Automated Diabetic Retinopathy Image Assessment Software: Diagnostic Accuracy and Cost-Effectiveness Compared with Human Graders

Different fundus imaging modalities and technical factors in AI screening for diabetic retinopathy: a review

BACKGROUND

Effective screening is a desirable method for the early detection and successful treatment for diabetic retinopathy, and fundus photography is currently the dominant medium for retinal imaging due to its convenience and accessibility. Manual screening using fundus photographs has however involved considerable costs for patients, clinicians and national health systems, which has limited its application particularly in less-developed countries. The advent of artificial intelligence, and in particular deep learning techniques, has however raised the possibility of widespread automated screening.

MAIN TEXT

In this review, we first briefly survey major published advances in retinal analysis using artificial intelligence. We take care to separately describe standard multiple-field fundus photography, and the newer modalities of ultra-wide field photography and smartphone-based photography. Finally, we consider several machine learning concepts that have been particularly relevant to the domain and illustrate their usage with extant works.

CONCLUSIONS

In the ophthalmology field, it was demonstrated that deep learning tools for diabetic retinopathy show clinically acceptable diagnostic performance when using colour retinal fundus images. Artificial intelligence models are among the most promising solutions to tackle the burden of diabetic retinopathy management in a comprehensive manner. However, future research is crucial to assess the potential clinical deployment, evaluate the cost-effectiveness of different DL systems in clinical practice and improve clinical acceptance.

Assessment of Training Outcomes of Nurse Readers for Diabetic Retinopathy Telescreening: Validation Study

BACKGROUND

With the high prevalence of diabetic retinopathy and its significant visual consequences if untreated, timely identification and management of diabetic retinopathy is essential. Teleophthalmology programs have assisted in screening a large number of individuals at risk for vision loss from diabetic retinopathy. Training nonophthalmological readers to assess remote fundus images for diabetic retinopathy may further improve the efficiency of such programs.

OBJECTIVE

This study aimed to evaluate the performance, safety implications, and progress of 2 ophthalmology nurses trained to read and assess diabetic retinopathy fundus images within a hospital diabetic retinopathy telescreening program.

METHODS

In this retrospective interobserver study, 2 ophthalmology nurses followed a specific training program within a hospital diabetic retinopathy telescreening program and were trained to assess diabetic retinopathy images at 2 levels of intervention: detection of diabetic retinopathy (level 1) and identification of referable disease (level 2). The reliability of the assessment by level 1-trained readers in 266 patients and of the identification of patients at risk of vision loss from diabetic retinopathy by level 2-trained readers in 559 more patients were measured. The learning curve, sensitivity, and specificity of the readings were evaluated using a group consensus gold standard.

RESULTS

An almost perfect agreement was measured in identifying the presence of diabetic retinopathy in both level 1 readers (κ=0.86 and 0.80) and in identifying referable diabetic retinopathy by level 2 readers (κ=0.80 and 0.83). At least substantial agreement was measured in the level 2 readers for macular edema (κ=0.79 and 0.88) for all eyes. Good screening threshold sensitivities and specificities were obtained for all level readers, with sensitivities of 90.6% and 96.9% and specificities of 95.1% and 85.1% for level 1 readers (readers A and B) and with sensitivities of 86.8% and 91.2% and specificities of 91.7% and 97.0% for level 2 readers (readers A and B). This performance was achieved immediately after training and remained stable throughout the study.

CONCLUSIONS

Notwithstanding the small number of trained readers, this study validates the screening performance of level 1 and level 2 diabetic retinopathy readers within this training program, emphasizing practical experience, and allows the establishment of an ongoing assessment clinic. This highlights the importance of supervised, hands-on experience and may help set parameters to further calibrate the training of diabetic retinopathy readers for safe screening programs.

Big Data and diabetes: the applications of Big Data for diabetes care now and in the future

We review current applications of Big Data in diabetes care and consider the future potential by carrying out a scoping study of the academic literature on Big Data and diabetes care. Healthcare data are being produced at ever-increasing rates, and this information has the potential to transform the provision of diabetes care. Big Data is beginning to have an impact on diabetes care through data research. The use of Big Data for routine clinical care is still a future application. Vast amounts of healthcare data are already being produced, and the key is harnessing these to produce actionable insights. Considerable development work is required to achieve these goals.

Development and Evaluation of a Deep Learning System for Screening Retinal Hemorrhage Based on Ultra-Widefield Fundus Images

Purpose

To develop and evaluate a deep learning (DL) system for retinal hemorrhage (RH) screening using ultra-widefield fundus (UWF) images.

Methods

A total of 16,827 UWF images from 11,339 individuals were used to develop the DL system. Three experienced retina specialists were recruited to grade UWF images independently. Three independent data sets from 3 different institutions were used to validate the effectiveness of the DL system. The data set from Zhongshan Ophthalmic Center (ZOC) was selected to compare the classification performance of the DL system and general ophthalmologists. A heatmap was generated to identify the most important area used by the DL model to classify RH and to discern whether the RH involved the anatomical macula.

Results

In the three independent data sets, the DL model for detecting RH achieved areas under the curve of 0.997, 0.998, and 0.999, with sensitivities of 97.6%, 96.7%, and 98.9% and specificities of 98.0%, 98.7%, and 99.4%. In the ZOC data set, the sensitivity of the DL model was better than that of the general ophthalmologists, although the general ophthalmologists had slightly higher specificities. The heatmaps highlighted RH regions in all true-positive images, and the RH within the anatomical macula was determined based on heatmaps.

Conclusions

Our DL system showed reliable performance for detecting RH and could be used to screen for RH-related diseases.

Translational Relevance

As a screening tool, this automated system may aid early diagnosis and management of RH-related retinal and systemic diseases by allowing timely referral.

Deep learning for detecting retinal detachment and discerning macular status using ultra-widefield fundus images

Retinal detachment can lead to severe visual loss if not treated timely. The early diagnosis of retinal detachment can improve the rate of successful reattachment and the visual results, especially before macular involvement. Manual retinal detachment screening is time-consuming and labour-intensive, which is difficult for large-scale clinical applications. In this study, we developed a cascaded deep learning system based on the ultra-widefield fundus images for automated retinal detachment detection and macula-on/off retinal detachment discerning. The performance of this system is reliable and comparable to an experienced ophthalmologist. In addition, this system can automatically provide guidance to patients regarding appropriate preoperative posturing to reduce retinal detachment progression and the urgency of retinal detachment repair. The implementation of this system on a global scale may drastically reduce the extent of vision impairment resulting from retinal detachment by providing timely identification and referral.

Li et al. develop a cascaded deep learning system for automated retinal detachment and macular status detection based on ultra-widefield fundus (UWF) images. With reliable and comparable performance to an experienced opthamologist, this system can also provide guidance to patients regarding appropriate preoperative posturing to reduce RD progression.

Applications of Artificial Intelligence for the Detection, Management, and Treatment of Diabetic Retinopathy

Rates of diabetic retinopathy (DR) and diabetic macular edema (DME), a common ocular complication of diabetes mellitus, are increasing worldwide. There is a substantial burden concerning the detection and management of this condition, particularly in low-resource settings, due to limitations such as the time, cost, and labor associated with current screening and treatment methods. Artificial intelligence (AI) is a modality of pattern recognition that has the potential to combat these limitations in a reliable and cost-effective way. This review explores the various applications of AI on the screening, management, and treatment of DR and DME. AI applications for detecting referable DR and DME have been the most thoroughly researched applications for this condition. While some studies exist using AI to stratify DR patients based on the risk of progression, predict treatment outcomes to anti-VEGF therapy, and explore the utilization of AI for clinical trials to develop new treatments for DR, further validation studies on larger datasets are warranted.

Simple, Mobile-based Artificial Intelligence Algoithm in the detection of Diabetic Retinopathy (SMART) study

INTRODUCTION

The aim of this study is to evaluate the performance of the offline smart phone-based Medios artificial intelligence (AI) algorithm in the diagnosis of diabetic retinopathy (DR) using non-mydriatic (NM) retinal images.

METHODS

This cross-sectional study prospectively enrolled 922 individuals with diabetes mellitus. NM retinal images (disc and macula centered) from each eye were captured using the Remidio NM fundus-on-phone (FOP) camera. The images were run offline and the diagnosis of the AI was recorded (DR present or absent). The diagnosis of the AI was compared with the image diagnosis of five retina specialists (majority diagnosis considered as ground truth).

RESULTS

Analysis included images from 900 individuals (252 had DR). For any DR, the sensitivity and specificity of the AI algorithm was found to be 83.3% (95% CI 80.9% to 85.7%) and 95.5% (95% CI 94.1% to 96.8%). The sensitivity and specificity of the AI algorithm in detecting referable DR (RDR) was 93% (95% CI 91.3% to 94.7%) and 92.5% (95% CI 90.8% to 94.2%).

CONCLUSION

The Medios AI has a high sensitivity and specificity in the detection of RDR using NM retinal images.

The Value of Automated Diabetic Retinopathy Screening with the EyeArt System: A Study of More Than 100,000 Consecutive Encounters from People with Diabetes

Background: Current manual diabetic retinopathy (DR) screening using eye care experts cannot scale to screen the growing population of diabetes patients who are at risk for vision loss. EyeArt system is an automated, cloud-based artificial intelligence (AI) eye screening technology designed to easily detect referral-warranted DR immediately through automated analysis of patient's retinal images. Methods: This retrospective study assessed the diagnostic efficacy of the EyeArt system v2.0 analyzing 850,908 fundus images from 101,710 consecutive patient visits, collected from 404 primary care clinics. Presence or absence of referral-warranted DR (more than mild nonproliferative DR [NPDR]) was automatically detected by the EyeArt system for each patient encounter, and its performance was compared against a clinical reference standard of quality-assured grading by rigorously trained certified ophthalmologists and optometrists. Results: Of the 101,710 visits, 75.7% were nonreferable, 19.3% were referable to an eye care specialist, and in 5.0%, the DR level was unknown as per the clinical reference standard. EyeArt screening had 91.3% (95% confidence interval [CI]: 90.9-91.7) sensitivity and 91.1% (95% CI: 90.9-91.3) specificity. For 5446 encounters with potentially treatable DR (more than moderate NPDR and/or diabetic macular edema), the system provided a positive "refer" output to 5363 encounters achieving sensitivity of 98.5%. Conclusions: This study captures variations in real-world clinical practice and shows that an AI DR screening system can be safe and effective in the real world. This study demonstrates the value of this easy-to-use, automated tool for endocrinologists, diabetologists, and general practitioners to address the growing need for DR screening and monitoring.

Comparison of Subjective Assessment and Precise Quantitative Assessment of Lesion Distribution in Diabetic Retinopathy

Importance

Predominantly peripheral disease in eyes with nonproliferative diabetic retinopathy (DR) is suggested as a potential strong risk factor for progression to proliferative disease. However, the reliability and optimal method for the assessment of lesion distribution are still uncertain.

Objective

To compare agreement between subjective assessment and precise quantification of lesion burden in ultrawidefield (UWF) images of eyes with DR.

Design, Setting, and Participants

This multisite cross-sectional study examines UWF pseudocolor images acquired from DR screening clinic patients from December 20, 2014, through August 1, 2014. Of 104 cases, 161 eyes with DR were included. Data analysis was conducted from June 1, 2016, through December 1, 2016 at the Doheny Image Reading Center.

Main Outcomes and Measures

Distribution of DR lesions in eyes was assessed subjectively and quantitatively, and eyes were classified as having predominantly central lesions (PCLs) or predominantly peripheral lesions (PPLs). The frequency and surface area (SA) of each lesion type were quantified. Intergrader and subjective vs quantitative classification were compared for level of agreement. Several methods of determining PPL distribution were also compared.

Results

On subjective frequency-based evaluation by graders, 133 eyes were classified as having PCL, and 28 eyes as having PPL. On exact quantification of lesion SA, 121 eyes were classified as PCL, and 40 eyes as having PPL. On SA-based quantification, 134 eyes were classified as having PCL, and 27 eyes as having PPL. There was a significant difference between qualitative and quantitative classification of DR lesion distribution for both frequency-based (mean difference [SD]: PCL, 6 [2]; PPL, 13 [6]; P < .001) and SA-based (mean difference [SD]: PCL, 6 [1]; PPL, 20 [7]; P < .001) methods. Both intergrader reproducibility and subjective vs quantitative agreement were higher with frequency-based classification.

Conclusions and Relevance

Subjective assessment of PPL DR lesions on UWF images differed in some cases from precise quantitative assessments, particularly when considering the area of lesions. These findings highlight the benefit of objective quantitative approaches to DR assessment, which may facilitate the development of a more precise DR scoring system.

Imaging and Biomarkers in Diabetic Macular Edema and Diabetic Retinopathy

PURPOSE OF REVIEW

Diabetic retinopathy (DR) is the leading cause of acquired vision loss in adults across the globe. Early identification and treatment of patients with DR is paramount for vision preservation. The aim of this review paper is to outline current and new imaging techniques and biomarkers that are valuable for clinical diagnosis and management of DR.

RECENT FINDINGS

Ultrawide field imaging and automated deep learning algorithms are recent advancements on traditional fundus photography and fluorescein angiography. Optical coherence tomography (OCT) and OCT angiography are techniques that image retinal anatomy and vasculature and OCT is routinely used to monitor response to treatment. Many circulating, vitreous, and genetic biomarkers have been studied to facilitate disease detection and development of new treatments. Recent advancements in retinal imaging and identification of promising new biomarkers for DR have the potential to increase detection, risk stratification, and treatment for patients with DR.

The diagnostic accuracy of an intelligent and automated fundus disease image assessment system with lesion quantitative function (SmartEye) in diabetic patients

BACKGROUND

With the diabetes mellitus (DM) prevalence increasing annually, the human grading of retinal images to evaluate DR has posed a substantial burden worldwide. SmartEye is a recently developed fundus image processing and analysis system with lesion quantification function for DR screening. It is sensitive to the lesion area and can automatically identify the lesion position and size. We reported the diabetic retinopathy (DR) grading results of SmartEye versus ophthalmologists in analyzing images captured with non-mydriatic fundus cameras in community healthcare centers, as well as DR lesion quantitative analysis results on different disease stages.

METHODS

This is a cross-sectional study. All the fundus images were collected from the Shanghai Diabetic Eye Study in Diabetics (SDES) program from Apr 2016 to Aug 2017. 19,904 fundus images were acquired from 6013 diabetic patients. The grading results of ophthalmologists and SmartEye are compared. Lesion quantification of several images at different DR stages is also presented.

RESULTS

The sensitivity for diagnosing no DR, mild NPDR (non-proliferative diabetic retinopathy), moderate NPDR, severe NPDR, PDR (proliferative diabetic retinopathy) are 86.19, 83.18, 88.64, 89.59, and 85.02%. The specificity are 63.07, 70.96, 64.16, 70.38, and 74.79%, respectively. The AUC are PDR, 0.80 (0.79, 0.81); severe NPDR, 0.80 (0.79, 0.80); moderate NPDR, 0.77 (0.76, 0.77); and mild NPDR, 0.78 (0.77, 0.79). Lesion quantification results showed that the total hemorrhage area, maximum hemorrhage area, total exudation area, and maximum exudation area increase with DR severity.

CONCLUSIONS

SmartEye has a high diagnostic accuracy in DR screening program using non-mydriatic fundus cameras. SmartEye quantitative analysis may be an innovative and promising method of DR diagnosis and grading.

Artificial Intelligence Screening for Diabetic Retinopathy: the Real-World Emerging Application

PURPOSE OF REVIEW

This paper systematically reviews the recent progress in diabetic retinopathy screening. It provides an integrated overview of the current state of knowledge of emerging techniques using artificial intelligence integration in national screening programs around the world. Existing methodological approaches and research insights are evaluated. An understanding of existing gaps and future directions is created.

RECENT FINDINGS

Over the past decades, artificial intelligence has emerged into the scientific consciousness with breakthroughs that are sparking increasing interest among computer science and medical communities. Specifically, machine learning and deep learning (a subtype of machine learning) applications of artificial intelligence are spreading into areas that previously were thought to be only the purview of humans, and a number of applications in ophthalmology field have been explored. Multiple studies all around the world have demonstrated that such systems can behave on par with clinical experts with robust diagnostic performance in diabetic retinopathy diagnosis. However, only few tools have been evaluated in clinical prospective studies. Given the rapid and impressive progress of artificial intelligence technologies, the implementation of deep learning systems into routinely practiced diabetic retinopathy screening could represent a cost-effective alternative to help reduce the incidence of preventable blindness around the world.

Development and validation of a deep-learning algorithm for the detection of neovascular age-related macular degeneration from colour fundus photographs

IMPORTANCE

Detection of early onset neovascular age-related macular degeneration (AMD) is critical to protecting vision.

BACKGROUND

To describe the development and validation of a deep-learning algorithm (DLA) for the detection of neovascular age-related macular degeneration.

DESIGN

Development and validation of a DLA using retrospective datasets.

PARTICIPANTS

We developed and trained the DLA using 56 113 retinal images and an additional 86 162 images from an independent dataset to externally validate the DLA. All images were non-stereoscopic and retrospectively collected.

METHODS

The internal validation dataset was derived from real-world clinical settings in China. Gold standard grading was assigned when consensus was reached by three individual ophthalmologists. The DLA classified 31 247 images as gradable and 24 866 as ungradable (poor quality or poor field definition). These ungradable images were used to create a classification model for image quality. Efficiency and diagnostic accuracy were tested using 86 162 images derived from the Melbourne Collaborative Cohort Study. Neovascular AMD and/or ungradable outcome in one or both eyes was considered referable.

MAIN OUTCOME MEASURES

Area under the receiver operating characteristic curve (AUC), sensitivity and specificity.

RESULTS

In the internal validation dataset, the AUC, sensitivity and specificity of the DLA for neovascular AMD was 0.995, 96.7%, 96.4%, respectively. Testing against the independent external dataset achieved an AUC, sensitivity and specificity of 0.967, 100% and 93.4%, respectively. More than 60% of false positive cases displayed other macular pathologies. Amongst the false negative cases (internal validation dataset only), over half (57.2%) proved to be undetected detachment of the neurosensory retina or RPE layer.

CONCLUSIONS AND RELEVANCE

This DLA shows robust performance for the detection of neovascular AMD amongst retinal images from a multi-ethnic sample and under different imaging protocols. Further research is warranted to investigate where this technology could be best utilized within screening and research settings.

Multiple Effects of Intravitreal Aflibercept on Microvascular Regression in Eyes with Diabetic Macular Edema

PURPOSE

To evaluate the effects of intravitreal aflibercept (IVA) on the number of microaneurysms and sizes of nonperfused areas (NPAs) in eyes with diabetic macular edema (DME).

DESIGN

Interventional, prospective study.

PARTICIPANTS

Twenty-five eyes of 25 DME patients (average age, 64.0±8.8 years) were treated with 3 consecutive monthly IVA injections.

METHODS

Fluorescein angiography (FA) and OCT were performed before the IVA injections (baseline) and at 1 week after the IVA treatment. The number of microaneurysms and the ischemic index (ISI), a measure of NPA, were determined. The correlations between central retinal thickness (CRT) and number of microaneurysms and the ISI were also determined.

MAIN OUTCOME MEASURES

The mean number of microaneurysms and NPA evaluated as the ISI.

RESULTS

At baseline, the mean CRT was 485.7±90.6 μm. After treatment, the mean CRT was reduced significantly to 376.9±81.6 μm (P = 0.1 × 10^-5, repeated analysis of variance). The mean number of microaneurysms was decreased significantly from 49.6±33.2 at baseline to 24.8±18.1 at 3 months after the initial treatment. This was a 50.4±21.2% reduction (P = 0.3 × 10^-5, paired t test). The mean ISI was also decreased significantly from 55.5±20.4% at baseline to 28.8±16.8% after treatment (P = 0.3 × 10^-5, paired t test). This was a reduction of 43.3±28.5%. A significant correlation was found between the CRT and number of microaneurysms at both baseline (r = 0.56; P = 0.004) and after treatment (r = 0.53; P = 0.006). A significant correlation was found between CRT and ISI at baseline (r = -0.39; P = 0.03) but not after treatment (r = -0.06; P = 0.79).

CONCLUSIONS

The reduction in the number of microaneurysms was correlated with reduction in CRT.

A Review of Machine Learning Techniques for Keratoconus Detection and Refractive Surgery Screening

Various machine learning techniques have been developed for keratoconus detection and refractive surgery screening. These techniques utilize inputs from a range of corneal imaging devices and are built with automated decision trees, support vector machines, and various types of neural networks. In general, these techniques demonstrate very good differentiation of normal and keratoconic eyes, as well as good differentiation of normal and form fruste keratoconus. However, it is difficult to directly compare these studies, as keratoconus represents a wide spectrum of disease. More importantly, no public dataset exists for research purposes. Despite these challenges, machine learning in keratoconus detection and refractive surgery screening is a burgeoning field of study, with significant potential for continued advancement as imaging devices and techniques become more sophisticated.

Digital image processing software for diagnosing diabetic retinopathy from fundus photograph

Objective

The aim of this study was to develop automated software for screening and diagnosing diabetic retinopathy (DR) from fundus photograph of patients with diabetes mellitus.

Methods

The extraction of clinically significant features to detect pathologies of DR and the severity classification were performed by using MATLAB R2015a with MATLAB Image Processing Toolbox. In addition, the graphic user interface was developed using the MATLAB GUI Toolbox. The accuracy of software was measured by comparing the obtained results to those of the diagnosis by the ophthalmologist.

Results

A set of 400 fundus images, containing 21 normal fundus images and 379 DR fundus images (162 non-proliferative DR and 217 proliferative DR), was interpreted by the ophthalmologist as a reference standard. The initial result showed that the sensitivity, specificity and accuracy of this software in detection of DR were 98%, 67% and 96.25%, respectively. However, the accuracy of this software in classifying non-proliferative and proliferative diabetic retinopathy was 66.58%. The average time for processing is 7 seconds for one fundus image.

Conclusion

The automated DR screening software was developed by using MATLAB programming and yielded 96.25% accuracy for the detection of DR when compared to that of the diagnosis by the ophthalmologist. It may be a helpful tool for DR screening in the distant rural area where ophthalmologist is not available.

[Can Big Data change our practices?]

The European Medicines Agency has defined Big Data by the "3 V's": Volume, Velocity and Variety. These large databases allow access to real life data on patient care. They are particularly suited for studies of adverse events and pharmacoepidemiology. Deep learning is a collection of algorithms used in machine learning, used to model high-level abstractions in data using model architectures, which are composed of multiple nonlinear transformations. This article shows how Big Data and Deep Learning can help in ophthalmology, pointing out their advantages and disadvantages. A literature review is presented in this article illustrating the uses of Deep Learning in ophthalmology.

Diagnostic Efficacy and Therapeutic Decision-making Capacity of an Artificial Intelligence Platform for Childhood Cataracts in Eye Clinics: A Multicentre Randomized Controlled Trial

BACKGROUND

CC-Cruiser is an artificial intelligence (AI) platform developed for diagnosing childhood cataracts and providing risk stratification and treatment recommendations. The high accuracy of CC-Cruiser was previously validated using specific datasets. The objective of this study was to compare the diagnostic efficacy and treatment decision-making capacity between CC-Cruiser and ophthalmologists in real-world clinical settings.

METHODS

This multicentre randomized controlled trial was performed in five ophthalmic clinics in different areas across China. Pediatric patients (aged ≤ 14 years) without a definitive diagnosis of cataracts or history of previous eye surgery were randomized (1:1) to receive a diagnosis and treatment recommendation from either CC-Cruiser or senior consultants (with over 5 years of clinical experience in pediatric ophthalmology). The experts who provided a gold standard diagnosis, and the investigators who performed slit-lamp photography and data analysis were blinded to the group assignments. The primary outcome was the diagnostic performance for childhood cataracts with reference to cataract experts' standards. The secondary outcomes included the evaluation of disease severity and treatment determination, the time required for the diagnosis, and patient satisfaction, which was determined by the mean rating. This trial is registered with ClinicalTrials.gov (NCT03240848).

FINDINGS

Between August 9, 2017 and May 25, 2018, 350 participants (700 eyes) were randomly assigned for diagnosis by CC-Cruiser (350 eyes) or senior consultants (350 eyes). The accuracies of cataract diagnosis and treatment determination were 87.4% and 70.8%, respectively, for CC-Cruiser, which were significantly lower than 99.1% and 96.7%, respectively, for senior consultants (p < 0.001, OR = 0.06 [95% CI 0.02 to 0.19]; and p < 0.001, OR = 0.08 [95% CI 0.03 to 0.25], respectively). The mean time for receiving a diagnosis from CC-Cruiser was 2.79 min, which was significantly less than 8.53 min for senior consultants (p < 0.001, mean difference 5.74 [95% CI 5.43 to 6.05]). The patients were satisfied with the overall medical service quality provided by CC-Cruiser, typically with its time-saving feature in cataract diagnosis.

INTERPRETATION

CC-Cruiser exhibited less accurate performance comparing to senior consultants in diagnosing childhood cataracts and making treatment decisions. However, the medical service provided by CC-Cruiser was less time-consuming and achieved a high level of patient satisfaction. CC-Cruiser has the capacity to assist human doctors in clinical practice in its current state.

FUNDING

National Key R&D Program of China (2018YFC0116500) and the Key Research Plan for the National Natural Science Foundation of China in Cultivation Project (91846109).

Fundus photograph-based deep learning algorithms in detecting diabetic retinopathy

Remarkable advances in biomedical research have led to the generation of large amounts of data. Using artificial intelligence, it has become possible to extract meaningful information from large volumes of data, in a shorter frame of time, with very less human interference. In effect, convolutional neural networks (a deep learning method) have been taught to recognize pathological lesions from images. Diabetes has high morbidity, with millions of people who need to be screened for diabetic retinopathy (DR). Deep neural networks offer a great advantage of screening for DR from retinal images, in improved identification of DR lesions and risk factors for diseases, with high accuracy and reliability. This review aims to compare the current evidences on various deep learning models for diagnosis of diabetic retinopathy (DR).

Deep Learning-Based Algorithms in Screening of Diabetic Retinopathy: A Systematic Review of Diagnostic Performance

TOPIC

Diagnostic performance of deep learning-based algorithms in screening patients with diabetes for diabetic retinopathy (DR). The algorithms were compared with the current gold standard of classification by human specialists.

CLINICAL RELEVANCE

Because DR is a common cause of visual impairment, screening is indicated to avoid irreversible vision loss. Automated DR classification using deep learning may be a suitable new screening tool that could improve diagnostic performance and reduce manpower.

METHODS

For this systematic review, we aimed to identify studies that incorporated the use of deep learning in classifying full-scale DR in retinal fundus images of patients with diabetes. The studies had to provide a DR grading scale, a human grader as a reference standard, and a deep learning performance score. A systematic search on April 5, 2018, through MEDLINE and Embase yielded 304 publications. To identify potentially missed publications, the reference lists of the final included studies were manually screened, yielding no additional publications. The Quality Assessment of Diagnostic Accuracy Studies 2 tool was used for risk of bias and applicability assessment.

RESULTS

By using objective selection, we included 11 diagnostic accuracy studies that validated the performance of their deep learning method using a new group of patients or retrospective datasets. Eight studies reported sensitivity and specificity of 80.28% to 100.0% and 84.0% to 99.0%, respectively. Two studies report accuracies of 78.7% and 81.0%. One study provides an area under the receiver operating curve of 0.955. In addition to diagnostic performance, one study also reported on patient satisfaction, showing that 78% of patients preferred an automated deep learning model over manual human grading.

CONCLUSIONS

Advantages of implementing deep learning-based algorithms in DR screening include reduction in manpower, cost of screening, and issues relating to intragrader and intergrader variability. However, limitations that may hinder such an implementation particularly revolve around ethical concerns regarding lack of trust in the diagnostic accuracy of computers. Considering both strengths and limitations, as well as the high performance of deep learning-based algorithms, automated DR classification using deep learning could be feasible in a real-world screening scenario.

Can Artificial Intelligence Make Screening Faster, More Accurate, and More Accessible?

Diabetic retinopathy, glaucoma, and age-related macular degeneration are leading causes of vision loss and blindness worldwide. They tend to be asymptomatic in the early phase of disease and therefore require active screening programs to identify the patients requiring referral and treatment. Deep learning-based artificial intelligence technology has recently become a major topic in the field of ophthalmology. This paper aimed to provide a general view of the major findings on the application of deep learning for the classification of eye diseases from common imaging modalities. In the future, it is expected that these technologies will be applied in real-world screening programs to improve their efficiency and affordability.

Artificial intelligence in retina

Major advances in diagnostic technologies are offering unprecedented insight into the condition of the retina and beyond ocular disease. Digital images providing millions of morphological datasets can fast and non-invasively be analyzed in a comprehensive manner using artificial intelligence (AI). Methods based on machine learning (ML) and particularly deep learning (DL) are able to identify, localize and quantify pathological features in almost every macular and retinal disease. Convolutional neural networks thereby mimic the path of the human brain for object recognition through learning of pathological features from training sets, supervised ML, or even extrapolation from patterns recognized independently, unsupervised ML. The methods of AI-based retinal analyses are diverse and differ widely in their applicability, interpretability and reliability in different datasets and diseases. Fully automated AI-based systems have recently been approved for screening of diabetic retinopathy (DR). The overall potential of ML/DL includes screening, diagnostic grading as well as guidance of therapy with automated detection of disease activity, recurrences, quantification of therapeutic effects and identification of relevant targets for novel therapeutic approaches. Prediction and prognostic conclusions further expand the potential benefit of AI in retina which will enable personalized health care as well as large scale management and will empower the ophthalmologist to provide high quality diagnosis/therapy and successfully deal with the complexity of 21st century ophthalmology.

Assessing the value of screening tools: reviewing the challenges and opportunities of cost-effectiveness analysis

Background

Screening is an important part of preventive medicine. Ideally, screening tools identify patients early enough to provide treatment and avoid or reduce symptoms and other consequences, improving health outcomes of the population at a reasonable cost. Cost-effectiveness analyses combine the expected benefits and costs of interventions and can be used to assess the value of screening tools.

Objective

This review seeks to evaluate the latest cost-effectiveness analyses on screening tools to identify the current challenges encountered and potential methods to overcome them.

Methods

A systematic literature search of EMBASE and MEDLINE identified cost-effectiveness analyses of screening tools published in 2017. Data extracted included the population, disease, screening tools, comparators, perspective, time horizon, discounting, and outcomes. Challenges and methodological suggestions were narratively synthesized.

Results

Four key categories were identified: screening pathways, pre-symptomatic disease, treatment outcomes, and non-health benefits. Not all studies included treatment outcomes; 15 studies (22%) did not include treatment following diagnosis. Quality-adjusted life years were used by 35 (51.4%) as the main outcome. Studies that undertook a societal perspective did not report non-health benefits and costs consistently. Two important challenges identified were (i) estimating the sojourn time, i.e., the time between when a patient can be identified by screening tests and when they would have been identified due to symptoms, and (ii) estimating the treatment effect and progression rates of patients identified early.

Conclusions

To capture all important costs and outcomes of a screening tool, screening pathways should be modeled including patient treatment. Also, false positive and false negative patients are likely to have important costs and consequences and should be included in the analysis. As these patients are difficult to identify in regular data sources, common treatment patterns should be used to determine how these patients are likely to be treated. It is important that assumptions are clearly indicated and that the consequences of these assumptions are tested in sensitivity analyses, particularly the assumptions of independence of consecutive tests and the level of patient and provider compliance to guidelines and sojourn times. As data is rarely available regarding the progression of undiagnosed patients, extrapolation from diagnosed patients may be necessary.

An intelligible deep convolution neural network based approach for classification of diabetic retinopathy

Deep convolution neural networks (CNNs) have demonstrated their capabilities in modern-day medical image classification and analysis. The vital edge of deep CNN over other techniques is their ability to train without expert knowledge. Time bound detection is very beneficial for the early cure of disease. In this paper, a deep CNN architecture is proposed to classify nondiabetic retinopathy and diabetic retinopathy fundus eye images. Kaggle 2015 diabetic retinopathy competition dataset and messier experiment dataset are used in this study. The proposed deep CNN algorithm produces significant results with 93% area under the curve (AUC) for the Kaggle dataset and 91% AUC for the Messidor dataset. The sensitivity and specificity for the Kaggle dataset are 90.22% and 85.13%, respectively; the corresponding values of the Messidor dataset are 91.07% and 80.23%, respectively. The results outperformed many existing studies. The present architecture is a promising tool for diabetic retinopathy image classification.

Deep learning applications in ophthalmology

PURPOSE OF REVIEW

To describe the emerging applications of deep learning in ophthalmology.

RECENT FINDINGS

Recent studies have shown that various deep learning models are capable of detecting and diagnosing various diseases afflicting the posterior segment of the eye with high accuracy. Most of the initial studies have centered around detection of referable diabetic retinopathy, age-related macular degeneration, and glaucoma.

SUMMARY

Deep learning has shown promising results in automated image analysis of fundus photographs and optical coherence tomography images. Additional testing and research is required to clinically validate this technology.

Retinal Telemedicine

PURPOSE OF REVIEW

An update and overview of the literature on current telemedicine applications in retina.

RECENT FINDINGS

The application of telemedicine to the field of Ophthalmology and Retina has been growing with advancing technologies in ophthalmic imaging. Retinal telemedicine has been most commonly applied to diabetic retinopathy and retinopathy of prematurity in adult and pediatric patients respectively. Telemedicine has the potential to alleviate the growing demand for clinical evaluation of retinal diseases. Subsequently, automated image analysis and deep learning systems may facilitate efficient processing of large, increasing numbers of images generated in telemedicine systems. Telemedicine may additionally improve access to education and standardized training through tele-education systems.

SUMMARY

Telemedicine has the potential to be utilized as a useful adjunct but not a complete replacement for physical clinical examinations. Retinal telemedicine programs should be carefully and appropriately integrated into current clinical systems.

Automated detection of diabetic retinopathy lesions on ultrawidefield pseudocolour images

PURPOSE

We examined the sensitivity and specificity of an automated algorithm for detecting referral-warranted diabetic retinopathy (DR) on Optos ultrawidefield (UWF) pseudocolour images.

METHODS

Patients with diabetes were recruited for UWF imaging. A total of 383 subjects (754 eyes) were enrolled. Nonproliferative DR graded to be moderate or higher on the 5-level International Clinical Diabetic Retinopathy (ICDR) severity scale was considered as grounds for referral. The software automatically detected DR lesions using the previously trained classifiers and classified each image in the test set as referral-warranted or not warranted. Sensitivity, specificity and the area under the receiver operating curve (AUROC) of the algorithm were computed.

RESULTS

The automated algorithm achieved a 91.7%/90.3% sensitivity (95% CI 90.1-93.9/80.4-89.4) with a 50.0%/53.6% specificity (95% CI 31.7-72.8/36.5-71.4) for detecting referral-warranted retinopathy at the patient/eye levels, respectively; the AUROC was 0.873/0.851 (95% CI 0.819-0.922/0.804-0.894).

CONCLUSION

Diabetic retinopathy (DR) lesions were detected from Optos pseudocolour UWF images using an automated algorithm. Images were classified as referral-warranted DR with a high degree of sensitivity and moderate specificity. Automated analysis of UWF images could be of value in DR screening programmes and could allow for more complete and accurate disease staging.

SS, Dhawan B, Palsule A, Kumar D, Chandel S, Sood S. Sensitivity and specificity of automated analysis of single-field non-mydriatic fundus photographs by Bosch DR Algorithm-Comparison with mydriatic fundus photography (ETDRS) for screening in undiagnosed diabetic retinopathy

Diabetic retinopathy (DR) is a leading cause of blindness among working-age adults. Early diagnosis through effective screening programs is likely to improve vision outcomes. The ETDRS seven-standard-field 35-mm stereoscopic color retinal imaging (ETDRS) of the dilated eye is elaborate and requires mydriasis, and is unsuitable for screening. We evaluated an image analysis application for the automated diagnosis of DR from non-mydriatic single-field images. Patients suffering from diabetes for at least 5 years were included if they were 18 years or older. Patients already diagnosed with DR were excluded. Physiologic mydriasis was achieved by placing the subjects in a dark room. Images were captured using a Bosch Mobile Eye Care fundus camera. The images were analyzed by the Retinal Imaging Bosch DR Algorithm for the diagnosis of DR. All subjects also subsequently underwent pharmacological mydriasis and ETDRS imaging. Non-mydriatic and mydriatic images were read by ophthalmologists. The ETDRS readings were used as the gold standard for calculating the sensitivity and specificity for the software. 564 consecutive subjects (1128 eyes) were recruited from six centers in India. Each subject was evaluated at a single outpatient visit. Forty-four of 1128 images (3.9%) could not be read by the algorithm, and were categorized as inconclusive. In four subjects, neither eye provided an acceptable image: these four subjects were excluded from the analysis. This left 560 subjects for analysis (1084 eyes). The algorithm correctly diagnosed 531 of 560 cases. The sensitivity, specificity, and positive and negative predictive values were 91%, 97%, 94%, and 95% respectively. The Bosch DR Algorithm shows favorable sensitivity and specificity in diagnosing DR from non-mydriatic images, and can greatly simplify screening for DR. This also has major implications for telemedicine in the use of screening for retinopathy in patients with diabetes mellitus.

Development and Validation of a Deep Learning System for Diabetic Retinopathy and Related Eye Diseases Using Retinal Images From Multiethnic Populations With Diabetes

IMPORTANCE

A deep learning system (DLS) is a machine learning technology with potential for screening diabetic retinopathy and related eye diseases.

OBJECTIVE

To evaluate the performance of a DLS in detecting referable diabetic retinopathy, vision-threatening diabetic retinopathy, possible glaucoma, and age-related macular degeneration (AMD) in community and clinic-based multiethnic populations with diabetes.

DESIGN, SETTING, AND PARTICIPANTS

Diagnostic performance of a DLS for diabetic retinopathy and related eye diseases was evaluated using 494 661 retinal images. A DLS was trained for detecting diabetic retinopathy (using 76 370 images), possible glaucoma (125 189 images), and AMD (72 610 images), and performance of DLS was evaluated for detecting diabetic retinopathy (using 112 648 images), possible glaucoma (71 896 images), and AMD (35 948 images). Training of the DLS was completed in May 2016, and validation of the DLS was completed in May 2017 for detection of referable diabetic retinopathy (moderate nonproliferative diabetic retinopathy or worse) and vision-threatening diabetic retinopathy (severe nonproliferative diabetic retinopathy or worse) using a primary validation data set in the Singapore National Diabetic Retinopathy Screening Program and 10 multiethnic cohorts with diabetes.

EXPOSURES

Use of a deep learning system.

MAIN OUTCOMES AND MEASURES

Area under the receiver operating characteristic curve (AUC) and sensitivity and specificity of the DLS with professional graders (retinal specialists, general ophthalmologists, trained graders, or optometrists) as the reference standard.

RESULTS

In the primary validation dataset (n = 14 880 patients; 71 896 images; mean [SD] age, 60.2 [2.2] years; 54.6% men), the prevalence of referable diabetic retinopathy was 3.0%; vision-threatening diabetic retinopathy, 0.6%; possible glaucoma, 0.1%; and AMD, 2.5%. The AUC of the DLS for referable diabetic retinopathy was 0.936 (95% CI, 0.925-0.943), sensitivity was 90.5% (95% CI, 87.3%-93.0%), and specificity was 91.6% (95% CI, 91.0%-92.2%). For vision-threatening diabetic retinopathy, AUC was 0.958 (95% CI, 0.956-0.961), sensitivity was 100% (95% CI, 94.1%-100.0%), and specificity was 91.1% (95% CI, 90.7%-91.4%). For possible glaucoma, AUC was 0.942 (95% CI, 0.929-0.954), sensitivity was 96.4% (95% CI, 81.7%-99.9%), and specificity was 87.2% (95% CI, 86.8%-87.5%). For AMD, AUC was 0.931 (95% CI, 0.928-0.935), sensitivity was 93.2% (95% CI, 91.1%-99.8%), and specificity was 88.7% (95% CI, 88.3%-89.0%). For referable diabetic retinopathy in the 10 additional datasets, AUC range was 0.889 to 0.983 (n = 40 752 images).

CONCLUSIONS AND RELEVANCE

In this evaluation of retinal images from multiethnic cohorts of patients with diabetes, the DLS had high sensitivity and specificity for identifying diabetic retinopathy and related eye diseases. Further research is necessary to evaluate the applicability of the DLS in health care settings and the utility of the DLS to improve vision outcomes.

Combination of Gray-Level and Moment Invariant for Automatic Blood Vessel Detection on Retinal Image

Segmentation of blood vessels in the retinal is a crucial step in the diagnosis of eye diseases such as diabetic retinopathy and glaucoma. This paper presents a supervised method for automatic segmentation of blood vessels in retinal images. The proposed method based on a hybrid combination between Gray-Level and Moment Invariant techniques. There are four steps involved, whereas preprocessing, feature extraction, classification, and post-processing. In the preprocessing, three stages are performed include vessel central light reflex removal, background homogenization, and vessel enhancement. The 7-D vector feature extraction was performed to compute that compose of gray-level and moment invariants-based features for pixel representation. The decision tree is used for classification step that characterized the pixel based on vessels and non-vessels. The final step is the post-processing which will remove the small artifacts appears after classification process. The proposed method was compared to the Vascular Tree method and Morphological method. Based on the objective evaluation, the proposed method achieved (sensitivity = 98.589, specificity = 55.544 and accuracy = 96.197).

Crowdsourcing and Automated Retinal Image Analysis for Diabetic Retinopathy

PURPOSE OF REVIEW

As the number of people with diabetic retinopathy (DR) in the USA is expected to increase threefold by 2050, the need to reduce health care costs associated with screening for this treatable disease is ever present. Crowdsourcing and automated retinal image analysis (ARIA) are two areas where new technology has been applied to reduce costs in screening for DR. This paper reviews the current literature surrounding these new technologies.

RECENT FINDINGS

Crowdsourcing has high sensitivity for normal vs abnormal images; however, when multiple categories for severity of DR are added, specificity is reduced. ARIAs have higher sensitivity and specificity, and some commercial ARIA programs are already in use. Deep learning enhanced ARIAs appear to offer even more improvement in ARIA grading accuracy. The utilization of crowdsourcing and ARIAs may be a key to reducing the time and cost burden of processing images from DR screening.