Prediction of retinopathy progression using deep learning on retinal images within the Scottish screening programme.

BACKGROUND/AIMS: National guidelines of many countries set screening intervals for diabetic retinopathy (DR) based on grading of the last screening retinal images. We explore the potential of deep learning (DL) on images to predict progression to referable DR beyond DR grading, and the potential impact on assigned screening intervals, within the Scottish screening programme. METHODS: We consider 21 346 and 247 233 people with type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM), respectively, each contributing on average 4.8 and 4.4 screening intervals of which 1339 and 4675 intervals concluded with a referable screening episode. Information extracted from fundus images using DL was used to predict referable status at the end of interval and its predictive value in comparison to screening-assigned DR grade was assessed. RESULTS: The DL predictor increased the area under the receiver operating characteristic curve in comparison to a predictor using current DR grades from 0.809 to 0.87 for T1DM and from 0.825 to 0.87 for T2DM. Expected sojourn time-the time from becoming referable to being rescreened-was found to be 3.4 (T1DM) and 2.7 (T2DM) weeks less for a DL-derived policy compared with the current recall policy. CONCLUSIONS: We showed that, compared with using the current retinopathy grade, DL of fundus images significantly improves the prediction of incident referable retinopathy before the next screening episode. This can impact screening recall interval policy positively, for example, by reducing the expected time with referable disease for a fixed workload-which we show as an exemplar. Additionally, it could be used to optimise workload for a fixed sojourn time.

Deep learning detection of diabetic retinopathy in Scotland's diabetic eye screening programme.

BACKGROUND/AIMS: Support vector machine-based automated grading (known as iGradingM) has been shown to be safe, cost-effective and robust in the diabetic retinopathy (DR) screening (DES) programme in Scotland. It triages screening episodes as gradable with no DR versus manual grading required. The study aim was to develop a deep learning-based autograder using images and gradings from DES and to compare its performance with that of iGradingM. METHODS: Retinal images, quality assurance (QA) data and routine DR grades were obtained from national datasets in 179 944 patients for years 2006-2016. QA grades were available for 744 images. We developed a deep learning-based algorithm to detect whether either eye contained ungradable images or any DR. The sensitivity and specificity were evaluated against consensus QA grades and routine grades. RESULTS: Images used in QA which were ungradable or with DR were detected by deep learning with better specificity compared with manual graders (p<0.001) and with iGradingM (p<0.001) at the same sensitivities. Any DR according to the DES final grade was detected with 89.19% (270 392/303 154) sensitivity and 77.41% (500 945/647 158) specificity. Observable disease and referable disease were detected with sensitivities of 96.58% (16 613/17 201) and 98.48% (22 600/22 948), respectively. Overall, 43.84% of screening episodes would require manual grading. CONCLUSION: A deep learning-based system for DR grading was evaluated in QA data and images from 11 years in 50% of people attending a national DR screening programme. The system could reduce the manual grading workload at the same sensitivity compared with the current automated grading system.

Can deep learning on retinal images augment known risk factors for cardiovascular disease prediction in diabetes? A prospective cohort study from the national screening programme in Scotland.

AIMS: This study's objective was to evaluate whether deep learning (DL) on retinal photographs from a diabetic retinopathy screening programme improve prediction of incident cardiovascular disease (CVD). METHODS: DL models were trained to jointly predict future CVD risk and CVD risk factors and used to output a DL score. Poisson regression models including clinical risk factors with and without a DL score were fitted to study cohorts with 2,072 and 38,730 incident CVD events in type 1 (T1DM) and type 2 diabetes (T2DM) respectively. RESULTS: DL scores were independently associated with incident CVD with adjusted standardised incidence rate ratios of 1.14 (P = 3 × 10-04 95 % CI (1.06, 1.23)) and 1.16 (P = 4 × 10-33 95 % CI (1.13, 1.18)) in T1DM and T2DM cohorts respectively. The differences in predictive performance between models with and without a DL score were statistically significant (differences in test log-likelihood 6.7 and 51.1 natural log units) but the increments in C-statistics from 0.820 to 0.822 and from 0.709 to 0.711 for T1DM and T2DM respectively, were small. CONCLUSIONS: These results show that in people with diabetes, retinal photographs contain information on future CVD risk. However for this to contribute appreciably to clinical prediction of CVD further approaches, including exploitation of serial images, need to be evaluated.

Multimodal imaging interpreted by graders to detect re-activation of diabetic eye disease in previously treated patients: the EMERALD diagnostic accuracy study.

BACKGROUND: Owing to the increasing prevalence of diabetes, the workload related to diabetic macular oedema and proliferative diabetic retinopathy is rising, making it difficult for hospital eye services to meet demands. OBJECTIVE: The objective was to evaluate the diagnostic performance, cost-effectiveness and acceptability of a new pathway using multimodal imaging interpreted by ophthalmic graders to detect reactivation of diabetic macular oedema/proliferative diabetic retinopathy in previously treated patients. DESIGN: This was a prospective, case-referent, cross-sectional diagnostic study. SETTING: The setting was ophthalmic clinics in 13 NHS hospitals. PARTICIPANTS: Adults with type 1 or type 2 diabetes with previously successfully treated diabetic macular oedema/proliferative diabetic retinopathy in one/both eyes in whom, at the time of enrolment, diabetic macular oedema/proliferative diabetic retinopathy could be active or inactive. METHODS: For the ophthalmic grader pathway, review of the spectral domain optical coherence tomography scans to detect diabetic macular oedema, and seven-field Early Treatment Diabetic Retinopathy Study/ultra-wide field fundus images to detect proliferative diabetic retinopathy, by trained ophthalmic graders. For the current standard care pathway (reference standard), ophthalmologists examined patients face to face by slit-lamp biomicroscopy for proliferative diabetic retinopathy and, in addition, spectral domain optical coherence tomography imaging for diabetic macular oedema. OUTCOME MEASURES: The primary outcome measure was sensitivity of the ophthalmic grader pathway to detect active diabetic macular oedema/proliferative diabetic retinopathy. The secondary outcomes were specificity, agreement between pathways, cost-consequences, acceptability and the proportion of patients requiring subsequent ophthalmologist assessment, unable to undergo imaging and with inadequate quality images/indeterminate findings. It was assumed for the main analysis that all patients in whom graders diagnosed active disease or were 'unsure' or images were 'ungradable' required examination by an ophthalmologist. RESULTS: Eligible participants with active and inactive diabetic macular oedema (152 and 120 participants, respectively) and active and inactive proliferative diabetic retinopathy (111 and 170 participants, respectively) were recruited. Under the main analysis, graders had a sensitivity of 97% (142/147) (95% confidence interval 92% to 99%) and specificity of 31% (35/113) (95% confidence interval 23% to 40%) to detect diabetic macular oedema. For proliferative diabetic retinopathy, graders had a similar sensitivity and specificity using seven-field Early Treatment Diabetic Retinopathy Study [sensitivity 85% (87/102), 95% confidence interval 77% to 91%; specificity 48% (77/160), 95% confidence interval 41% to 56%] or ultra-wide field imaging [sensitivity 83% (87/105), 95% confidence interval 75% to 89%; specificity 54% (86/160), 95% confidence interval 46% to 61%]. Participants attending focus groups expressed preference for face-to-face evaluations by ophthalmologists. In the ophthalmologists' absence, patients voiced the need for immediate feedback following grader's assessments, maintaining periodic evaluations by ophthalmologists. Graders and ophthalmologists were supportive of the new pathway. When compared with the reference standard (current standard pathway), the new grader pathway could save £1390 per 100 patients in the review of people with diabetic macular oedema and, depending on the imaging modality used, between £461 and £1189 per 100 patients in the review of people with proliferative diabetic retinopathy. CONCLUSIONS: For people with diabetic macular oedema, the ophthalmic grader pathway appears safe and cost saving. The sensitivity of the new pathway to detect active proliferative diabetic retinopathy was lower, but may still be considered acceptable for patients with proliferative diabetic retinopathy previously treated with laser. Suggestions from focus group discussions should be taken into consideration if the new pathway is introduced to ensure its acceptability to users. LIMITATIONS: Lack of fundus fluorescein angiography to confirm diagnosis of active proliferative diabetic retinopathy. FUTURE WORK: Could refinement of the new pathway increase its sensitivity to detect proliferative diabetic retinopathy? Could artificial intelligence be used for automated reading of images in this previously treated population? TRIAL REGISTRATION: Current Controlled Trials ISRCTN10856638 and ClinicalTrials.gov NCT03490318. FUNDING: This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology AssessmentVol. 25, No. 32. See the NIHR Journals Library website for further project information.

More and more people are developing diabetes. Diabetic macular oedema and proliferative diabetic retinopathy are complications of diabetes, which could cause blindness. Thus, people with diabetic macular oedema and proliferative diabetic retinopathy need to be treated in a timely manner and reviewed in clinic for life. The population in the world is ageing. As a result, there are more people with eye diseases. There are also more treatments now for people with eye diseases. The workload in hospitals is increasing, making it difficult for the NHS to cope with the demand. There are not enough ophthalmologists (eye doctors) to look after patients. Delayed appointments and treatment mean that patients may lose sight. The goal of EMERALD (Effectiveness of Multimodal imaging for the Evaluation of Retinal oedema And new vesseLs in Diabetic retinopathy) was to see if patients with treated and stable diabetic macular oedema or proliferative diabetic retinopathy could be followed by 'ophthalmic graders', who are not doctors but are trained to diagnose diabetic macular oedema and proliferative diabetic retinopathy. In EMERALD, trained ophthalmic graders examined photographs of the back of the eye of people with diabetic macular oedema and proliferative diabetic retinopathy. They checked if diabetic macular oedema and proliferative diabetic retinopathy remain inactive. If so, patients could continue follow-up with the ophthalmic graders. If diabetic macular oedema or proliferative diabetic retinopathy were active, graders would immediately refer patients to ophthalmologists. EMERALD found that graders were excellent at detecting diabetic macular oedema, and this could give ophthalmologists time to see other patients. Graders were not quite as good at detecting active proliferative diabetic retinopathy. However, considering that patients had already had treatment, this may still be safe. Patients participating in focus group discussions mentioned that they would prefer to see ophthalmologists, so they could ask questions about their eye condition. If this was not possible, they would like to have immediate results from graders and still see the ophthalmologist from time to time.

Cataract Surgery Redesign: Meeting Increasing Demand, Training, Audit and Patient-Centered Care.

OBJECTIVE: The demand for cataract surgery in Fife (a well-defined region in southeast Scotland) was steadily increasing over 15 years. Cataract surgery was therefore being outsourced to meet demand with consequences on list mix, training needs, patient experience and staff morale. We aimed to redesign our services to meet local demand, retain a patient-centered service and continue to fulfil training needs. METHODS: We quantified cataract surgery delivery over an 18-month period: before, during and after redesign of services. We studied numbers of operations, trainee cases and number of outsourced cases. We also considered the economic implications of the redesign. RESULTS: We studied three periods (each of six months duration): before redesign (BR), redesign period (RP) and post-redesign (PR). Data were collected on total operation numbers, number of cases performed by trainees, and numbers performed out with normal working hours (weekend lists) and external providers. An economic analysis examined the cost of outsourcing cataracts during BR and RP and the costs of the redesign, including building, equipment and additional nursing staff. CONCLUSION: Regional fulfilment of cataract surgery provision remains a continuous challenge within the NHS. We show that with minimal investment, smart redesign process and collaborative working, increased local provision is possible while fulfilling trainee needs and achieving the necessary clinical audits and national standards.

Evaluation of a New Model of Care for People with Complications of Diabetic Retinopathy: The EMERALD Study.

PURPOSE: The increasing diabetes prevalence and advent of new treatments for its major visual-threatening complications (diabetic macular edema [DME] and proliferative diabetic retinopathy [PDR]), which require frequent life-long follow-up, have increased hospital demands markedly. Subsequent delays in patient's evaluation and treatment are causing sight loss. Strategies to increase capacity are needed urgently. The retinopathy (EMERALD) study tested diagnostic accuracy, acceptability, and costs of a new health care pathway for people with previously treated DME or PDR. DESIGN: Prospective, multicenter, case-referent, cross-sectional, diagnostic accuracy study undertaken in 13 hospitals in the United Kingdom. PARTICIPANTS: Adults with type 1 or 2 diabetes previously successfully treated DME or PDR who, at the time of enrollment, had active or inactive disease. METHODS: A new health care pathway entailing multimodal imaging (spectral-domain OCT for DME, and 7-field Early Treatment Diabetic Retinopathy Study [ETDRS] and ultra-widefield [UWF] fundus images for PDR) interpreted by trained nonmedical staff (ophthalmic graders) to detect reactivation of disease was compared with the current standard care (face-to-face examination by ophthalmologists). MAIN OUTCOME MEASURES: Primary outcome: sensitivity of the new pathway. SECONDARY OUTCOMES: specificity; agreement between pathways; costs; acceptability; proportions requiring subsequent ophthalmologist assessment, unable to undergo imaging, and with inadequate images or indeterminate findings. RESULTS: The new pathway showed sensitivity of 97% (95% confidence interval [CI], 92%-99%) and specificity of 31% (95% CI, 23%-40%) to detect DME. For PDR, sensitivity and specificity using 7-field ETDRS images (85% [95% CI, 77%-91%] and 48% [95% CI, 41%-56%], respectively) or UWF images (83% [95% CI, 75%-89%] and 54% [95% CI, 46%-61%], respectively) were comparable. For detection of high-risk PDR, sensitivity and specificity were higher when using UWF images (87% [95% CI, 78%-93%] and 49% [95% CI, 42%-56%], respectively, for UWF versus 80% [95% CI, 69-88%] and 40% [95% CI, 34%-47%], respectively, for 7-field ETDRS images). Participants preferred ophthalmologists' assessments; in their absence, they preferred immediate feedback by graders, maintaining periodic ophthalmologist evaluations. When compared with the current standard of care, the new pathway could save £1390 per 100 DME visits and between £461 and £1189 per 100 PDR visits. CONCLUSIONS: The new pathway has acceptable sensitivity and would release resources. Users' suggestions should guide implementation.

Use of personalised risk-based screening schedules to optimise workload and sojourn time in screening programmes for diabetic retinopathy: A retrospective cohort study.

BACKGROUND: National guidelines in most countries set screening intervals for diabetic retinopathy (DR) that are insufficiently informed by contemporary incidence rates. This has unspecified implications for interval disease risks (IDs) of referable DR, disparities in ID between groups or individuals, time spent in referable state before screening (sojourn time), and workload. We explored the effect of various screening schedules on these outcomes and developed an open-access interactive policy tool informed by contemporary DR incidence rates. METHODS AND FINDINGS: Scottish Diabetic Retinopathy Screening Programme data from 1 January 2007 to 31 December 2016 were linked to diabetes registry data. This yielded 128,606 screening examinations in people with type 1 diabetes (T1D) and 1,384,360 examinations in people with type 2 diabetes (T2D). Among those with T1D, 47% of those without and 44% of those with referable DR were female, mean diabetes duration was 21 and 23 years, respectively, and mean age was 26 and 24 years, respectively. Among those with T2D, 44% of those without and 42% of those with referable DR were female, mean diabetes duration was 9 and 14 years, respectively, and mean age was 58 and 52 years, respectively. Individual probability of developing referable DR was estimated using a generalised linear model and was used to calculate the intervals needed to achieve various IDs across prior grade strata, or at the individual level, and the resultant workload and sojourn time. The current policy in Scotland-screening people with no or mild disease annually and moderate disease every 6 months-yielded large differences in ID by prior grade (13.2%, 3.6%, and 0.6% annually for moderate, mild, and no prior DR strata, respectively, in T1D) and diabetes type (2.4% in T1D and 0.6% in T2D overall). Maintaining these overall risks but equalising risk across prior grade strata would require extremely short intervals in those with moderate DR (1-2 months) and very long intervals in those with no prior DR (35-47 months), with little change in workload or average sojourn time. Changing to intervals of 12, 9, and 3 months in T1D and to 24, 9, and 3 months in T2D for no, mild, and moderate DR strata, respectively, would substantially reduce disparity in ID across strata and between diabetes types whilst reducing workload by 26% and increasing sojourn time by 2.3 months. Including clinical risk factor data gave a small but significant increment in prediction of referable DR beyond grade (increase in C-statistic of 0.013 in T1D and 0.016 in T2D, both p < 0.001). However, using this model to derive personalised intervals did not have substantial workload or sojourn time benefits over stratum-specific intervals. The main limitation is that the results are pertinent only to countries that share broadly similar rates of retinal disease and risk factor distributions to Scotland. CONCLUSIONS: Changing current policies could reduce disparities in ID and achieve substantial reductions in workload within the range of IDs likely to be deemed acceptable. Our tool should facilitate more rational policy setting for screening.

Effectiveness of Multimodal imaging for the Evaluation of Retinal oedema And new vesseLs in Diabetic retinopathy (EMERALD).

INTRODUCTION: Diabetic macular oedema (DMO) and proliferative diabetic retinopathy (PDR) are the major causes of sight loss in people with diabetes. Due to the increased prevalence of diabetes, the workload related to these complications is increasing making it difficult for Hospital Eye Services (HSE) to meet demands. METHODS AND ANALYSIS: Effectiveness of Multimodal imaging for the Evaluation of Retinal oedema And new vesseLs in Diabetic retinopathy (EMERALD) is a prospective, case-referent, cross-sectional diagnostic study. It aims at determining the diagnostic performance, cost-effectiveness and acceptability of a new form of surveillance for people with stable DMO and/or PDR, which entails multimodal imaging and image review by an ophthalmic grader, using the current standard of care (evaluation of patients in clinic by an ophthalmologist) as the reference standard. If safe, cost-effective and acceptable, this pathway could help HES by freeing ophthalmologist time. The primary outcome of EMERALD is sensitivity of the new surveillance pathway in detecting active DMO/PDR. Secondary outcomes include specificity, agreement between new and the standard care pathway, positive and negative likelihood ratios, cost-effectiveness, acceptability, proportion of patients requiring subsequent full clinical assessment, unable to undergo imaging, with inadequate quality images or indeterminate findings. ETHICS AND DISSEMINATION: Ethical approval was obtained for this study from the Office for Research Ethics Committees Northern Ireland (reference 17/NI/0124). Study results will be published as a Health Technology Assessment monograph, in peer-reviewed national and international journals and presented at national/international conferences and to patient groups. TRIAL REGISTRATION NUMBER: NCT03490318 and ISRCTN:10856638.

Cataract surgery: factors influencing decision to treat and implications for training (south-east Scotland 2008-2014).

PURPOSE: To describe the population referred for cataract surgery, identify factors that influenced decision to treat, and patients suitable for ophthalmic training. PATIENTS AND METHODS: A total of 2,693 consecutive referrals over 6 years were interrogated using Business Objects software on cataract electronic patient records. RESULTS: A total of 2,693 patients were referred for cataract surgery (group A). Of these patients 2,132 (79%) had surgery (group B) and 561 (21%) did not (group C). Age for group B vs group C: 672 (32%) vs 115 (20%) ≤69 years, P<0.001; 803 (38%) vs 225 (40%) 70-79 years, P=0.48; 586 (27%) vs 203 (36%) 80-89 years, P<0.05; 71 (3%) vs 18 (3%) ≥90 years, P=1.0. Visual acuity, group B vs group C: 556 (26%) vs 664 (59%) 6/12 or better; 1,275 (60%) vs 367 (33%) 6/18-6/60; 266 (12%) vs 64 (6%) counting fingers or worse, P<0.05. Medical history for group B vs C: cognitive impairment: 55 (2.6%) vs 29 (5.2%), P<0.05; cardiovascular accident: 158 (7.4%) vs 60 (10.7%), P<0.05; diabetes: 372 (17.4%) vs 96 (17.1%), P=0.87; COPD/asthma: 382 (17.9%) vs 93 (16.6%), P=0.53; heart disease: 535 (25.1%) vs 155 (27.6%), P=0.35; hypertension: 971 (45.5%) vs 263 (46.9%), P=0.73. Ocular history for group B vs C was significant (P<0.05) for age-related macular degeneration 255 (12.0%) vs 93 (16.6%), other macular pathology 38 (1.8%) vs 25 (4.5%), corneal pathology 92 (4.3%) vs 36 (6.4%), amblyopia 37 (1.7%) vs 22 (3.9%). Detailed data on presenting complaint, ophthalmic history, and social status is discussed. CONCLUSION: We observed that surgery at a younger age with good levels of visual acuity was a factor in deferring cataract surgery. Cognitive impairment, cardiovascular accident, amblyopia, corneal and macular pathology significantly affected decision not to operate. We estimate that 80% of patients would be suitable for ophthalmic training.

Improving the cost-effectiveness of photographic screening for diabetic macular oedema: a prospective, multi-centre, UK study.

BACKGROUND/AIMS: Retinal screening programmes in England and Scotland have similar photographic grading schemes for background (non-proliferative) and proliferative diabetic retinopathy, but diverge over maculopathy. We looked for the most cost-effective method of identifying diabetic macular oedema from retinal photographs including the role of automated grading and optical coherence tomography, a technology that directly visualises oedema. METHODS: Patients from seven UK centres were recruited. The following features in at least one eye were required for enrolment: microaneurysms/dot haemorrhages or blot haemorrhages within one disc diameter, or exudates within one or two disc diameters of the centre of the macula. Subjects had optical coherence tomography and digital photography. Manual and automated grading schemes were evaluated. Costs and QALYs were modelled using microsimulation techniques. RESULTS: 3540 patients were recruited, 3170 were analysed. For diabetic macular oedema, England's scheme had a sensitivity of 72.6% and specificity of 66.8%; Scotland's had a sensitivity of 59.5% and specificity of 79.0%. When applying a ceiling ratio of £30,000 per quality adjusted life years (QALY) gained, Scotland's scheme was preferred. Assuming automated grading could be implemented without increasing grading costs, automation produced a greater number of QALYS for a lower cost than England's scheme, but was not cost effective, at the study's operating point, compared with Scotland's. The addition of optical coherence tomography, to each scheme, resulted in cost savings without reducing health benefits. CONCLUSIONS: Retinal screening programmes in the UK should reconsider the screening pathway to make best use of existing and new technologies.

First versus second eye intravitreal ranibizumab therapy for wet AMD.

PURPOSE: To evaluate the short-term visual outcomes after intravitreal ranibizumab for wet age-related macular degeneration, when used in first eyes (good vision in the untreated eye) compared with second eyes (significant visual impairment in the untreated eye). METHODS: Seventy-five consecutive patients who received intravitreal ranibizumab injection were divided into Group A, comprising 35 first eyed patients and Group B, comprising 40 second eyes. Visual acuity and contrast sensitivity was compared before treatment, and 3 months after the 3rd injection. Results were compared at 95% confidence interval. RESULTS: Mean pretreatment logMar visual acuity was 0.86 (standard deviation 0.28) in Group A whereas Group B was 0.66 (standard deviation 0.36) (P = 0.007). Posttreatment the mean visual acuity in Group A was 0.63 (standard deviation 0.37) and in Group B was 0.44 (standard deviation 0.33) (P = 0.02). The mean numbers of letters gained per patient were 11.1 (Group A) and 10.6 (Group B). Half of all patients showed significant improvement of visual acuity (> or =15 letters gain). Contrast sensitivity significantly improved in both groups and was usually, but not always, associated with visual gain. CONCLUSION: Second eye patients tend to present to clinical diagnosis at a better visual acuity than first ones and subsequently have better chances for better posttreatment visual acuity. However, both groups have an equal chance of significant visual improvement.

Evaluating the use of a scanning laser-derived oedema index to grade diabetic retinopathy and maculopathy.

PURPOSE: To evaluate the Heidelburg Retina Tomograph II (HRTII) retinal module as a tool for grading severity of retinopathy in a diabetic retinal screening and treatment service. METHODS: Seventy-seven consecutive patients with type 2 diabetes underwent scanning laser tomography using the HRTII. Scan data were analysed using the proprietary macular module software and oedema indices calculated for each of nine topographic macular zones. Two consultant ophthalmologists, masked to the result of the HRTII scans, graded each subject for severity of retinopathy and presence of macular oedema. The oedema indices were analysed statistically to determine whether these correlated with severity of retinopathy and presence of macular oedema. RESULTS: There is an increased oedema index in severe non-proliferative diabetic retinopathy in the outer temporal zone compared with lesser grades of diabetic retinopathy (P = 0.001). In patients with clinically detectable macular oedema, the oedema index from the 500-microm-diameter central zone was significantly higher than those without (P = 0.03). CONCLUSION: The scanning laser-derived oedema index differentiated between moderate and severe non-proliferative diabetic retinopathy in this series and detected diabetic macular oedema. Further development of this technology may provide an important tool to supplement retinal photographic surveillance in eye clinics overwhelmed by an increasing prevalence of type 2 diabetes.