has been proven to be a major hurdle in access to medical care services with screening and treatment goals repeatedly being missed. Telemedicine in ophthalmology provides the opportunity to increase the availability of high quality and timely access to healthcare within. Recent years has also seen the introduction of artificial intelligence (AI) in ophthalmology, particularly in the screening of diseases. AI will hopefully increase the number of appropriate referrals, reduce travel time for patients and ensure timely triage given the low number of qualified optometrists and ophthalmologists. Telemedicine and AI has been introduced in a number of countries and has led to tremendous benefits and advantages when compared to standard practices. This paper summarises current practices in telemedicine and AI and the future of this technology in improving patient care in the field of ophthalmology.

Antibiotic prophylaxis for corneal abrasion.

BACKGROUND: Corneal abrasion is a common disorder frequently faced by ophthalmologists, emergency physicians, and primary care physicians. Ocular antibiotics are one of the management options for corneal abrasion. A comprehensive summary and synthesis of the evidence on antibiotic prophylaxis in traumatic corneal abrasion is thus far unavailable, therefore we conducted this review to evaluate the current evidence regarding this important issue. OBJECTIVES: To assess the safety and efficacy of topical antibiotic prophylaxis following corneal abrasion. Our objectives were 1) to investigate the incidence of infection with antibiotics versus placebo or alternative antibiotics in people with corneal abrasion; and 2) to investigate time to clinical cure, defined as complete healing (re-epithelialization) of the epithelium, with antibiotics versus placebo or alternative antibiotics in people with corneal abrasion. SEARCH METHODS: We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (which contains the Cochrane Eyes and Vision Trials Register; 2021, Issue 4), Ovid MEDLINE, Embase.com, PubMed, the Latin American and Caribbean Health Sciences Literature database (LILACS), ClinicalTrials.gov, and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP). We did not use any date or language restrictions in the electronic search for trials. We last searched the electronic databases on 25 April 2021. SELECTION CRITERIA: We included randomized controlled trials (RCTs) comparing antibiotic with another antibiotic or placebo in children and adults with corneal abrasion due to any cause. DATA COLLECTION AND ANALYSIS: We used standard Cochrane methodology and assessed the certainty of the body of evidence for the prespecified outcomes using the GRADE classification. MAIN RESULTS: Our search of the electronic databases yielded 8661 records. We screened 7690 titles and abstracts after removal of duplicates. We retrieved 32 full-text reports for further review. We included two studies that randomized a total of 527 eyes of 527 participants in the review. One study was conducted in Denmark, and one was conducted in India. The two studies did not examine most of our prespecified primary and secondary outcomes. The first study was a parallel-group RCT comparing chloramphenicol ocular ointment with fusidic acid ocular gels (frequency was not clearly reported). This study enrolled 153 participants older than 5 years of age with corneal abrasion in Denmark with a one-day follow-up duration. No participants had secondary infection in the fusidic acid group, whereas three (4.1%) participants in the chloramphenicol group had a slight reaction (risk ratio [RR] 0.15, 95% confidence interval [CI] 0.01 to 2.79; 144 participants; very low certainty evidence). Thirty-one (44.3%) participants in the fusidic acid arm and 34 (46.6%) participants in the chloramphenicol arm were cured (defined as the area of abrasion zero and no infection) at day 1 (RR 0.94, 95% CI 0.65 to 1.34; 144 participants; very low certainty evidence). Without providing specific data, the study reported that the degree of pain was not affected by the interventions received. The most common adverse events reported were itching and discomfort of the eye, which occurred in approximately one-third of participants in each group (low certainty evidence). A second multicenter, two-arm RCT conducted in India enrolled 374 participants older than 5 years of age with corneal abrasion who presented within 48 hours after injury. This study investigated the effect of a three-day course of either ocular ointment combinations of chloramphenicol-clotrimazole or chloramphenicol-placebo (all three times daily). At day 3, 169 (100%) participants in the chloramphenicol-clotrimazole arm and 203 (99%) out of 205 participants in the chloramphenicol-placebo arm were cured without any complication, defined as complete epithelialization of the cornea without evidence of infection (RR 1.01, 95% CI 0.99 to 1.03; 374 participants; very low certainty evidence). Four participants assigned to the chloramphenicol-placebo arm experienced mild adverse events: two participants (1%) had mild chemosis and irritation, and two (1%) had small single sterile corneal infiltrates (low certainty evidence). AUTHORS' CONCLUSIONS: Given the low to very low certainty of the available evidence, any beneficial effects of antibiotic prophylaxis in preventing ocular infection or accelerating epithelial healing following a corneal abrasion remain unclear. Moreover, the current evidence is insufficient to support any antibiotic regimen being superior to another. There is a need for a well-designed RCT assessing the efficacy and safety of ocular antibiotics in the treatment of corneal abrasion with a particular focus on high-risk populations and formulation of interventions.

Benefits of Integrating Telemedicine and Artificial Intelligence Into Outreach Eye Care: Stepwise Approach and Future Directions.

Telemedicine has traditionally been applied within remote settings to overcome geographical barriers to healthcare access, providing an alternate means of connecting patients to specialist services. The coronavirus 2019 pandemic has rapidly expanded the use of telemedicine into metropolitan areas and enhanced global telemedicine capabilities. Through our experience of delivering real-time telemedicine over the past decade within a large outreach eye service, we have identified key themes for successful implementation which may be relevant to services facing common challenges. We present our journey toward establishing a comprehensive teleophthalmology model built on the principles of collaborative care, with a focus on delivering practical lessons for service design. Artificial intelligence is an emerging technology that has shown potential to further address resource limitations. We explore the applications of artificial intelligence and the need for targeted research within underserved settings in order to meet growing healthcare demands. Based on our rural telemedicine experience, we make the case that similar models may be adapted to urban settings with the aim of reducing surgical waitlists and improving efficiency.

Intravitreal dexamethasone versus bevacizumab in Aboriginal and Torres Strait Islander patients with diabetic macular oedema: The OASIS study (a randomised control trial).

BACKGROUND: Frequent intravitreal anti-VEGF injections are impractical for many Aboriginal patients with diabetic macular oedema (DMO). The longer acting intravitreal dexamethasone implant (DEX-implant) is approved for DMO but has not been assessed in an Aboriginal population. METHODS: This was a prospective, multicentre, randomised, single-masked, non-inferiority clinical trial. Aboriginal adults from Western Australia with DMO were randomised to receive 3-monthly DEX-implant, or monthly intravitreal bevacizumab. The primary outcome was the change in best corrected visual acuity (BCVA) at 12 months. RESULTS: The final endpoint was analysed for 24 DEX-implant and 28 bevacizumab injection eyes. Mean BCVA improved by 4.0 letters (-0.08 LogMAR) in the DEX-implant group and worsened by 5.5 letters (0.11 LogMAR) in the bevacizumab group. Before adjusting for cataract surgery, the upper bound of the two-sided 90% CI for the DEX-implant was 3.5 letters (0.07 LogMAR), which met non-inferiority criteria. The BCVA of remote participants who received the DEX-implant improved by 5.5 letters (0.11 LogMAR), compared to an 18.5 letter (0.37 LogMAR) decline for bevacizumab (p = 0.04). The incidence of steroid-induced ocular hypertension for the DEX-implant was 33.3%. CONCLUSIONS: Before adjusting for the effect of cataract surgery, the DEX-implant was non-inferior to bevacizumab for treating DMO in Aboriginal participants. In remote participants, the DEX-implant surpassed non-inferiority to achieve superior outcomes to bevacizumab. The incidence of steroid-induced hypertension was comparable to that reported in non-Aboriginal populations. We provide guidelines for the judicious use of DEX-implant among Aboriginal people, and a framework for performing ophthalmic clinical trials in Aboriginal communities.

Deep Learning to Detect OCT-derived Diabetic Macular Edema from Color Retinal Photographs: A Multicenter Validation Study.

PURPOSE: To validate the generalizability of a deep learning system (DLS) that detects diabetic macular edema (DME) from 2-dimensional color fundus photographs (CFP), for which the reference standard for retinal thickness and fluid presence is derived from 3-dimensional OCT. DESIGN: Retrospective validation of a DLS across international datasets. PARTICIPANTS: Paired CFP and OCT of patients from diabetic retinopathy (DR) screening programs or retina clinics. The DLS was developed using data sets from Thailand, the United Kingdom, and the United States and validated using 3060 unique eyes from 1582 patients across screening populations in Australia, India, and Thailand. The DLS was separately validated in 698 eyes from 537 screened patients in the United Kingdom with mild DR and suspicion of DME based on CFP. METHODS: The DLS was trained using DME labels from OCT. The presence of DME was based on retinal thickening or intraretinal fluid. The DLS's performance was compared with expert grades of maculopathy and to a previous proof-of-concept version of the DLS. We further simulated the integration of the current DLS into an algorithm trained to detect DR from CFP. MAIN OUTCOME MEASURES: The superiority of specificity and noninferiority of sensitivity of the DLS for the detection of center-involving DME, using device-specific thresholds, compared with experts. RESULTS: The primary analysis in a combined data set spanning Australia, India, and Thailand showed the DLS had 80% specificity and 81% sensitivity, compared with expert graders, who had 59% specificity and 70% sensitivity. Relative to human experts, the DLS had significantly higher specificity (P = 0.008) and noninferior sensitivity (P < 0.001). In the data set from the United Kingdom, the DLS had a specificity of 80% (P < 0.001 for specificity of >50%) and a sensitivity of 100% (P = 0.02 for sensitivity of > 90%). CONCLUSIONS: The DLS can generalize to multiple international populations with an accuracy exceeding that of experts. The clinical value of this DLS to reduce false-positive referrals, thus decreasing the burden on specialist eye care, warrants a prospective evaluation.

Differences in retinopathy prevalence and progression between Anglo-Celt and Aboriginal Australians: the Fremantle Diabetes Study Phase II.

BACKGROUND: Indigenous populations have higher rates of diabetes and diabetic complications, yet there is a paucity of contemporary data on diabetic retinopathy (DR) prevalence and incidence in urban dwelling Aboriginal Australians. AIMS: The aim of the study was to compare the prevalence of DR and incidence of new or worsening DR between Aboriginal Australians and Anglo-Celts with Type 2 diabetes. METHODS: Participants from the community-based Fremantle Diabetes Study Phase II (817 Anglo-Celts, 94 Aboriginal people) recruited between 2008 and 2011 underwent fundus photography at baseline and biennial reviews. The prevalence of any DR and moderate non-proliferative DR (NPDR), and the incidence of new or worsening DR were ascertained using baseline and 4-year follow-up data. RESULTS: Compared with Anglo-Celts, the Aboriginal participants had a higher prevalence of any DR (33.0% vs 52.1%) and moderate NPDR or worse (5.1% vs 24.4%), and new or worsening DR during follow up (6.7% vs 23.5%). The unadjusted odds ratios (95% confidence interval) of any DR and moderate NPDR at baseline were 2.21 (1.43, 3.39) and 5.98 (3.40, 10.50), respectively, and of new or worsening DR 4.32 (1.33, 13.98). In adjusted models, Aboriginal ethnicity was only associated with the prevalence of moderate NPDR or worse (5.58 (2.44, 12.76)). CONCLUSIONS: Aboriginal participants had a higher prevalence of DR and new or worsening DR, reflecting conventional risk factors including suboptimal glycaemic control. Their significantly higher odds of moderate NPDR or worse in adjusted models suggest ethnic-specific determinants of DR severity. These findings highlight the need for equitable, culturally appropriate diabetes/ophthalmic care.

Real-world artificial intelligence-based opportunistic screening for diabetic retinopathy in endocrinology and indigenous healthcare settings in Australia.

This study investigated the diagnostic performance, feasibility, and end-user experiences of an artificial intelligence (AI)-assisted diabetic retinopathy (DR) screening model in real-world Australian healthcare settings. The study consisted of two components: (1) DR screening of patients using an AI-assisted system and (2) in-depth interviews with health professionals involved in implementing screening. Participants with type 1 or type 2 diabetes mellitus attending two endocrinology outpatient and three Aboriginal Medical Services clinics between March 2018 and May 2019 were invited to a prospective observational study. A single 45-degree (macula centred), non-stereoscopic, colour retinal image was taken of each eye from participants and were instantly screened for referable DR using a custom offline automated AI system. A total of 236 participants, including 174 from endocrinology and 62 from Aboriginal Medical Services clinics, provided informed consent and 203 (86.0%) were included in the analysis. A total of 33 consenting participants (14%) were excluded from the primary analysis due to ungradable or missing images from small pupils (n = 21, 63.6%), cataract (n = 7, 21.2%), poor fixation (n = 2, 6.1%), technical issues (n = 2, 6.1%), and corneal scarring (n = 1, 3%). The area under the curve, sensitivity, and specificity of the AI system for referable DR were 0.92, 96.9% and 87.7%, respectively. There were 51 disagreements between the reference standard and index test diagnoses, including 29 which were manually graded as ungradable, 21 false positives, and one false negative. A total of 28 participants (11.9%) were referred for follow-up based on new ocular findings, among whom, 15 (53.6%) were able to be contacted and 9 (60%) adhered to referral. Of 207 participants who completed a satisfaction questionnaire, 93.7% specified they were either satisfied or extremely satisfied, and 93.2% specified they would be likely or extremely likely to use this service again. Clinical staff involved in screening most frequently noted that the AI system was easy to use, and the real-time diagnostic report was useful. Our study indicates that AI-assisted DR screening model is accurate and well-accepted by patients and clinicians in endocrinology and indigenous healthcare settings. Future deployments of AI-assisted screening models would require consideration of downstream referral pathways.