Five-Year Clinical Trial of the Low-Concentration Atropine for Myopia Progression (LAMP) Study: Phase 4 Report.

PURPOSE: To evaluate (1) the long-term efficacy of low-concentration atropine over 5 years, (2) the proportion of children requiring re-treatment and associated factors, and (3) the efficacy of pro re nata (PRN) re-treatment using 0.05% atropine from years 3 to 5. DESIGN: Randomized, double-masked extended trial. PARTICIPANTS: Children 4 to 12 years of age originally from the Low-Concentration Atropine for Myopia Progression (LAMP) study. METHODS: Children 4 to 12 years of age originally from the LAMP study were followed up for 5 years. During the third year, children in each group originally receiving 0.05%, 0.025%, and 0.01% atropine were randomized to continued treatment and treatment cessation. During years 4 and 5, all continued treatment subgroups were switched to 0.05% atropine for continued treatment, whereas all treatment cessation subgroups followed a PRN re-treatment protocol to resume 0.05% atropine for children with myopic progressions of 0.5 diopter (D) or more over 1 year. Generalized estimating equations were used to compare the changes in spherical equivalent (SE) progression and axial length (AL) elongation among groups. MAIN OUTCOMES MEASURES: (1) Changes in SE and AL in different groups over 5 years, (2) the proportion of children who needed re-treatment, and (3) changes in SE and AL in the continued treatment and PRN re-treatment groups from years 3 to 5. RESULTS: Two hundred seventy (82.8%) of 326 children (82.5%) from the third year completed 5 years of follow-up. Over 5 years, the cumulative mean SE progressions were -1.34 ± 1.40 D, -1.97 ± 1.03 D, and -2.34 ± 1.71 D for the continued treatment groups with initial 0.05%, 0.025%, and 0.01% atropine, respectively (P = 0.02). Similar trends were observed in AL elongation (P = 0.01). Among the PRN re-treatment group, 87.9% of children (94/107) needed re-treatment. The proportion of re-treatment across all studied concentrations was similar (P = 0.76). The SE progressions for continued treatment and PRN re-treatment groups from years 3 to 5 were -0.97 ± 0.82 D and -1.00 ± 0.74 D (P = 0.55) and the AL elongations were 0.51 ± 0.34 mm and 0.49 ± 0.32 mm (P = 0.84), respectively. CONCLUSIONS: Over 5 years, the continued 0.05% atropine treatment demonstrated good efficacy for myopia control. Most children needed to restart treatment after atropine cessation at year 3. Restarted treatment with 0.05% atropine achieved similar efficacy as continued treatment. Children should be considered for re-treatment if myopia progresses after treatment cessation. FINANCIAL DISCLOSURE(S): The author(s) have no proprietary or commercial interest in any materials discussed in this article.

Analysis of Secondhand Smoke Exposure and Myopia Among Children Aged 6 to 8 Years in Hong Kong.

Importance: Secondhand smoke (SHS) exposure potentially threatens ocular health; however, its association with myopia is unknown. Objective: To examine the association between SHS exposure and childhood myopia. Design, Setting, and Participants: Cross-sectional data from the population-based Hong Kong Children Eye Study were used. Data were collected from March 5, 2015, to September 12, 2021, at The Chinese University of Hong Kong Eye Center. Participants included children aged 6 to 8 years. Secondhand smoke exposure was evaluated using a validated questionnaire. All participants underwent comprehensive ophthalmic and physical examinations. Exposure: Secondhand smoke exposure. Main Outcomes and Measures: Generalized estimating equations were constructed to examine the association of SHS exposure with spherical equivalent and axial length; logistic regression models, with myopia rate; and linear regression models, with myopia onset. Results: A total of 12 630 children (mean [SD] age, 7.37 [0.88] years; 53.2% boys) were included in the analysis. Among the participants, 4092 (32.4%) had SHS exposure. After adjusting for age, sex, parental myopia, body mass index, near-work time, outdoor time, and family income, SHS exposure was associated with greater myopic refraction (ß = -0.09 [95% CI, -0.14 to -0.03]) and longer axial length (ß = 0.05 [95% CI, 0.02-0.08]). Children with SHS exposure were more likely to develop moderate (odds ratio [OR], 1.30 [95% CI, 1.06-1.59]) and high myopia (OR, 2.64 [95% CI, 1.48-4.69]). The association of SHS exposure with spherical equivalence and axial length was magnified in younger children. For each younger year of a child's exposure to SHS, SHS exposure was associated with a 0.07-D decrease in spherical equivalence (ß = 0.07 [95% CI, 0.01-0.13]) and a 0.05-mm increase in axial length (ß = -0.05 [95% CI, -0.08 to -0.01]). Exposure to SHS was associated with an earlier mean (SD) age at onset of myopia (72.8 [0.9] vs 74.6 [0.6] months; P = .01). Every increase in SHS exposure in units of 10 cigarettes per day was associated with greater myopic refraction (ß = -0.07 [95% CI, -0.11 to -0.02]), axial length (ß = 0.04 [95% CI, 0.01-0.06]), and likelihood of developing moderate (OR, 1.23 [95% CI, 1.05-1.44]) and high myopia (OR, 1.75 [95% CI, 1.20-2.56]), and earlier myopia onset (ß = -1.30 [95% CI, -2.32 to -0.27]). Conclusions and Relevance: The findings of this cross-sectional study suggest that SHS exposure was associated with greater myopic refraction, longer axial length, greater likelihood of developing moderate and high myopia, and earlier myopia onset. The larger the quantity of SHS exposure and the younger the child, the more advanced myopia development and progression with which SHS exposure is associated.

Prevalence of Myopia in Children Before, During, and After COVID-19 Restrictions in Hong Kong.

Importance: Childhood myopia increased during the COVID-19 pandemic. Limited evidence exists about whether myopia development was reversed or worsened after the lockdown. Objective: To determine the prevalence of myopia and its associated factors before, during, and after COVID-19 restrictions. Design, Setting, and Participants: This population-based, repeated cross-sectional study evaluated children aged 6 to 8 years from the Hong Kong Children Eye Study between 2015 and 2021 in 3 cohorts: before COVID-19 (2015-2019), during COVID-19 restrictions (2020), and after COVID-19 restrictions were lifted (2021). Exposures: All the children received ocular examinations, including cycloplegic autorefraction and axial length. Data about the children's lifestyle, including time spent outdoors, near-work time, and screen time, were collected from a standardized questionnaire. Main Outcomes and Measures: The main outcomes were the prevalence of myopia, mean spherical equivalent refraction, axial length, changes in lifestyle, and the associated factors over 7 years. Data were analyzed using descriptive statistics, logistic regression, and generalized estimating equations. Results: Of 20 527 children (mean [SD] age, 7.33 [0.89] years; 52.8% boys and 47.2% girls), myopia prevalence was stable from 2015 to 2019 (23.5%-24.9%; P = .90) but increased to 28.8% (P < .001) in 2020 and 36.2% (P < .001) in 2021. The mean (SD) time spent outdoors was much lower in 2020 (0.85 [0.53] h/d; P < .001) and 2021 (1.26 [0.48] h/d; P < .001) compared with pre-COVID-19 levels (1.40 [0.47]-1.46 [0.65] h/d). The trend was reversed for total near-work time and screen time. High myopia prevalence was associated with the COVID-19 pandemic (odds ratio [OR], 1.40; 95% CI, 1.28-1.54; P < .001), younger age (OR, 1.84; 95% CI, 1.76-1.93; P < .001), male sex (OR, 1.11; 95% CI, 1.03-1.21; P = .007), lower family income (OR, 1.05; 95% CI, 1.00-1.09; P = .04), and parental myopia (OR, 1.61; 95% CI, 1.52-1.70; P < .001). During the pandemic, mean (SD) near-work and screen times in children from lower-income families were 5.16 (2.05) h/d and 3.44 (1.97) h/d, more than from higher-income families (4.83 [1.85] and 2.90 [1.61] h/d, respectively). Conclusions and Relevance: The findings of this cross-sectional study revealed that after COVID-19 restrictions were lifted in Hong Kong, myopia prevalence among children was higher than before the pandemic, and lifestyle did not return to pre-COVID-19 levels. Younger children and those from low-income families were at a higher risk of myopia development during the pandemic, suggesting that collective efforts for myopia control should be advocated for these groups.

Effect of Low-Concentration Atropine Eyedrops vs Placebo on Myopia Incidence in Children: The LAMP2 Randomized Clinical Trial.

Importance: Early onset of myopia is associated with high myopia later in life, and myopia is irreversible once developed. Objective: To evaluate the efficacy of low-concentration atropine eyedrops at 0.05% and 0.01% concentration for delaying the onset of myopia. Design, Setting, and Participants: This randomized, placebo-controlled, double-masked trial conducted at the Chinese University of Hong Kong Eye Centre enrolled 474 nonmyopic children aged 4 through 9 years with cycloplegic spherical equivalent between +1.00 D to 0.00 D and astigmatism less than -1.00 D. The first recruited participant started treatment on July 11, 2017, and the last participant was enrolled on June 4, 2020; the date of the final follow-up session was June 4, 2022. Interventions: Participants were assigned at random to the 0.05% atropine (n = 160), 0.01% atropine (n = 159), and placebo (n = 155) groups and had eyedrops applied once nightly in both eyes over 2 years. Main Outcomes and Measures: The primary outcomes were the 2-year cumulative incidence rate of myopia (cycloplegic spherical equivalent of at least -0.50 D in either eye) and the percentage of participants with fast myopic shift (spherical equivalent myopic shift of at least 1.00 D). Results: Of the 474 randomized patients (mean age, 6.8 years; 50% female), 353 (74.5%) completed the trial. The 2-year cumulative incidence of myopia in the 0.05% atropine, 0.01% atropine, and placebo groups were 28.4% (33/116), 45.9% (56/122), and 53.0% (61/115), respectively, and the percentages of participants with fast myopic shift at 2 years were 25.0%, 45.1%, and 53.9%. Compared with the placebo group, the 0.05% atropine group had significantly lower 2-year cumulative myopia incidence (difference, 24.6% [95% CI, 12.0%-36.4%]) and percentage of patients with fast myopic shift (difference, 28.9% [95% CI, 16.5%-40.5%]). Compared with the 0.01% atropine group, the 0.05% atropine group had significantly lower 2-year cumulative myopia incidence (difference, 17.5% [95% CI, 5.2%-29.2%]) and percentage of patients with fast myopic shift (difference, 20.1% [95% CI, 8.0%-31.6%]). The 0.01% atropine and placebo groups were not significantly different in 2-year cumulative myopia incidence or percentage of patients with fast myopic shift. Photophobia was the most common adverse event and was reported by 12.9% of participants in the 0.05% atropine group, 18.9% in the 0.01% atropine group, and 12.2% in the placebo group in the second year. Conclusions and Relevance: Among children aged 4 to 9 years without myopia, nightly use of 0.05% atropine eyedrops compared with placebo resulted in a significantly lower incidence of myopia and lower percentage of participants with fast myopic shift at 2 years. There was no significant difference between 0.01% atropine and placebo. Further research is needed to replicate the findings, to understand whether this represents a delay or prevention of myopia, and to assess longer-term safety. Trial Registration: Chinese Clinical Trial Registry: ChiCTR-IPR-15006883.

Deep semi-supervised multiple instance learning with self-correction for DME classification from OCT images.

Supervised deep learning has achieved prominent success in various diabetic macular edema (DME) recognition tasks from optical coherence tomography (OCT) volumetric images. A common problematic issue that frequently occurs in this field is the shortage of labeled data due to the expensive fine-grained annotations, which increases substantial difficulty in accurate analysis by supervised learning. The morphological changes in the retina caused by DME might be distributed sparsely in B-scan images of the OCT volume, and OCT data is often coarsely labeled at the volume level. Hence, the DME identification task can be formulated as a multiple instance classification problem that could be addressed by multiple instance learning (MIL) techniques. Nevertheless, none of previous studies utilize unlabeled data simultaneously to promote the classification accuracy, which is particularly significant for a high quality of analysis at the minimum annotation cost. To this end, we present a novel deep semi-supervised multiple instance learning framework to explore the feasibility of leveraging a small amount of coarsely labeled data and a large amount of unlabeled data to tackle this problem. Specifically, we come up with several modules to further improve the performance according to the availability and granularity of their labels. To warm up the training, we propagate the bag labels to the corresponding instances as the supervision of training, and propose a self-correction strategy to handle the label noise in the positive bags. This strategy is based on confidence-based pseudo-labeling with consistency regularization. The model uses its prediction to generate the pseudo-label for each weakly augmented input only if it is highly confident about the prediction, which is subsequently used to supervise the same input in a strongly augmented version. This learning scheme is also applicable to unlabeled data. To enhance the discrimination capability of the model, we introduce the Student-Teacher architecture and impose consistency constraints between two models. For demonstration, the proposed approach was evaluated on two large-scale DME OCT image datasets. Extensive results indicate that the proposed method improves DME classification with the incorporation of unlabeled data and outperforms competing MIL methods significantly, which confirm the feasibility of deep semi-supervised multiple instance learning at a low annotation cost.

Optimizing the protocol for retinal vascular permeability mapping from fluorescein videoangiography data.

An intact blood-retinal barrier is critical to maintaining the function of the retina. Many diseases of the eye have been directly associated with impairment in vascular permeability, and methods to measure vascular permeability could offer a window into early detection of disease; however, there exist no direct measures of vascular permeability that have be translated to the clinic. This work details a complete clinical workflow to quantify vascular permeability and volumetric blood flow from fluorescein videoangiography data, with validation through realistic simulations. For optimizing the protocol, this study carried on frame rate of fluorescein videoangiography to generate a high-resolution image while minimizing the error.

A MULTITASK DEEP-LEARNING SYSTEM FOR ASSESSMENT OF DIABETIC MACULAR ISCHEMIA ON OPTICAL COHERENCE TOMOGRAPHY ANGIOGRAPHY IMAGES.

PURPOSE: We aimed to develop and test a deep-learning system to perform image quality and diabetic macular ischemia (DMI) assessment on optical coherence tomography angiography (OCTA) images. METHODS: This study included 7,194 OCTA images with diabetes mellitus for training and primary validation and 960 images from three independent data sets for external testing. A trinary classification for image quality assessment and the presence or absence of DMI for DMI assessment were labeled on all OCTA images. Two DenseNet-161 models were built for both tasks for OCTA images of superficial and deep capillary plexuses, respectively. External testing was performed on three unseen data sets in which one data set using the same model of OCTA device as of the primary data set and two data sets using another brand of OCTA device. We assessed the performance by using the area under the receiver operating characteristic curves with sensitivities, specificities, and accuracies and the area under the precision-recall curves with precision. RESULTS: For the image quality assessment, analyses for gradability and measurability assessment were performed. Our deep-learning system achieved the area under the receiver operating characteristic curves >0.948 and area under the precision-recall curves >0.866 for the gradability assessment, area under the receiver operating characteristic curves >0.960 and area under the precision-recall curves >0.822 for the measurability assessment, and area under the receiver operating characteristic curves >0.939 and area under the precision-recall curves >0.899 for the DMI assessment across three external validation data sets. Grad-CAM demonstrated the capability of our deep-learning system paying attention to regions related to DMI identification. CONCLUSION: Our proposed multitask deep-learning system might facilitate the development of a simplified assessment of DMI on OCTA images among individuals with diabetes mellitus at high risk for visual loss.

Optical Coherence Tomography Classification Systems for Diabetic Macular Edema and Their Associations With Visual Outcome and Treatment Responses - An Updated Review.

ABSTRACT: Optical coherence tomography (OCT) is an invaluable imaging tool in detecting and assessing diabetic macular edema (DME). Over the past decade, there have been different proposed OCT-based classification systems for DME. In this review, we present an update of spectral-domain OCT (SDOCT)-based DME classifications over the past 5 years. In addition, we attempt to summarize the proposed OCT qualitative and quantitative parameters from different classification systems in relation to disease severity, risk of progression, and treatment outcome. Although some OCT-based measurements were found to have prognostic value on visual outcome, there has been a lack of consensus or guidelines on which parameters can be reliably used to predict treatment outcomes. We also summarize recent literatures on the prognostic value of these parameters including quantitative measures such as macular thickness or volume, central subfield thickness or foveal thickness, and qualitative features such as the morphology of the vitreoretinal interface, disorganization of retinal inner layers, ellipsoid zone disruption integrity, and hyperreflec-tive foci. In addition, we discuss that a framework to assess the validity of biomarkers for treatment outcome is essentially important in assessing the prognosis before deciding on treatment in DME. Finally, we echo with other experts on the demand for updating the current diabetic retinal disease classification.

A Multitask Deep-Learning System to Classify Diabetic Macular Edema for Different Optical Coherence Tomography Devices: A Multicenter Analysis.

OBJECTIVE: Diabetic macular edema (DME) is the primary cause of vision loss among individuals with diabetes mellitus (DM). We developed, validated, and tested a deep learning (DL) system for classifying DME using images from three common commercially available optical coherence tomography (OCT) devices. RESEARCH DESIGN AND METHODS: We trained and validated two versions of a multitask convolution neural network (CNN) to classify DME (center-involved DME [CI-DME], non-CI-DME, or absence of DME) using three-dimensional (3D) volume scans and 2D B-scans, respectively. For both 3D and 2D CNNs, we used the residual network (ResNet) as the backbone. For the 3D CNN, we used a 3D version of ResNet-34 with the last fully connected layer removed as the feature extraction module. A total of 73,746 OCT images were used for training and primary validation. External testing was performed using 26,981 images across seven independent data sets from Singapore, Hong Kong, the U.S., China, and Australia. RESULTS: In classifying the presence or absence of DME, the DL system achieved area under the receiver operating characteristic curves (AUROCs) of 0.937 (95% CI 0.920-0.954), 0.958 (0.930-0.977), and 0.965 (0.948-0.977) for the primary data set obtained from CIRRUS, SPECTRALIS, and Triton OCTs, respectively, in addition to AUROCs >0.906 for the external data sets. For further classification of the CI-DME and non-CI-DME subgroups, the AUROCs were 0.968 (0.940-0.995), 0.951 (0.898-0.982), and 0.975 (0.947-0.991) for the primary data set and >0.894 for the external data sets. CONCLUSIONS: We demonstrated excellent performance with a DL system for the automated classification of DME, highlighting its potential as a promising second-line screening tool for patients with DM, which may potentially create a more effective triaging mechanism to eye clinics.

Referral for disease-related visual impairment using retinal photograph-based deep learning: a proof-of-concept, model development study.

BACKGROUND: In current approaches to vision screening in the community, a simple and efficient process is needed to identify individuals who should be referred to tertiary eye care centres for vision loss related to eye diseases. The emergence of deep learning technology offers new opportunities to revolutionise this clinical referral pathway. We aimed to assess the performance of a newly developed deep learning algorithm for detection of disease-related visual impairment. METHODS: In this proof-of-concept study, using retinal fundus images from 15 175 eyes with complete data related to best-corrected visual acuity or pinhole visual acuity from the Singapore Epidemiology of Eye Diseases Study, we first developed a single-modality deep learning algorithm based on retinal photographs alone for detection of any disease-related visual impairment (defined as eyes from patients with major eye diseases and best-corrected visual acuity of <20/40), and moderate or worse disease-related visual impairment (eyes with disease and best-corrected visual acuity of <20/60). After development of the algorithm, we tested it internally, using a new set of 3803 eyes from the Singapore Epidemiology of Eye Diseases Study. We then tested it externally using three population-based studies (the Beijing Eye study [6239 eyes], Central India Eye and Medical study [6526 eyes], and Blue Mountains Eye Study [2002 eyes]), and two clinical studies (the Chinese University of Hong Kong's Sight Threatening Diabetic Retinopathy study [971 eyes] and the Outram Polyclinic Study [1225 eyes]). The algorithm's performance in each dataset was assessed on the basis of the area under the receiver operating characteristic curve (AUC). FINDINGS: In the internal test dataset, the AUC for detection of any disease-related visual impairment was 94·2% (95% CI 93·0-95·3; sensitivity 90·7% [87·0-93·6]; specificity 86·8% [85·6-87·9]). The AUC for moderate or worse disease-related visual impairment was 93·9% (95% CI 92·2-95·6; sensitivity 94·6% [89·6-97·6]; specificity 81·3% [80·0-82·5]). Across the five external test datasets (16 993 eyes), the algorithm achieved AUCs ranging between 86·6% (83·4-89·7; sensitivity 87·5% [80·7-92·5]; specificity 70·0% [66·7-73·1]) and 93·6% (92·4-94·8; sensitivity 87·8% [84·1-90·9]; specificity 87·1% [86·2-88·0]) for any disease-related visual impairment, and the AUCs for moderate or worse disease-related visual impairment ranged between 85·9% (81·8-90·1; sensitivity 84·7% [73·0-92·8]; specificity 74·4% [71·4-77·2]) and 93·5% (91·7-95·3; sensitivity 90·3% [84·2-94·6]; specificity 84·2% [83·2-85·1]). INTERPRETATION: This proof-of-concept study shows the potential of a single-modality, function-focused tool in identifying visual impairment related to major eye diseases, providing more timely and pinpointed referral of patients with disease-related visual impairment from the community to tertiary eye hospitals. FUNDING: National Medical Research Council, Singapore.

Detection of Diabetic Retinopathy from Ultra-Widefield Scanning Laser Ophthalmoscope Images: A Multicenter Deep Learning Analysis.

PURPOSE: To develop a deep learning (DL) system that can detect referable diabetic retinopathy (RDR) and vision-threatening diabetic retinopathy (VTDR) from images obtained on ultra-widefield scanning laser ophthalmoscope (UWF-SLO). DESIGN: Observational, cross-sectional study. PARTICIPANTS: A total of 9392 UWF-SLO images of 1903 eyes from 1022 subjects with diabetes from Hong Kong, the United Kingdom, India, and Argentina. METHODS: All images were labeled according to the presence or absence of RDR and the presence or absence of VTDR. Labeling was performed by retina specialists from fundus examination, according to the International Clinical Diabetic Retinopathy Disease Severity Scale. Three convolutional neural networks (ResNet50) were trained with a transfer-learning procedure for assessing gradability and identifying VTDR and RDR. External validation was performed on 4 datasets spanning different geographical regions. MAIN OUTCOME MEASURES: Area under the receiver operating characteristic curve (AUROC); area under the precision-recall curve (AUPRC); sensitivity, specificity, and accuracy of the DL system in gradability assessment; and detection of RDR and VTDR. RESULTS: For gradability assessment, the system achieved an AUROC of 0.923 (95% confidence interval [CI], 0.892-0.947), sensitivity of 86.5% (95% CI, 77.6-92.8), and specificity of 82.1% (95% CI, 77.3-86.2) for the primary validation dataset, and >0.82 AUROCs, >79.6% sensitivity, and >70.4% specificity for the geographical external validation datasets. For detecting RDR and VTDR, the AUROCs were 0.981 (95% CI, 0.977-0.984) and 0.966 (95% CI, 0.961-0.971), with sensitivities of 94.9% (95% CI, 92.3-97.9) and 87.2% (95% CI, 81.5-91.6), specificities of 95.1% (95% CI, 90.6-97.9) and 95.8% (95% CI, 93.3-97.6), and positive predictive values (PPVs) of 98.0% (95% CI, 96.1-99.0) and 91.1% (95% CI, 86.3-94.3) for the primary validation dataset, respectively. The AUROCs and accuracies for detecting both RDR and VTDR were >0.9% and >80%, respectively, for the geographical external validation datasets. The AUPRCs were >0.9, and sensitivities, specificities, and PPVs were >80% for the geographical external validation datasets for RDR and VTDR detection. CONCLUSIONS: The excellent performance achieved with this DL system for image quality assessment and detection of RDR and VTDR in UWF-SLO images highlights its potential as an efficient and effective diabetic retinopathy screening tool.

Global assessment of arteriolar, venular and capillary changes in normal tension glaucoma.

Microcirculatory insufficiency has been hypothesized in glaucoma pathogenesis. There is a scarcity of data to comprehensively examine the changes in retinal microvasculature and its role in normal tension glaucoma (NTG). We conducted a cross-sectional case-control study and included 168 eyes from 100 NTG patients and 68 healthy subjects. Quantitative retinal arteriolar and venular metrics were measured from retinal photographs using a computer-assisted program. Radial peripapillary capillary network was imaged with OCT-A and quantitative capillary metrics (circumpapillary vessel density (cpVD) and circumpapillary fractal dimension (cpFD)) were measured with a customized MATLAB program. We found that NTG was associated with decreased arteriolar and venular tortuosity, arteriolar branching angle, cpVD and cpFD. Decreased venular caliber, arteriolar and venular branching angles, cpVD and cpFD were associated with thinner average RNFL thickness. Decreased arteriolar and venular branching angles, cpVD and cpFD were also associated with worse standard automated perimetry measurements (mean deviation and visual field index). Compared with retinal arteriolar and venular metrics, regression models based on OCT-A capillary metrics consistently showed stronger associations with NTG and structural and functional measurements in NTG. We concluded that NTG eyes showed generalized microvascular attenuations, in which OCT-A capillary metrics attenuations were more prominent and strongly associated with NTG.

UD-MIL: Uncertainty-Driven Deep Multiple Instance Learning for OCT Image Classification.

Deep learning has achieved remarkable success in the optical coherence tomography (OCT) image classification task with substantial labelled B-scan images available. However, obtaining such fine-grained expert annotations is usually quite difficult and expensive. How to leverage the volume-level labels to develop a robust classifier is very appealing. In this paper, we propose a weakly supervised deep learning framework with uncertainty estimation to address the macula-related disease classification problem from OCT images with the only volume-level label being available. First, a convolutional neural network (CNN) based instance-level classifier is iteratively refined by using the proposed uncertainty-driven deep multiple instance learning scheme. To our best knowledge, we are the first to incorporate the uncertainty evaluation mechanism into multiple instance learning (MIL) for training a robust instance classifier. The classifier is able to detect suspicious abnormal instances and abstract the corresponding deep embedding with high representation capability simultaneously. Second, a recurrent neural network (RNN) takes instance features from the same bag as input and generates the final bag-level prediction by considering the individually local instance information and globally aggregated bag-level representation. For more comprehensive validation, we built two large diabetic macular edema (DME) OCT datasets from different devices and imaging protocols to evaluate the efficacy of our method, which are composed of 30,151 B-scans in 1,396 volumes from 274 patients (Heidelberg-DME dataset) and 38,976 B-scans in 3,248 volumes from 490 patients (Triton-DME dataset), respectively. We compare the proposed method with the state-of-the-art approaches, and experimentally demonstrate that our method is superior to alternative methods, achieving volume-level accuracy, F1-score and area under the receiver operating characteristic curve (AUC) of 95.1%, 0.939 and 0.990 on Heidelberg-DME and those of 95.1%, 0.935 and 0.986 on Triton-DME, respectively. Furthermore, the proposed method also yields competitive results on another public age-related macular degeneration OCT dataset, indicating the high potential as an effective screening tool in the clinical practice.

OCT Angiography Metrics Predict Progression of Diabetic Retinopathy and Development of Diabetic Macular Edema: A Prospective Study.

PURPOSE: To prospectively determine the relationship of OCT angiography (OCTA) metrics to diabetic retinopathy (DR) progression and development of diabetic macular edema (DME). DESIGN: Prospective, observational study. PARTICIPANTS: A total of 205 eyes from 129 patients with diabetes mellitus followed up for at least 2 years. METHODS: All participants underwent OCTA with a swept-source OCT device (DRI-OCT Triton, Topcon, Inc, Tokyo, Japan). Individual OCTA images of superficial capillary plexus (SCP) and deep capillary plexus (DCP) were generated by IMAGEnet6 (Basic License 10). After a quality check, automated measurements of foveal avascular zone (FAZ) area, FAZ circularity, vessel density (VD), and fractal dimension (FD) of both SCP and DCP were then obtained. MAIN OUTCOME MEASURES: Progression of DR and development of DME. RESULTS: Over a median follow-up of 27.14 months (interquartile range, 24.16-30.41 months), 28 of the 205 eyes (13.66%) developed DR progression. Of the 194 eyes without DME at baseline, 17 (8.76%) developed DME. Larger FAZ area (hazard ratio [HR], 1.829 per SD increase; 95% confidence interval [CI], 1.332-2.512), lower VD (HR, 1.908 per SD decrease; 95% CI, 1.303-2.793), and lower FD (HR, 4.464 per SD decrease; 95% CI, 1.337-14.903) of DCP were significantly associated with DR progression after adjusting for established risk factors (DR severity, glycated hemoglobin, duration of diabetes, age, and mean arterial blood pressure at baseline). Lower VD of SCP (HR, 1.789 per SD decrease; 95% CI, 1.027-4.512) was associated with DME development. Compared with the model with established risk factors alone, the addition of OCTA metrics improved the predictive discrimination of DR progression (FAZ area of DCP, C-statistics 0.723 vs. 0.677, P < 0.001; VD of DCP, C-statistics 0.727 vs. 0.677, P = 0.001; FD of DCP, C-statistics 0.738 vs. 0.677, P < 0.001) and DME development (VD of SCP, C-statistics 0.904 vs. 0.875, P = 0.036). CONCLUSIONS: The FAZ area, VD, and FD of DCP predict DR progression, whereas VD of SCP predicts DME development. Our findings provide evidence to support that OCTA metrics improve the evaluation of risk of DR progression and DME development beyond traditional risk factors.

Artificial Intelligence in Diabetic Eye Disease Screening.

Systematic or national screening programs for diabetic retinopathy (DR) and diabetic macular edema (DME), using digital fundus photography and optical coherence tomography (OCT), are currently implemented at primary care level, aiming to provide timely referral for vision-threatening DR and DME to ophthalmologists for timely treatment and vision loss prevention. However, interpretation of retinal images requires specialized knowledge and expertise in diabetic eye disease. Furthermore, current DR screening programs are capital- and labor-intensive, which makes it difficult to rapidly scale up and expand diabetic eye screening to meet the needs of this growing global epidemic. Deep learning (DL), a new branch of machine learning technology under the broad term of artificial intelligence (AI), has made remarkable breakthrough in medical imaging in particular for pattern recognition and image classification. In ophthalmology, AI and DL technology has been developed from big image datasets in assessment of retinal photographs for detection and screening of DR as well as the segmentation and assessment of OCT images for diagnosis and screening of DME. This review aimed to summarize the current progress and the development of using AI and DL technology for diabetic eye disease screening as well as current challenges in the actual implementation of DL in screening programs, and translating DL research into direct clinical applications of screening in a community setting.

Non-mydriatic ultrawide field scanning laser ophthalmoscopy compared with dilated fundal examination for assessment of diabetic retinopathy and diabetic macular oedema in Chinese individuals with diabetes mellitus.

AIMS: To evaluate the performance of ultrawide field scanning laser ophthalmoscopy (UWF-SLO) for assessing diabetic retinopathy (DR) and diabetic macular oedema (DME) in a Chinese population, compared with clinical examination. METHODS: This is a retrospective cohort study. A series of 322 eyes from 164 patients with DM were included. Each patient underwent both dilated fundal examination with DR and DME grading by retina specialist and non-mydriatic 200° UWF-SLO (Daytona, Optos, Dunfermline, UK). The severity of DR and DME from UWF-SLO images was further graded by ophthalmologists, according to both international clinical DR and DME disease severity scales and the standard 7-field Early Treatment Diabetic Retinopathy Study (ETDRS) scale. Any DR, DME and vision-threatening DR (VTDR) were treated as endpoints for this study. RESULTS: 23 out of 322 images (7.14%), including all four cases with proliferative DR on clinical examinations, were determined as ungradable. When the international scale was used for grading UWF-SLO images, the sensitivity of any DR, DME and VTDR was 67.7%, 67.4% and 72.6%, respectively; the specificity of any DR, DME and VTDR was 97.8%, 97.3% and 97.8%, respectively. The agreement with clinical grading in picking up any DR, DME and VTDR was substantial, with κ-values of 0.634, 0.694 and 0.707, respectively. The performance of UWF-SLO was shown to be lower when ETDRS scale was used for grading the images. CONCLUSION: The performance of non-mydriatic UWF-SLO is comparable in identifying DR with that of clinical examination in a Chinese cohort. However, whether UWF-SLO can be considered as tool for screening DR is still undetermined.

Relationship of intercapillary area with visual acuity in diabetes mellitus: an optical coherence tomography angiography study.

AIM: To examine the correlation of best-corrected visual acuity (BCVA) with intercapillary area (ICA) measured from optical coherence tomography angiography (OCT-A) in patients with diabetes, and to compare the strength of associations between BCVA with ICA and other OCT-A metrics. METHODS: A cross-sectional study involved 447 eyes from 299 patients with diabetes. All participants underwent OCT-A with a swept-source OCT (Triton DRI-OCT, Topcon, Tokyo, Japan). An automated customised MATLAB programme was used to quantify ICA (the mean of the 10 largest areas including foveal avascular zone (FAZ) area (ICA10_FAZ) and excluding FAZ area (ICA10_excFAZ)) and other OCT-A metrics (FAZ area, FAZ circularity and vessel density) from the macular OCT-A images. BCVA was measured using Snellen chart for the patients and then converted to logarithm of the minimum angle of resolution (logMAR) VA. We further defined 'good VA' as Snellen >0.7 and 'poor VA' as Snellen ≤0.7 as a binary VA outcome for logistic regression analysis. RESULTS: In univariate regression analysis, increased ICA10_FAZ and ICA10_excFAZ were significantly correlated with logMAR (p values <0.05). In multivariate regression analysis, only the association between ICA10_FAZ and logMAR persisted (ß=0.103, p=0.024). In multivariable logistic regression analysis, increased ICA10_FAZ (OR=1.300, 95% CI 1.076 to 1.679, p=0.044) and FAZ circularity (OR=1.285, 95% CI 1.031 to 1.603, p=0.026) showed significant associations with poor VA. CONCLUSIONS: Increased ICA measured from OCT-A, describing enlargement of capillary rarefaction or closure at macular area, is independently associated with BCVA, suggesting that ICA is a potential marker to quantify retinal microvascular abnormalities relating to vision among individuals with diabetes.

Factors Associated With Long-term Intraocular Pressure Fluctuation in Primary Angle Closure Disease: The CUHK PACG Longitudinal (CUPAL) Study.

PURPOSE: To determine the demographic, ocular, and systemic factors associated with long-term intraocular pressure (IOP) fluctuation in primary angle closure disease (PACD). METHODS: This prospective cohort study included 422 PACD eyes from 269 Chinese patients, including 274 primary angle closure glaucoma (PACG) eyes and 152 primary angle closure/primary angle closure suspect (PAC/PACS) eyes. Long-term IOP fluctuation defined as the SD of all IOP measurements over 2 years (at least 5 measurements in total). Chinese patients with PACD were recruited and followed up 3 monthly. Eyes with IOP-lowering surgery or lens extraction performed within the 2-year study period were excluded. Patient demographics, received treatments, ocular biometry, retinal nerve fiber layer thickness, and systemic factors (eg, hypertension, smoking) were evaluated. Generalized estimating equations adjusting for inter-eye correlation were used to determine the associations. RESULTS: Eyes with PACG had significantly higher IOP fluctuation than PAC/PACS (2.4±1.2 versus 2.1±0.9 mm Hg; P=0.04). In the multivariate analysis with PACG eyes, higher baseline IOP (P<0.001), greater number of IOP-lowering medications (P<0.001), previous trabeculectomy (P=0.002), and current smoking (P=0.03) were significantly associated with larger IOP fluctuation, whereas diabetes mellitus was associated with lower IOP fluctuation (P=0.03). Among PAC/PACS eyes, younger age group (P<0.001), male sex (P=0.002), and higher baseline IOP (P<0.001) were significantly associated with larger IOP fluctuation. CONCLUSIONS: PACG eyes have greater IOP fluctuation than PAC/PACS eyes. Certain demographic, ocular, and systemic factors are associated with IOP fluctuation in PACD eyes.