Large-scale machine-learning-based phenotyping significantly improves genomic discovery for optic nerve head morphology.

Genome-wide association studies (GWASs) require accurate cohort phenotyping, but expert labeling can be costly, time intensive, and variable. Here, we develop a machine learning (ML) model to predict glaucomatous optic nerve head features from color fundus photographs. We used the model to predict vertical cup-to-disc ratio (VCDR), a diagnostic parameter and cardinal endophenotype for glaucoma, in 65,680 Europeans in the UK Biobank (UKB). A GWAS of ML-based VCDR identified 299 independent genome-wide significant (GWS; p ≤ 5 × 10-8) hits in 156 loci. The ML-based GWAS replicated 62 of 65 GWS loci from a recent VCDR GWAS in the UKB for which two ophthalmologists manually labeled images for 67,040 Europeans. The ML-based GWAS also identified 93 novel loci, significantly expanding our understanding of the genetic etiologies of glaucoma and VCDR. Pathway analyses support the biological significance of the novel hits to VCDR: select loci near genes involved in neuronal and synaptic biology or harboring variants are known to cause severe Mendelian ophthalmic disease. Finally, the ML-based GWAS results significantly improve polygenic prediction of VCDR and primary open-angle glaucoma in the independent EPIC-Norfolk cohort.

Association of severity of primary open-angle glaucoma with serum vitamin D levels in patients of African descent.

Purpose: To study the relationship between primary open-angle glaucoma (POAG) in a cohort of patients of African descent (AD) and serum vitamin D levels. Methods: A subset of the AD and glaucoma evaluation study III (ADAGES III) cohort, consisting of 357 patients with a diagnosis of POAG and 178 normal controls of self-reported AD, were included in this analysis. Demographic information, family history, and blood samples were collected from all the participants. All the subjects underwent clinical evaluation, including visual field (VF) mean deviation (MD), central cornea thickness (CCT), intraocular pressure (IOP), and height and weight measurements. POAG patients were classified into early and advanced phenotypes based on the severity of their visual field damage, and they were matched for age, gender, and history of hypertension and diabetes. Serum 25-Hydroxy (25-OH) vitamin D levels were measured by enzyme-linked immunosorbent assay (ELISA). The association of serum vitamin D levels with the development and severity of POAG was tested by analysis of variance (ANOVA) and the paired t-test. Results: The 178 early POAG subjects had a visual field MD of better than -4.0 dB, and the 179 advanced glaucoma subjects had a visual field MD of worse than -10 dB. The mean (95% confidence interval [CI]) levels of vitamin D of the subjects in the control (8.02 ± 6.19 pg/ml) and early phenotype (7.56 ± 5.74 pg/ml) groups were significantly or marginally significantly different from the levels observed in subjects with the advanced phenotype (6.35 ± 4.76 pg/ml; p = 0.0117 and 0.0543, respectively). In contrast, the mean serum vitamin D level in controls was not significantly different from that of the subjects with the early glaucoma phenotype (p = 0.8508). Conclusions: In this AD cohort, patients with advanced glaucoma had lower serum levels of vitamin D compared with early glaucoma and normal subjects.

Deep Learning and Glaucoma Specialists: The Relative Importance of Optic Disc Features to Predict Glaucoma Referral in Fundus Photographs.

PURPOSE: To develop and validate a deep learning (DL) algorithm that predicts referable glaucomatous optic neuropathy (GON) and optic nerve head (ONH) features from color fundus images, to determine the relative importance of these features in referral decisions by glaucoma specialists (GSs) and the algorithm, and to compare the performance of the algorithm with eye care providers. DESIGN: Development and validation of an algorithm. PARTICIPANTS: Fundus images from screening programs, studies, and a glaucoma clinic. METHODS: A DL algorithm was trained using a retrospective dataset of 86 618 images, assessed for glaucomatous ONH features and referable GON (defined as ONH appearance worrisome enough to justify referral for comprehensive examination) by 43 graders. The algorithm was validated using 3 datasets: dataset A (1205 images, 1 image/patient; 18.1% referable), images adjudicated by panels of GSs; dataset B (9642 images, 1 image/patient; 9.2% referable), images from a diabetic teleretinal screening program; and dataset C (346 images, 1 image/patient; 81.7% referable), images from a glaucoma clinic. MAIN OUTCOME MEASURES: The algorithm was evaluated using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity for referable GON and glaucomatous ONH features. RESULTS: The algorithm's AUC for referable GON was 0.945 (95% confidence interval [CI], 0.929-0.960) in dataset A, 0.855 (95% CI, 0.841-0.870) in dataset B, and 0.881 (95% CI, 0.838-0.918) in dataset C. Algorithm AUCs ranged between 0.661 and 0.973 for glaucomatous ONH features. The algorithm showed significantly higher sensitivity than 7 of 10 graders not involved in determining the reference standard, including 2 of 3 GSs, and showed higher specificity than 3 graders (including 1 GS), while remaining comparable to others. For both GSs and the algorithm, the most crucial features related to referable GON were: presence of vertical cup-to-disc ratio of 0.7 or more, neuroretinal rim notching, retinal nerve fiber layer defect, and bared circumlinear vessels. CONCLUSIONS: A DL algorithm trained on fundus images alone can detect referable GON with higher sensitivity than and comparable specificity to eye care providers. The algorithm maintained good performance on an independent dataset with diagnoses based on a full glaucoma workup.

Using a Deep Learning Algorithm and Integrated Gradients Explanation to Assist Grading for Diabetic Retinopathy.

PURPOSE: To understand the impact of deep learning diabetic retinopathy (DR) algorithms on physician readers in computer-assisted settings. DESIGN: Evaluation of diagnostic technology. PARTICIPANTS: One thousand seven hundred ninety-six retinal fundus images from 1612 diabetic patients. METHODS: Ten ophthalmologists (5 general ophthalmologists, 4 retina specialists, 1 retina fellow) read images for DR severity based on the International Clinical Diabetic Retinopathy disease severity scale in each of 3 conditions: unassisted, grades only, or grades plus heatmap. Grades-only assistance comprised a histogram of DR predictions (grades) from a trained deep-learning model. For grades plus heatmap, we additionally showed explanatory heatmaps. MAIN OUTCOME MEASURES: For each experiment arm, we computed sensitivity and specificity of each reader and the algorithm for different levels of DR severity against an adjudicated reference standard. We also measured accuracy (exact 5-class level agreement and Cohen's quadratically weighted κ), reader-reported confidence (5-point Likert scale), and grading time. RESULTS: Readers graded more accurately with model assistance than without for the grades-only condition (P < 0.001). Grades plus heatmaps improved accuracy for patients with DR (P < 0.001), but reduced accuracy for patients without DR (P = 0.006). Both forms of assistance increased readers' sensitivity moderate-or-worse DR: unassisted: mean, 79.4% [95% confidence interval (CI), 72.3%-86.5%]; grades only: mean, 87.5% [95% CI, 85.1%-89.9%]; grades plus heatmap: mean, 88.7% [95% CI, 84.9%-92.5%] without a corresponding drop in specificity (unassisted: mean, 96.6% [95% CI, 95.9%-97.4%]; grades only: mean, 96.1% [95% CI, 95.5%-96.7%]; grades plus heatmap: mean, 95.5% [95% CI, 94.8%-96.1%]). Algorithmic assistance increased the accuracy of retina specialists above that of the unassisted reader or model alone; and increased grading confidence and grading time across all readers. For most cases, grades plus heatmap was only as effective as grades only. Over the course of the experiment, grading time decreased across all conditions, although most sharply for grades plus heatmap. CONCLUSIONS: Deep learning algorithms can improve the accuracy of, and confidence in, DR diagnosis in an assisted read setting. They also may increase grading time, although these effects may be ameliorated with experience.

The African Descent and Glaucoma Evaluation Study (ADAGES) III: Contribution of Genotype to Glaucoma Phenotype in African Americans: Study Design and Baseline Data.

PURPOSE: To describe the study protocol and baseline characteristics of the African Descent and Glaucoma Evaluation Study (ADAGES) III. DESIGN: Cross-sectional, case-control study. PARTICIPANTS: Three thousand two hundred sixty-six glaucoma patients and control participants without glaucoma of African or European descent were recruited from 5 study centers in different regions of the United States. METHODS: Individuals of African descent (AD) and European descent (ED) with primary open-angle glaucoma (POAG) and control participants completed a detailed demographic and medical history interview. Standardized height, weight, and blood pressure measurements were obtained. Saliva and blood samples to provide serum, plasma, DNA, and RNA were collected for standardized processing. Visual fields, stereoscopic disc photographs, and details of the ophthalmic examination were obtained and transferred to the University of California, San Diego, Data Coordinating Center for standardized processing and quality review. MAIN OUTCOME MEASURES: Participant gender, age, race, body mass index, blood pressure, history of smoking and alcohol use in POAG patients and control participants were described. Ophthalmic measures included intraocular pressure, visual field mean deviation, central corneal thickness, glaucoma medication use, or past glaucoma surgery. Ocular conditions, including diabetic retinopathy, age-related macular degeneration, and past cataract surgery, were recorded. RESULTS: The 3266 ADAGES III study participants in this report include 2146 AD POAG patients, 695 ED POAG patients, 198 AD control participants, and 227 ED control participants. The AD POAG patients and control participants were significantly younger (both, 67.4 years) than ED POAG patients and control participants (73.4 and 70.2 years, respectively). After adjusting for age, AD POAG patients had different phenotypic characteristics compared with ED POAG patients, including higher intraocular pressure, worse visual acuity and visual field mean deviation, and thinner corneas (all P < 0.001). Family history of glaucoma did not differ between AD and ED POAG patients. CONCLUSIONS: With its large sample size, extensive specimen collection, and deep phenotyping of AD and ED glaucoma patients and control participants from different regions in the United States, the ADAGES III genomics study will address gaps in our knowledge of the genetics of POAG in this high-risk population.

'Blue bubble' technique: an ab interno approach for Descemet separation in deep anterior lamellar keratoplasty using trypan blue stained viscoelastic device.

BACKGROUND: In this study, we examined a novel variant of 'big-bubble' deep anterior lamellar keratoplasty using trypan-blue-stained viscoelastic device for the creation of a pre-descemetic bubble. METHODS: Ten corneoscleral rims were mounted on an artificial anterior chamber (AC). The AC was filled with air through a limbal paracentesis. A Melles' triangulated spatula was inserted through the paracentesis, with its tip penetrating the AC, was then slightly retracted and pushed into the deep stroma above the roof of the paracentesis. A mixture of trypan blue and viscoelastic device (Healon, Abbott Medical Optics, Abbott Park, Illinois) was injected into this intra-stromal pocket using a 27-G cannula to create a pre-descemetic separation bubble. Bubble type and visualization of dyed viscoelastic device were noted. The method was later employed in three cases. RESULTS: In all 10 corneoscleral rims, the technique successfully created a visible pre-descemetic (type 1) bubble that could be expanded up to the predicted diameter of trephination. Subsequent trephination and the removal of corneal stroma were uneventful. In two out of four clinical cases, a type 1 bubble was created, while in two others, visco-dissection failed and dyed viscoelastic was seen in the AC. CONCLUSIONS: The presented technique holds promise of being a relatively easy to perform, predictable and well-controlled alternative for achieving a type 1 bubble during deep anterior lamellar keratoplasty surgery. The trypan-blue-stained viscoelastic device facilitates proper visualization and control of the separation bubble and assists in identifying the penetrance to the separation bubble prior to removal of the stromal cap.

Diagnostic Accuracy of the Spectralis and Cirrus Reference Databases in Differentiating between Healthy and Early Glaucoma Eyes.

PURPOSE: To evaluate and compare the diagnostic accuracy of global and sector analyses for detection of early visual field (VF) damage using the retinal nerve fiber layer (RNFL) reference databases of the Spectralis (Heidelberg Engineering, Heidelberg, Germany) and Cirrus (Carl Zeiss Meditec, Dublin, CA) spectral-domain optical coherence tomography (SD OCT) devices. METHODS: Healthy subjects and glaucoma suspects from the Diagnostic Innovations in Glaucoma Study (DIGS) and African Descent and Glaucoma Evaluation Study (ADAGES) with at least 2 years of follow-up were included. Global and sectoral RNFL measures were classified as within normal limits, borderline (BL), and outside normal limits (ONL) on the basis of the device reference databases. The sensitivity of ONL classification was estimated in glaucoma suspect eyes that developed repeatable VF damage. RESULTS: A total of 353 glaucoma suspect eyes and 279 healthy eyes were included. A total of 34 (9.6%) of the glaucoma suspect eyes developed VF damage. In glaucoma suspect eyes, Spectralis and Cirrus ONL classification was present in 47 eyes (13.3%) and 24 eyes (6.8%), respectively. The sensitivity of the global RNFL ONL classification among eyes that developed VF damage was 23.5% for Cirrus and 32.4% for Spectralis. The specificity of within-normal-limits global classification in healthy eyes was 100% for Cirrus and 99.6% for Spectralis. There was moderate to substantial agreement between Cirrus and Spectralis classification as ONL. CONCLUSIONS: The Spectralis and Cirrus reference databases have a high specificity for identifying healthy eyes and good agreement for detection of eyes with early glaucoma damage.

Rate and Pattern of Rim Area Loss in Healthy and Progressing Glaucoma Eyes.

PURPOSE: To characterize the rate and pattern of age-related and glaucomatous neuroretinal rim area changes in subjects of African and European descent. DESIGN: Prospective longitudinal study. PARTICIPANTS: Two hundred ninety-six eyes of 157 healthy subjects (88 patients of African descent and 69 of European descent) and 73 progressing glaucoma eyes of 67 subjects (24 patients of African descent and 43 of European descent) from the Diagnostic Innovations in Glaucoma Study and the African Descent and Glaucoma Evaluation Study were included. METHODS: Global and sectoral rim areas were measured using confocal laser scanning ophthalmoscopy. Masked stereophotograph review determined progression of glaucomatous optic disc damage. The rates of absolute rim area loss and percentage rim area loss in healthy and progressing glaucomatous eyes were compared using multivariate, nested, mixed-effects models. MAIN OUTCOME MEASURES: Rate of rim area loss over time. RESULTS: The median follow-up time was 5.0 years (interquartile range, 2.0-7.4 years) for healthy eyes and 8.3 years (interquartile range, 7.5-9.9 years) for progressing glaucoma eyes. The mean rate of global rim area loss was significantly faster in progressing glaucomatous eyes compared with healthy eyes for both rim area loss (-10.2×10(-3) vs. -2.8×10(-3) mm(2)/year, respectively; P < 0.001) and percentage rim area loss (-1.1% vs. -0.2%/year, respectively; P < 0.001), but considerable overlap existed between the 2 groups. Sixty-three percent of progressing glaucoma eyes had a rate of change faster than the fifth quantile of healthy eyes. For both healthy and progressing eyes, the pattern of rim area loss and percentage rim area loss were similar, tending to be fastest in the superior temporal and inferior temporal sectors. The rate of change was similar in progressing eyes of patients of African or European descent. CONCLUSIONS: Compared with healthy eyes, the mean rate of global rim area loss was 3.7 times faster and the mean rate of global percentage rim area loss was 5.4 times faster in progressing glaucoma eyes. A reference database of healthy eyes can be used to help clinicians distinguish age-related rim area loss from rim area loss resulting from glaucoma.

Rates of retinal nerve fiber layer thinning in glaucoma suspect eyes.

PURPOSE: To compare the rates of retinal nerve fiber layer (RNFL) loss in patients suspected of having glaucoma who developed visual field damage (VFD) with those who did not develop VFD and to determine whether the rate of RNFL loss can be used to predict the development of VFD. DESIGN: Prospective, observational cohort study. PARTICIPANTS: Glaucoma suspects, defined as having glaucomatous optic neuropathy or ocular hypertension (intraocular pressure, >21 mmHg) without repeatable VFD at baseline, from the Diagnostic Innovations in Glaucoma Study and the African Descent and Glaucoma Evaluation Study. METHODS: Global and quadrant RNFL thickness (RNFLT) were measured with the Spectralis spectral-domain optical coherence tomography (SD-OCT; Spectralis HRA+OCT [Heidelberg Engineering, Heidelberg, Germany]). Visual field damage was defined as having 3 consecutive abnormal visual fields. The rate of RNFL loss in eyes developing VFD was compared to eyes not developing VFD using multivariate linear mixed-effects models. A joint longitudinal survival model used the estimated RNFLT slope to predict the risk of developing VFD, while adjusting for potential confounding variables. MAIN OUTCOME MEASURES: The rate of RNFL thinning and the probability of developing VFD. RESULTS: Four hundred fifty-four eyes of 294 glaucoma suspects were included. The average number of SD-OCT examinations was 4.6 (range, 2-9), with median follow-up of 2.2 years (0.4-4.1 years). Forty eyes (8.8%) developed VFD. The estimated mean rate of global RNFL loss was significantly faster in eyes that developed VFD compared with eyes that did not develop VFD (-2.02 µm/year vs. -0.82 µm/year; P<0.001). The joint longitudinal survival model showed that each 1-µm/year faster rate of global RNFL loss corresponded to a 2.05-times higher risk of developing VFD (hazard ratio, 2.05; 95% confidence interval, 1.14-3.71; P = 0.017). CONCLUSIONS: The rate of global RNFL loss was more than twice as fast in eyes that developed VFD compared with eyes that did not develop VFD. A joint longitudinal survival model showed that a 1-µm/year faster rate of RNFLT loss corresponded to a 2.05-times higher risk of developing VFD. These results suggest that measuring the rate of SD-OCT RNFL loss may be a useful tool to help identify patients who are at a high risk of developing visual field loss.

Using generative AI to investigate medical imagery models and datasets.

BACKGROUND: AI models have shown promise in performing many medical imaging tasks. However, our ability to explain what signals these models have learned is severely lacking. Explanations are needed in order to increase the trust of doctors in AI-based models, especially in domains where AI prediction capabilities surpass those of humans. Moreover, such explanations could enable novel scientific discovery by uncovering signals in the data that aren't yet known to experts. METHODS: In this paper, we present a workflow for generating hypotheses to understand which visual signals in images are correlated with a classification model's predictions for a given task. This approach leverages an automatic visual explanation algorithm followed by interdisciplinary expert review. We propose the following 4 steps: (i) Train a classifier to perform a given task to assess whether the imagery indeed contains signals relevant to the task; (ii) Train a StyleGAN-based image generator with an architecture that enables guidance by the classifier ("StylEx"); (iii) Automatically detect, extract, and visualize the top visual attributes that the classifier is sensitive towards. For visualization, we independently modify each of these attributes to generate counterfactual visualizations for a set of images (i.e., what the image would look like with the attribute increased or decreased); (iv) Formulate hypotheses for the underlying mechanisms, to stimulate future research. Specifically, present the discovered attributes and corresponding counterfactual visualizations to an interdisciplinary panel of experts so that hypotheses can account for social and structural determinants of health (e.g., whether the attributes correspond to known patho-physiological or socio-cultural phenomena, or could be novel discoveries). FINDINGS: To demonstrate the broad applicability of our approach, we present results on eight prediction tasks across three medical imaging modalities-retinal fundus photographs, external eye photographs, and chest radiographs. We showcase examples where many of the automatically-learned attributes clearly capture clinically known features (e.g., types of cataract, enlarged heart), and demonstrate automatically-learned confounders that arise from factors beyond physiological mechanisms (e.g., chest X-ray underexposure is correlated with the classifier predicting abnormality, and eye makeup is correlated with the classifier predicting low hemoglobin levels). We further show that our method reveals a number of physiologically plausible, previously-unknown attributes based on the literature (e.g., differences in the fundus associated with self-reported sex, which were previously unknown). INTERPRETATION: Our approach enables hypotheses generation via attribute visualizations and has the potential to enable researchers to better understand, improve their assessment, and extract new knowledge from AI-based models, as well as debug and design better datasets. Though not designed to infer causality, importantly, we highlight that attributes generated by our framework can capture phenomena beyond physiology or pathophysiology, reflecting the real world nature of healthcare delivery and socio-cultural factors, and hence interdisciplinary perspectives are critical in these investigations. Finally, we will release code to help researchers train their own StylEx models and analyze their predictive tasks of interest, and use the methodology presented in this paper for responsible interpretation of the revealed attributes. FUNDING: Google.

A deep learning model for novel systemic biomarkers in photographs of the external eye: a retrospective study.

BACKGROUND: Photographs of the external eye were recently shown to reveal signs of diabetic retinal disease and elevated glycated haemoglobin. This study aimed to test the hypothesis that external eye photographs contain information about additional systemic medical conditions. METHODS: We developed a deep learning system (DLS) that takes external eye photographs as input and predicts systemic parameters, such as those related to the liver (albumin, aspartate aminotransferase [AST]); kidney (estimated glomerular filtration rate [eGFR], urine albumin-to-creatinine ratio [ACR]); bone or mineral (calcium); thyroid (thyroid stimulating hormone); and blood (haemoglobin, white blood cells [WBC], platelets). This DLS was trained using 123 130 images from 38 398 patients with diabetes undergoing diabetic eye screening in 11 sites across Los Angeles county, CA, USA. Evaluation focused on nine prespecified systemic parameters and leveraged three validation sets (A, B, C) spanning 25 510 patients with and without diabetes undergoing eye screening in three independent sites in Los Angeles county, CA, and the greater Atlanta area, GA, USA. We compared performance against baseline models incorporating available clinicodemographic variables (eg, age, sex, race and ethnicity, years with diabetes). FINDINGS: Relative to the baseline, the DLS achieved statistically significant superior performance at detecting AST >36·0 U/L, calcium <8·6 mg/dL, eGFR <60·0 mL/min/1·73 m2, haemoglobin <11·0 g/dL, platelets <150·0 × 103/µL, ACR ≥300 mg/g, and WBC <4·0 × 103/µL on validation set A (a population resembling the development datasets), with the area under the receiver operating characteristic curve (AUC) of the DLS exceeding that of the baseline by 5·3-19·9% (absolute differences in AUC). On validation sets B and C, with substantial patient population differences compared with the development datasets, the DLS outperformed the baseline for ACR ≥300·0 mg/g and haemoglobin <11·0 g/dL by 7·3-13·2%. INTERPRETATION: We found further evidence that external eye photographs contain biomarkers spanning multiple organ systems. Such biomarkers could enable accessible and non-invasive screening of disease. Further work is needed to understand the translational implications. FUNDING: Google.

Detection of signs of disease in external photographs of the eyes via deep learning.

Retinal fundus photographs can be used to detect a range of retinal conditions. Here we show that deep-learning models trained instead on external photographs of the eyes can be used to detect diabetic retinopathy (DR), diabetic macular oedema and poor blood glucose control. We developed the models using eye photographs from 145,832 patients with diabetes from 301 DR screening sites and evaluated the models on four tasks and four validation datasets with a total of 48,644 patients from 198 additional screening sites. For all four tasks, the predictive performance of the deep-learning models was significantly higher than the performance of logistic regression models using self-reported demographic and medical history data, and the predictions generalized to patients with dilated pupils, to patients from a different DR screening programme and to a general eye care programme that included diabetics and non-diabetics. We also explored the use of the deep-learning models for the detection of elevated lipid levels. The utility of external eye photographs for the diagnosis and management of diseases should be further validated with images from different cameras and patient populations.

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.

Predicting the risk of developing diabetic retinopathy using deep learning.

BACKGROUND: Diabetic retinopathy screening is instrumental to preventing blindness, but scaling up screening is challenging because of the increasing number of patients with all forms of diabetes. We aimed to create a deep-learning system to predict the risk of patients with diabetes developing diabetic retinopathy within 2 years. METHODS: We created and validated two versions of a deep-learning system to predict the development of diabetic retinopathy in patients with diabetes who had had teleretinal diabetic retinopathy screening in a primary care setting. The input for the two versions was either a set of three-field or one-field colour fundus photographs. Of the 575 431 eyes in the development set 28 899 had known outcomes, with the remaining 546 532 eyes used to augment the training process via multitask learning. Validation was done on one eye (selected at random) per patient from two datasets: an internal validation (from EyePACS, a teleretinal screening service in the USA) set of 3678 eyes with known outcomes and an external validation (from Thailand) set of 2345 eyes with known outcomes. FINDINGS: The three-field deep-learning system had an area under the receiver operating characteristic curve (AUC) of 0·79 (95% CI 0·77-0·81) in the internal validation set. Assessment of the external validation set-which contained only one-field colour fundus photographs-with the one-field deep-learning system gave an AUC of 0·70 (0·67-0·74). In the internal validation set, the AUC of available risk factors was 0·72 (0·68-0·76), which improved to 0·81 (0·77-0·84) after combining the deep-learning system with these risk factors (p<0·0001). In the external validation set, the corresponding AUC improved from 0·62 (0·58-0·66) to 0·71 (0·68-0·75; p<0·0001) following the addition of the deep-learning system to available risk factors. INTERPRETATION: The deep-learning systems predicted diabetic retinopathy development using colour fundus photographs, and the systems were independent of and more informative than available risk factors. Such a risk stratification tool might help to optimise screening intervals to reduce costs while improving vision-related outcomes. FUNDING: Google.

Changes in anterior segment parameters following insertion of ExPRESS mini glaucoma implant vs. trabeculectomy.

PURPOSE: To compare changes in anterior segment parameters following ExPRESS Mini Glaucoma Shunt surgery vs. trabeculectomy using the Pentacam rotating Scheimpflug camera. METHODS: In this prospective, comparative study, 27 patients with glaucoma treated at the Rabin Medical Center from 2009 to 2013 were enrolled in this prospective comparative study: 19 participants (19 eyes) underwent ExPRESS shunt implantation and 12 (13 eyes) underwent trabeculectomy. Changes in anterior chamber parameters at postoperative day 1 and postoperative month 3 were evaluated on Scheimpflug images. RESULTS: Intraocular pressure decreased significantly from baseline in both groups. The decrease in both groups was similar at postoperative month 3 (p=0.82). ExPRESS surgery caused a transient increase in posterior corneal astigmatism (p=0.008) and a transient decrease in anterior chamber depth (p=0.016) and volume (p=0.006) on postoperative day 1. At postoperative month 3, these parameters were no longer statistically significant (p=0.65, p=0.51, and p=0.57 respectively). Trabeculectomy caused a transient increase in anterior and posterior corneal astigmatism on postoperative day 1 (p=0.003 and p=0.005, respectively), which were not evident at postoperative month 3 (p=1.0 and p=1.0, respectively). At postoperative month 3, both ExPRESS and trabeculectomy showed similar changes in anterior chamber parameters. CONCLUSIONS: Both ExPRESS mini glaucoma implant and trabeculectomy significantly decreased intraocular pressure and had transient effects on anterior segment parameters, with minor differences between the methods.

Author Correction: Detection of anaemia from retinal fundus images via deep learning.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Detection of anaemia from retinal fundus images via deep learning.

Owing to the invasiveness of diagnostic tests for anaemia and the costs associated with screening for it, the condition is often undetected. Here, we show that anaemia can be detected via machine-learning algorithms trained using retinal fundus images, study participant metadata (including race or ethnicity, age, sex and blood pressure) or the combination of both data types (images and study participant metadata). In a validation dataset of 11,388 study participants from the UK Biobank, the fundus-image-only, metadata-only and combined models predicted haemoglobin concentration (in g dl-1) with mean absolute error values of 0.73 (95% confidence interval: 0.72-0.74), 0.67 (0.66-0.68) and 0.63 (0.62-0.64), respectively, and with areas under the receiver operating characteristic curve (AUC) values of 0.74 (0.71-0.76), 0.87 (0.85-0.89) and 0.88 (0.86-0.89), respectively. For 539 study participants with self-reported diabetes, the combined model predicted haemoglobin concentration with a mean absolute error of 0.73 (0.68-0.78) and anaemia an AUC of 0.89 (0.85-0.93). Automated anaemia screening on the basis of fundus images could particularly aid patients with diabetes undergoing regular retinal imaging and for whom anaemia can increase morbidity and mortality risks.

Remote Tool-Based Adjudication for Grading Diabetic Retinopathy.

PURPOSE: To present and evaluate a remote, tool-based system and structured grading rubric for adjudicating image-based diabetic retinopathy (DR) grades. METHODS: We compared three different procedures for adjudicating DR severity assessments among retina specialist panels, including (1) in-person adjudication based on a previously described procedure (Baseline), (2) remote, tool-based adjudication for assessing DR severity alone (TA), and (3) remote, tool-based adjudication using a feature-based rubric (TA-F). We developed a system allowing graders to review images remotely and asynchronously. For both TA and TA-F approaches, images with disagreement were reviewed by all graders in a round-robin fashion until disagreements were resolved. Five panels of three retina specialists each adjudicated a set of 499 retinal fundus images (1 panel using Baseline, 2 using TA, and 2 using TA-F adjudication). Reliability was measured as grade agreement among the panels using Cohen's quadratically weighted kappa. Efficiency was measured as the number of rounds needed to reach a consensus for tool-based adjudication. RESULTS: The grades from remote, tool-based adjudication showed high agreement with the Baseline procedure, with Cohen's kappa scores of 0.948 and 0.943 for the two TA panels, and 0.921 and 0.963 for the two TA-F panels. Cases adjudicated using TA-F were resolved in fewer rounds compared with TA (P < 0.001; standard permutation test). CONCLUSIONS: Remote, tool-based adjudication presents a flexible and reliable alternative to in-person adjudication for DR diagnosis. Feature-based rubrics can help accelerate consensus for tool-based adjudication of DR without compromising label quality. TRANSLATIONAL RELEVANCE: This approach can generate reference standards to validate automated methods, and resolve ambiguous diagnoses by integrating into existing telemedical workflows.

Comparing optical coherence tomography radial and cube scan patterns for measuring Bruch's membrane opening minimum rim width (BMO-MRW) in glaucoma and healthy eyes: cross-sectional and longitudinal analysis.

AIM: To compare the cube and radial scan patterns of the spectral domain optical coherence tomography (SD-OCT) for quantifying the Bruch's membrane opening minimum rim width (BMO-MRW). METHODS: Sixty healthy eyes and 189 glaucomatous eyes were included. The optic nerve head cube and radial pattern scans were acquired using Spectralis SD-OCT. BMO-MRWs were automatically delineated using the San Diego Automated Layer Segmentation Algorithm. The BMO-MRW diagnostic accuracy for glaucoma detection and rates of change derived from the two scan patterns were compared. RESULTS: There was a significant difference between the baseline global BMO-MRW measurements of cube and radial scans for healthy (301.9±57.8 µm and 334.7±61.8 µm, respectively, p<0.003) and glaucoma eyes (181.2±63.0 µm and 210.2±67.2 µm, respectively, p<0.001). The area under the receiver operating characteristic curve for differentiating between healthy and glaucoma eyes was 0.90 for both the radial scan-based and cube scan-based BMO-MRW. No significant difference in the rate of BMO-MRW change (mean follow-up years) by scan pattern was found among both healthy (cube: -1.47 µm/year, radial: -1.53 µm/year; p=0.48) (1.6 years) and glaucoma eyes (cube: -2.37 µm/year, radial: -2.28 µm/year; p=0.45) (2.6 years). CONCLUSION: Although the cube scan-based BMO-MRW was significantly smaller than the radial scan-based BMO-MRW, we found no significant difference between the two scan patterns for detecting glaucoma, identifying BMO location and measuring the rate of BMO-MRW change. These results suggest that although BMO-MRW estimates are not interchangeable, both scan patterns can be used for monitoring BMO-MRW changes over time.