Search

Abstract Diabetic retinopathy (DR) at risk of vision loss (referable DR) needs to be identified by retinal screening and referred to an ophthalmologist. Existing automated algorithms have mostly been developed from images acquired with high cost mydriatic retinal cameras and cannot be applied in the settings used in most low- and middle-income countries. In this prospective multicentre study, we developed a deep learning system (DLS) that detects referable DR from retinal images acquired using handheld non-mydriatic fundus camera by non-technical field workers in 20 sites across India. Macula-centred and optic-disc-centred images from 16,247 eyes (9778 participants) were used to train and cross-validate the DLS and risk factor based logistic regression models. The DLS achieved an AUROC of 0.99 (1000 times bootstrapped 95% CI 0.98–0.99) using two-field retinal images, with 93.86 (91.34–96.08) sensitivity and 96.00 (94.68–98.09) specificity at the Youden’s index operational point. With single field inputs, the DLS reached AUROC of 0.98 (0.98–0.98) for the macula field and 0.96 (0.95–0.98) for the optic-disc field. Intergrader performance was 90.01 (88.95–91.01) sensitivity and 96.09 (95.72–96.42) specificity. The image based DLS outperformed all risk factor-based models. This DLS demonstrated a clinically acceptable performance for the identification of referable DR despite challenging image capture conditions.

Abstract To examine the retinal structure and function in relation to diabetes duration and glycemia in patients without diabetic retinopathy (DR). 85 adults with type 2 diabetes without DR or macular edema underwent dilated indirect ophthalmoscopy, optical coherence tomography (OCT), ultra-wide field fundus photography, multifocal electroretinography (mfERG) and HbA1C assessment. Patients were stratified as those with diabetes duration

Abstract Prediction models for population-based screening need, for global usage, to be resource-driven, involving predictors that are affordably resourced. Here, we report the development and validation of three resource-driven risk models to identify people with type 2 diabetes (T2DM) at risk of stage 3 CKD defined by a decline in estimated glomerular filtration rate (eGFR) to below 60 mL/min/1.73m2. The observational study cohort used for model development consisted of data from a primary care dataset of 20,510 multi-ethnic individuals with T2DM from London, UK (2007–2018). Discrimination and calibration of the resulting prediction models developed using cox regression were assessed using the c-statistic and calibration slope, respectively. Models were internally validated using tenfold cross-validation and externally validated on 13,346 primary care individuals from Wales, UK. The simplest model was simplified into a risk score to enable implementation in community-based medicine. The derived full model included demographic, laboratory parameters, medication-use, cardiovascular disease history (CVD) and sight threatening retinopathy status (STDR). Two less resource-intense models were developed by excluding CVD and STDR in the second model and HbA1c and HDL in the third model. All three 5-year risk models had good internal discrimination and calibration (optimism adjusted C-statistics were each 0.85 and calibration slopes 0.999–1.002). In Wales, models achieved excellent discrimination(c-statistics ranged 0.82–0.83). Calibration slopes at 5-years suggested models over-predicted risks, however were successfully updated to accommodate reduced incidence of stage 3 CKD in Wales, which improved their alignment with the observed rates in Wales (E/O ratios near to 1). The risk score demonstrated similar model performance compared to direct evaluation of the cox model. These resource-driven risk prediction models may enable universal screening for Stage 3 CKD to enable targeted early optimisation of risk factors for CKD.

Abstract Screening effectively identifies patients at risk of sight-threatening diabetic retinopathy (STDR) when retinal images are captured through dilated pupils. Pharmacological mydriasis is not logistically feasible in non-clinical, community DR screening, where acquiring gradable retinal images using handheld devices exhibits high technical failure rates, reducing STDR detection. Deep learning (DL) based gradability predictions at acquisition could prompt device operators to recapture insufficient quality images, increasing gradable image proportions and consequently STDR detection. Non-mydriatic retinal images were captured as part of SMART India, a cross-sectional, multi-site, community-based, house-to-house DR screening study between August 2018 and December 2019 using the Zeiss Visuscout 100 handheld camera. From 18,277 patient eyes (40,126 images), 16,170 patient eyes (35,319 images) were eligible and 3261 retinal images (1490 patient eyes) were sampled then labelled by two ophthalmologists. Compact DL model area under the receiver operator characteristic curve was 0.93 (0.01) following five-fold cross-validation. Compact DL model agreement (Kappa) were 0.58, 0.69 and 0.69 for high specificity, balanced sensitivity/specificity and high sensitivity operating points compared to an inter-grader agreement of 0.59. Compact DL gradability model performance was favourable compared to ophthalmologists. Compact DL models can effectively classify non-mydriatic, handheld retinal image gradability with potential applications within community-based DR screening.