Dataset Information

Systematic review and meta-analysis of diagnostic accuracy of detection of any level of diabetic retinopathy using digital retinal imaging.

ABSTRACT:

Background

Visual impairment from diabetic retinopathy (DR) is an increasing global public health concern, which is preventable with screening and early treatment. Digital retinal imaging has become a preferred choice as it enables higher coverage of screening. The aim of this review is to evaluate how different characteristics of the DR screening (DRS) test impact on diagnostic test accuracy (DTA) and its relevance to a low-income setting.

Methods

We conducted a systematic literature search to identify clinic-based studies on DRS using digital retinal imaging of people with DM (PwDM). Summary estimates of different sub-groups were calculated using DTA values weighted according to the sample size. The DTA of each screening method was derived after exclusion of ungradable images and considering the eye as the unit of analysis. The meta-analysis included studies which measured DTA of detecting any level of DR. We also examined the effect on detection from using different combinations of retinal fields, pupil status, index test graders and setting.

Results

Six thousand six hundred forty-six titles and abstracts were retrieved, and data were extracted from 122 potentially eligible full reports. Twenty-six studies were included in the review, and 21 studies, mostly from high-income settings (18/21, 85.7%), were included in the meta-analysis. The highest sensitivity was observed in the mydriatic greater than two field strategy (92%, 95% CI 90-94%). The highest specificity was observed in greater than two field methods (94%, 95% CI 93-96%) where mydriasis did not affect specificity. Overall, there was no difference in sensitivity between non-mydriatic and mydriatic methods (86%, 95% CI 85-87) after exclusion of ungradable images. The highest DTA (sensitivity 90%, 95% CI 88-91%; specificity 95%, 95% CI 94-96%) was observed when screening was delivered at secondary/tertiary level clinics.

Conclusions

Non-mydriatic two-field strategy could be a more pragmatic approach in starting DRS programmes for facility-based PwDM in low-income settings, with dilatation of the pupils of those who have ungradable images. There was insufficient evidence in primary studies to draw firm conclusions on how graders' background influences DTA. Conducting more context-specific DRS validation studies in low-income and non-ophthalmic settings can be recommended.

SUBMITTER: Piyasena MMPN

PROVIDER: S-EPMC6222985 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Diabetic Retinopathy (DR) may result in various degrees of vision loss and even blindness if not diagnosed in a timely manner. Therefore, having an annual eye exam helps early detection to prevent vision loss in earlier stages, especially for diabetic patients. Recent technological advances made smartphone-based retinal imaging systems available on the market to perform small-sized, low-powered, and affordable DR screening in diverse environments. However, the accuracy of DR detection depends on the field of view and image quality. Since smartphone-based retinal imaging systems have much more compact designs than a traditional fundus camera, captured images are likely to be the low quality with a smaller field of view. Our motivation in this paper is to develop an automatic DR detection model for smartphone-based retinal images using the deep learning approach with the ResNet50 network. This study first utilized the well-known AlexNet, GoogLeNet, and ResNet50 architectures, using the transfer learning approach. Second, these frameworks were retrained with retina images from several datasets including EyePACS, Messidor, IDRiD, and Messidor-2 to investigate the effect of using images from the single, cross, and multiple datasets. Third, the proposed ResNet50 model is applied to smartphone-based synthetic images to explore the DR detection accuracy of smartphone-based retinal imaging systems. Based on the vision-threatening diabetic retinopathy detection results, the proposed approach achieved a high classification accuracy of 98.6%, with a 98.2% sensitivity and a 99.1% specificity while its AUC was 0.9978 on the independent test dataset. As the main contributions, DR detection accuracy was improved using the deep transfer learning approach for the ResNet50 network with publicly available datasets and the effect of the field of view in smartphone-based retinal imaging was studied. Although a smaller number of images were used in the training set compared with the existing studies, considerably acceptable high accuracies for validation and testing data were obtained.

Project description:BackgroundThe evidence on diagnostic test accuracy (DTA) of diabetic retinopathy (DR) screening utilising photographic studies by non-ophthalmologist personnel in low and middle-income country (LMIC) settings is scarce. We aimed to assess DTA of DR screening using a nonmydriatic hand-held digital camera by trained general physicians in a non-ophthalmic setting.MethodsThis study is a validation of a screening intervention. We selected 700 people with diabetes (PwDM) > 18 years of age, not previously screened or treated for DR, presenting at a tertiary medical clinic in Sri Lanka. Two-field retinal imaging was used to capture fundus images before and after pupil dilatation, using a hand-held non-mydriatic (Visuscout 100®-Germany) digital retinal camera. The images were captured and graded by two trained, masked independent physician graders. The DTA of different levels of DR was assessed comparing physician's grading with a retinologist's clinical examination by mydriatic bio-microscopy, according to a locally adopted guideline.ResultsSeven hundred eligible PwDM were screened by physician graders. The mean age of participants was 60.8 years (SD ±10.08) and mean duration of DM was 9.9 years (SD ±8.09). Ungradable image proportion in non-mydriatic imaging was 43.4% (either eye-31.3%, both eyes 12.1%). This decreased to 12.8% (either eye-11.6%, both eyes-1.2%) following pupil dilatation. In comparison to detection of any level of DR, a referable level DR (moderate non-proliferative DR and levels above) showed a higher level of DTA. The sensitivity of the defined referable DR was 88.7% (95% CI 81.7-93.8%) for grader 1 (positive predictive value [PPV] 59.1%) and 92.5% (95% CI 86.4-96.5%) for grader 2 (PPV 68%), using mydriatic imaging, after including ungradable images as screen positives. The specificity was 94.9% (95% CI 93.6-96.0%) for grader 1 (negative predictive value [NPV] 99%) and 96.4% (95% CI 95.3-97.3%) for grader 2 (NPV 99.4%).ConclusionsThe Physicians grading of images from a digital hand-held non-mydriatic camera at a medical clinic, with dilatation of pupil of those who have ungradable images, provides a valid modality to identify referable level of DR. This could be a feasible alternative modality to the existing opportunistic screening to improve the access and coverage.Trial registrationCurrent Controlled Trials ISRCTN47559703 . Date of Registration 18th March 2019, Retrospectively registered.

Project description:BackgroundDiabetic retinopathy (DR), a common complication of diabetes mellitus, is the leading cause of impaired vision in adults worldwide. Smartphone ophthalmoscopy involves using a smartphone camera for digital retinal imaging. Utilizing smartphones to detect DR is potentially more affordable, accessible, and easier to use than conventional methods.ObjectiveThis study aimed to determine the diagnostic accuracy of various smartphone ophthalmoscopy approaches for detecting DR in diabetic patients.MethodsWe performed an electronic search on the Medical Literature Analysis and Retrieval System Online (MEDLINE), EMBASE, and Cochrane Library for literature published from January 2000 to November 2018. We included studies involving diabetic patients, which compared the diagnostic accuracy of smartphone ophthalmoscopy for detecting DR to an accurate or commonly employed reference standard, such as indirect ophthalmoscopy, slit-lamp biomicroscopy, and tabletop fundus photography. Two reviewers independently screened studies against the inclusion criteria, extracted data, and assessed the quality of included studies using the Quality Assessment of Diagnostic Accuracy Studies-2 tool, with disagreements resolved via consensus. Sensitivity and specificity were pooled using the random effects model. A summary receiver operating characteristic (SROC) curve was constructed. This review is reported in line with the Preferred Reporting Items for a Systematic Review and Meta-analysis of Diagnostic Test Accuracy Studies guidelines.ResultsIn all, nine studies involving 1430 participants were included. Most studies were of high quality, except one study with limited applicability because of its reference standard. The pooled sensitivity and specificity for detecting any DR was 87% (95% CI 74%-94%) and 94% (95% CI 81%-98%); mild nonproliferative DR (NPDR) was 39% (95% CI 10%-79%) and 95% (95% CI 91%-98%); moderate NPDR was 71% (95% CI 57%-81%) and 95% (95% CI 88%-98%); severe NPDR was 80% (95% CI 49%-94%) and 97% (95% CI 88%-99%); proliferative DR (PDR) was 92% (95% CI 79%-97%) and 99% (95% CI 96%-99%); diabetic macular edema was 79% (95% CI 63%-89%) and 93% (95% CI 82%-97%); and referral-warranted DR was 91% (95% CI 86%-94%) and 89% (95% CI 56%-98%). The area under SROC curve ranged from 0.879 to 0.979. The diagnostic odds ratio ranged from 11.3 to 1225.ConclusionsWe found heterogeneous evidence showing that smartphone ophthalmoscopy performs well in detecting DR. The diagnostic accuracy for PDR was highest. Future studies should standardize reference criteria and classification criteria and evaluate other available forms of smartphone ophthalmoscopy in primary care settings.

Project description:ObjectivesTo compare the accuracy of trained level 1 diabetic retinopathy (DR) graders (nurses, endocrinologists and one general practitioner), level 2 graders (midlevel ophthalmologists) and level 3 graders (senior ophthalmologists) in Vietnam against a reference standard from the UK and assess the impact of supplementary targeted grader training.DesignDiagnostic test accuracy study.SettingSecondary care hospitals in Southern Vietnam.ParticipantsDR training was delivered to Vietnamese graders in February 2018 by National Health Service (NHS) UK graders. Two-field retinal images (412 patient images) were graded by 14 trained graders in Vietnam between August and October 2018 and then regraded retrospectively by an NHS-certified reference standard UK optometrist (phase I). Further DR training based on phase I results was delivered to graders in November 2019. After training, a randomised subset of images from January to October 2020 (115 patient images) was graded by six of the original cohort (phase II). The reference grader regraded all images from phase I and II retrospectively in masked fashion.Primary and secondary outcome measuresSensitivity was calculated at the two different time points, and χ2 was used to test significance.ResultsIn phase I, the sensitivity for detecting any DR for all grader groups in Vietnam was low (41.8-42.2%) and improved in phase II after additional training was delivered (51.3-87.2%). The greatest improvement was seen among level 1 graders (p<0.001), and the lowest improvement was observed among level 3 graders (p=0.326). There was a statistically significant improvement in sensitivity for detecting referable DR and referable diabetic macular oedema between all grader levels. The post-training values ranged from 40.0 to 61.5% (including ungradable images) and 55.6%-90.0% (excluding ungradable images).ConclusionsThis study demonstrates that targeted training interventions can improve accuracy of DR grading. These findings have important implications for improving service delivery in DR screening programmes in low-resource settings.

Project description:Diabetic retinopathy (DR) is a late microvascular complication of Diabetes Mellitus (DM) that could lead to permanent blindness in patients, without early detection. Although adequate management of DM via regular eye examination can preserve vision in in 98% of the DR cases, DR screening and diagnoses based on clinical lesion features devised by expert clinicians; are costly, time-consuming and not sufficiently accurate. This raises the requirements for Artificial Intelligent (AI) systems which can accurately detect DR automatically and thus preventing DR before affecting vision. Hence, such systems can help clinician experts in certain cases and aid ophthalmologists in rapid diagnoses. To address such requirements, several approaches have been proposed in the literature that use Machine Learning (ML) and Deep Learning (DL) techniques to develop such systems. However, these approaches ignore the highly valuable clinical lesion features that could contribute significantly to the accurate detection of DR. Therefore, in this study we introduce a framework called DR-detector that employs the Extreme Gradient Boosting (XGBoost) ML model trained via the combination of the features extracted by the pretrained convolutional neural networks commonly known as transfer learning (TL) models and the clinical retinal lesion features for accurate detection of DR. The retinal lesion features are extracted via image segmentation technique using the UNET DL model and captures exudates (EXs), microaneurysms (MAs), and hemorrhages (HEMs) that are relevant lesions for DR detection. The feature combination approach implemented in DR-detector has been applied to two common TL models in the literature namely VGG-16 and ResNet-50. We trained the DR-detector model using a training dataset comprising of 1,840 color fundus images collected from e-ophtha, retinal lesions and APTOS 2019 Kaggle datasets of which 920 images are healthy. To validate the DR-detector model, we test the model on external dataset that consists of 81 healthy images collected from High-Resolution Fundus (HRF) dataset and MESSIDOR-2 datasets and 81 images with DR signs collected from Indian Diabetic Retinopathy Image Dataset (IDRID) dataset annotated for DR by expert. The experimental results show that the DR-detector model achieves a testing accuracy of 100% in detecting DR after training it with the combination of ResNet-50 and lesion features and 99.38% accuracy after training it with the combination of VGG-16 and lesion features. More importantly, the results also show a higher contribution of specific lesion features toward the performance of the DR-detector model. For instance, using only the hemorrhages feature to train the model, our model achieves an accuracy of 99.38 in detecting DR, which is higher than the accuracy when training the model with the combination of all lesion features (89%) and equal to the accuracy when training the model with the combination of all lesions and VGG-16 features together. This highlights the possibility of using only the clinical features, such as lesions that are clinically interpretable, to build the next generation of robust artificial intelligence (AI) systems with great clinical interpretability for DR detection. The code of the DR-detector framework is available on GitHub at https://github.com/Janga-Lab/DR-detector and can be readily employed for detecting DR from retinal image datasets.

Project description:ObjectiveDiabetic retinopathy (DR) is a common diabetic microvascular complication and a major cause of acquired vision loss. Finding effective biomarkers for the early identification and diagnosis of DR is crucial. This study aimed to comprehensively evaluate the accuracy of microRNAs (miRNAs) in the diagnosis of DR via a meta-analysis of previously published diagnostic studies. This study has been registered on the PROSPERO website, with the number CRD42022323238.MethodsWe searched PubMed, Cochrane Library, Embase, Web of Science, China Wanfang database, and China Knowledge Network database to identify relevant articles published from the time of database creation to April 10, 2022. Stata 14.0 software was used to calculate the pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic ratio (DOR), and area under the summary receiver operating characteristic (ROC) curve to assess the accuracy of miRNAs in the diagnosis of DR. Heterogeneity between studies was assessed using Cochran-Q test and I2 statistic for quantitative analysis. The random-effect model was selected due to significant heterogeneity. Subgroup analysis and regression analysis were also performed to determine the potential sources of heterogeneity.ResultsWe included 25 articles detailing 52 studies with 1987 patients with DR and 1771 non-DR controls. The findings demonstrated overall sensitivity (0.82, 95% CI: 0.78 ~ 0.85), specificity (0.84, 95% CI: 0.81 ~ 0.86), PLR (5.0, 95% CI: 4.2 ~ 5.9), NLR (0.22, 95% CI: 0.18 ~ 0.26), and the area under the summary ROC curve (0.90, 95% CI: 0.87 ~ 0.92). Furthermore, we performed subgroup analysis and found that panels of multiple miRNAs could enhance the pooled sensitivity (sensitivity, specificity, and AUC values were 0.89, 0.87, and 0.94, respectively).ConclusionThe meta-analysis showed that miRNAs can be used as potential diagnostic markers for DR, with high accuracy of diagnoses observed with the detection of miRNAs in plasma and serum.