Predicting optical coherence tomography-derived diabetic macular edema grades from fundus photographs using deep learning.
Ontology highlight
ABSTRACT:
SUBMITTER: Varadarajan AV
PROVIDER: S-EPMC6949287 | biostudies-literature | 2020 Jan
REPOSITORIES: biostudies-literature
Publications
Predicting optical coherence tomography-derived diabetic macular edema grades from fundus photographs using deep learning.
Varadarajan Avinash V AV Bavishi Pinal P Ruamviboonsuk Paisan P Chotcomwongse Peranut P Venugopalan Subhashini S Narayanaswamy Arunachalam A Cuadros Jorge J Kanai Kuniyoshi K Bresnick George G Tadarati Mongkol M Silpa-Archa Sukhum S Limwattanayingyong Jirawut J Nganthavee Variya V Ledsam Joseph R JR Keane Pearse A PA Corrado Greg S GS Peng Lily L Webster Dale R DR
Nature communications 20200108 1
Center-involved diabetic macular edema (ci-DME) is a major cause of vision loss. Although the gold standard for diagnosis involves 3D imaging, 2D imaging by fundus photography is usually used in screening settings, resulting in high false-positive and false-negative calls. To address this, we train a deep learning model to predict ci-DME from fundus photographs, with an ROC-AUC of 0.89 (95% CI: 0.87-0.91), corresponding to 85% sensitivity at 80% specificity. In comparison, retinal specialists ha ...[more]
Similar Datasets
Project description:Diabetic macular edema (DME) is the abnormal accumulation of fluid in the subretinal or intraretinal spaces in the macula in patients with diabetic retinopathy and leads to severely impaired central vision. Technical developments in retinal imaging systems have led to many advances in the study of DME. In particular, optical coherence tomography (OCT) can provide longitudinal and microstructural analysis of the macula. A comprehensive review was provided regarding the role of inflammation using OCT-based classification of DME and current and ongoing therapeutic approaches. In this review, we first describe the pathogenesis of DME, then discuss the classification of DME based on OCT findings and the association of different types of DME with inflammation, and finally describe current and ongoing therapeutic approaches using OCT-based classification of DME. Inflammation has an important role in the pathogenesis of DME, but its role appears to differ among the DME phenotypes, as determined by OCT. It is important to determine how the different DME subtypes respond to intravitreal injections of steroids, antivascular endothelial growth factor agents, and other drugs to improve prognosis and responsiveness to treatment.
| S-EPMC6939426 | biostudies-literature
Project description:This retrospective study was performed to classify diabetic macular edema (DME) based on the localization and area of the fluid and to investigate the relationship of the classification with visual acuity (VA). The fluid was visualized using en face optical coherence tomography (OCT) images constructed using swept-source OCT. A total of 128 eyes with DME were included. The retina was segmented into: Segment 1, mainly comprising the inner nuclear layer and outer plexiform layer, including Henle's fiber layer; and Segment 2, mainly comprising the outer nuclear layer. DME was classified as: foveal cystoid space at Segment 1 and no fluid at Segment 2 (n = 24), parafoveal cystoid space at Segment 1 and no fluid at Segment 2 (n = 25), parafoveal cystoid space at Segment 1 and diffuse fluid at Segment 2 (n = 16), diffuse fluid at both segments (n = 37), and diffuse fluid at both segments with subretinal fluid (n = 26). Eyes with diffuse fluid at Segment 2 showed significantly poorer VA, higher ellipsoid zone disruption rates, and greater central subfield thickness than did those without fluid at Segment 2 (P < 0.001 for all). These results indicate the importance of the localization and area of the fluid for VA in DME.
| S-EPMC8026626 | biostudies-literature
Project description:Several studies published so far used highly selective image datasets from unclear sources to train computer vision models and that may lead to overestimated results, while those studies conducted in real-life remain scarce. To avoid image selection bias, we stacked convolutional and recurrent neural networks (CNN-RNN) to analyze complete optical coherence tomography (OCT) cubes in a row and predict diabetic macular edema (DME), in a real-world diabetic retinopathy screening program. A retrospective cohort study was carried out. Throughout 4-years, 5314 OCT cubes from 4408 subjects who attended to the diabetic retinopathy (DR) screening program were included. We arranged twenty-two (22) pre-trained CNNs in parallel with a bidirectional RNN layer stacked at the bottom, allowing the model to make a prediction for the whole OCT cube. The staff of retina experts built a ground truth of DME later used to train a set of these CNN-RNN models with different configurations. For each trained CNN-RNN model, we performed threshold tuning to find the optimal cut-off point for binary classification of DME. Finally, the best models were selected according to sensitivity, specificity, and area under the receiver operating characteristics curve (AUROC) with their 95% confidence intervals (95%CI). An ensemble of the best models was also explored. 5188 cubes were non-DME and 126 were DME. Three models achieved an AUROC of 0.94. Among these, sensitivity, and specificity (95%CI) ranged from 84.1-90.5 and 89.7-93.3, respectively, at threshold 1, from 89.7-92.1 and 80-83.1 at threshold 2, and from 80.2-81 and 93.8-97, at threshold 3. The ensemble model improved these results, and lower specificity was observed among subjects with sight-threatening DR. Analysis by age, gender, or grade of DME did not vary the performance of the models. CNN-RNN models showed high diagnostic accuracy for detecting DME in a real-world setting. This engine allowed us to detect extra-foveal DMEs commonly overlooked in other studies, and showed potential for application as the first filter of non-referable patients in an outpatient center within a population-based DR screening program, otherwise ended up in specialized care.
| S-EPMC11291805 | biostudies-literature
Project description:BackgroundDiabetic macular edema (DME) is a major cause of vision loss in diabetics that is currently mainly treated by antivascular endothelial growth factor (VEGF) agents. The effect of these agents on macular perfusion (MP) is a current concern. Optical coherence tomography angiography (OCTA) is an imaging modality that allows noninvasive high-resolution retinal microvasculature imaging. Several recent studies evaluated the effect of anti-VEGF agents on the MP of DME patients using OCTA. Our aim is to provide a systematic review of these studies.MethodsMultiple databases were searched including PubMed, Ovid Medline, EMBASE, and Google Scholar for relevant studies published between January 2016 and November 2020 which were included in this review. Studies were compared regarding their design, the number of included patients, the machine and scanning protocol used, the inclusion and exclusion criteria, the number of injections given, the type of anti-VEGF agent used, the outcome measures assessed, and the effect of injections on different MP parameters.ResultsA total of 16 studies were included. The studies assessed various OCTA parameters that define MP including the foveal avascular zone area and superficial and deep vascular density and yielded conflicting results. Seven studies showed stable or improved MP following treatment, while 7 studies showed worsening MP following treatment, and 2 studies showed inconclusive results. This could have been due to differences in study design, inclusion criteria, type of anti-VEGF agents used, treatment duration, and methods of image analysis and vascular density quantification. All identified studies were noncomparative case series, and 14 of them (87.5%) used the RTVue XR Avanti OCTA machine. Only one study compared OCTA to fluorescein angiography findings.ConclusionAnalysis of MP changes following VEGF inhibition for DME could benefit from a unified scanning protocol and analysis approach that uses similar study designs to eliminate potential sources of bias. This may provide more definitive conclusions regarding the effect of treatment on MP.
| S-EPMC8169275 | biostudies-literature
Project description:PurposeTo assess the relationship between best-corrected visual acuity (BCVA) and optical coherence tomography (OCT) features in noninfectious uveitis (NIU)-related macular edema.DesignClinical cohort study from post hoc analysis of 2 phase 3 clinical trials.MethodsCorrelation and longitudinal treatment analyses were performed. Of 198 patients with NIU, 134 received suprachoroidal CLS-TA (proprietary formulation of a triamcinolone acetonide injectable suspension), and 64 received sham, with 12.9% and 72%, respectively, receiving rescue therapy.ResultsAt baseline, mean BCVA progressively worsened with each ordinal drop in central subfield ellipsoid zone (EZ) integrity. Eyes with normal baseline EZ experienced greater 24-week change in BCVA versus those with some degree of baseline EZ disruption (11.9 vs 9.4 letters, P = .006). In contrast, eyes with baseline central subfield cystoid spaces and/or subretinal fluid showed more improvement (13.7 or 17.2 letters, respectively) at 24 weeks, versus those without such findings (5.5 [P = .012] or 9.5 letters [P < .001], respectively). Longitudinal modeling for CLS-TA-treated eyes showed that central subfield thickness (CST) reached 90% of maximal improvement by week 3, whereas 90% maximal response in BCVA was not reached until week 9. CLS-TA-treated eyes that showed CST reduction of ≥50 µm at 4 weeks experienced a greater 24-week improvement in BCVA versus those without such an early response (14.6 vs 6.5 letters, P = .006 for difference).ConclusionsPretreatment EZ integrity and the presence of central subfield cystoid spaces or subretinal fluid each predict improved therapeutic response to treatment in eyes with NIU. In CLS-TA treated eyes, longitudinal modeling shows CST improvement preceding BCVA improvement.
| S-EPMC10919549 | biostudies-literature
Project description:ObjectiveTo evaluate agreement between fluorescein angiography (FA) and optical coherence tomography (OCT) results for diagnosis of macular edema in patients with uveitis.DesignMulticenter cross-sectional study.ParticipantsFour hundred seventy-nine eyes with uveitis from 255 patients.MethodsThe macular status of dilated eyes with intermediate uveitis, posterior uveitis, or panuveitis was assessed via Stratus-3 OCT and FA. To evaluate agreement between the diagnostic approaches, ? statistics were used.Main outcome measuresMacular thickening (MT; center point thickness, ? 240 ?m per reading center grading of OCT images) and macular leakage (ML; central subfield fluorescein leakage, ? 0.44 disc areas per reading center grading of FA images), and agreement between these outcomes in diagnosing macular edema.ResultsOptical coherence tomography (90.4%) more frequently returned usable information regarding macular edema than FA (77%) or biomicroscopy (76%). Agreement in diagnosis of MT and ML (? = 0.44) was moderate. Macular leakage was present in 40% of cases free of MT, whereas MT was present in 34% of cases without ML. Biomicroscopic evaluation for macular edema failed to detect 40% and 45% of cases of MT and ML, respectively, and diagnosed 17% and 17% of cases with macular edema that did not have MT or ML, respectively; these results may underestimate biomicroscopic errors (ophthalmologists were not explicitly masked to OCT and FA results). Among eyes free of ML, phakic eyes without cataract rarely (4%) had MT. No factors were found that effectively ruled out ML when MT was absent.ConclusionsOptical coherence tomography and FA offered only moderate agreement regarding macular edema status in uveitis cases, probably because what they measure (MT and ML) are related but nonidentical macular pathologic characteristics. Given its lower cost, greater safety, and greater likelihood of obtaining usable information, OCT may be the best initial test for evaluation of suspected macular edema. However, given that ML cannot be ruled out if MT is absent and vice versa, obtaining the second test after negative results on the first seems justified when detection of ML or MT would alter management. Given that biomicroscopic evaluation for macular edema erred frequently, ancillary testing for macular edema seems indicated when knowledge of ML or MT status would affect management.Financial disclosure(s)Proprietary or commercial disclosure may be found after the references.
| S-EPMC3758459 | biostudies-literature
Project description:PurposeDevelop equations to convert Cirrus central subfield thickness (CST) to Spectralis CST equivalents and vice versa in eyes with diabetic macular edema (DME).MethodsThe DRCR Retina Network Protocol O data were split randomly to train (70% sample) and validate (30% sample) conversion equations. Data from an independent study (CADME) also validated the equations. Bland-Altman 95% limits of agreement between predicted and observed values evaluated the equations.ResultsProtocol O included 374 CST scan pairs from 187 eyes (107 participants). The CADME study included 150 scan pairs of 37 eyes (37 participants). Proposed conversion equations are Spectralis = 40.78 + 0.95 × Cirrus and Cirrus = 1.82 + 0.94 × Spectralis regardless of age, sex, or CST. Predicted values were within 10% of observed values in 101 (90%) of Spectralis and 99 (88%) of Cirrus scans in the validation data; and in 136 (91%) of the Spectralis and 148 (99%) of the Cirrus scans in the CADME data. Adjusting for within-eye correlations, 95% of conversions are estimated to be within 17% (95% confidence interval, 14%-21%) of CST on Spectralis and within 22% (95% confidence interval, 18%-28%) of CST on Cirrus.ConclusionsConversion equations developed in this study allow the harmonization of CST measurements for eyes with DME using a mix of current Cirrus and Spectralis device images.Translational relevanceThe CSTs measured on Cirrus and Spectralis devices are not directly comparable owing to outer boundary segmentation differences. Converting CST values across spectral domain optical coherence tomography instruments should benefit both clinical research and standard care efforts.
| S-EPMC8727314 | biostudies-literature
Project description:ImportanceCoats disease is a rare pediatric vitreoretinopathy that can cause devastating visual and anatomic outcomes.ObjectiveTo compare optical coherence tomography (OCT) with fundus photographs, fluorescein angiography (FA), and histopathologic findings in Coats disease.Design, setting, and participantsThis retrospective cohort study was conducted in a single tertiary institution (Duke Eye Center) and identified 28 children with Coats disease through a review of medical records from December 2002 to January 2018. Four eyes were obtained from a biorepository for histopathologic analysis.Main outcomes and measuresMacular OCT, fundus photographs, and FA results were reviewed and compared for morphological changes. These were compared with retinal histopathological findings.ResultsThe mean (SD) age was 9.5 (5.5) years for the 28 children (and 29 eyes) with clinical imaging results, and 24 (86%) were boys. A comparison between imaging modalities revealed OCT features that were not visible in photographs or FA, including exudates in multiple retinal layers (23 [82.1%]), small pockets of subretinal fluid (4 [14.3%]), an outer retinal atrophy overlying fibrotic nodules (7 [25.0%]), and small preretinal hyperreflective OCT dots (25 [89.3%]). Next, a comparison with light micrographs introduced an association of OCT findings with possible pathological features, including hyperreflective linear structures on OCT that appeared consistent with cholesterol crystals, small hyperreflective dots with macrophages, outer retinal tubulations with rosettes, and analogous OCT histopathology features such as intraretinal vessels entering fibrotic nodules and retinal pigment epithelium excrescences under the subretinal fluid. An OCT analysis revealed intraretinal cystoid spaces in 19 eyes, but in 9 of 19 (47.4) this was not associated with cystoid macular leakage; rather, fluorescein leakage was observed from peripheral telangiectatic vessels. Additionally, exudates were intraretinal only (6 [21.4%]) or both intraretinal and subretinal (17 [60.7%]); none were subretinal only. In eyes with follow-up results, new fibrosis developed in 8 of 17 eyes (47.1%). Fibrosis developed in 5 of 5 eyes (100%) with baseline subretinal fluid vs 3 of 12 without (25%; 95% CI, 22%-92%) and in 7 of 9 eyes (77.8%) with subretinal exudates vs 1 of 8 (12.5%) without (95% CI, 16%-89%).Conclusions and relevanceOptical coherence tomography may show the transient and permanent effects of Coats disease on the retina. These results suggest that exudates and fluid in the macular subretinal space appear later in the disease and may result in fibrosis formation. Further studies are needed to confirm if early treatment could prevent vision-threatening macular fibrosis.
| S-EPMC6439851 | biostudies-literature
Project description:PurposeTo develop an automated method based on deep learning (DL) to classify macular edema (ME) from the evaluation of optical coherence tomography (OCT) scans.MethodsA total of 4230 images were obtained from data repositories of patients attended in an ophthalmology clinic in Colombia and two free open-access databases. They were annotated with four biomarkers (BMs) as intraretinal fluid, subretinal fluid, hyperreflective foci/tissue, and drusen. Then the scans were labeled as control or ocular disease among diabetic macular edema (DME), neovascular age-related macular degeneration (nAMD), and retinal vein occlusion (RVO) by two expert ophthalmologists. Our method was developed by following four consecutive phases: segmentation of BMs, the combination of BMs, feature extraction with convolutional neural networks to achieve binary classification for each disease, and, finally, multiclass classification of diseases and control images.ResultsThe accuracy of our model for nAMD was 97%, and for DME, RVO, and control were 94%, 93%, and 93%, respectively. Area under curve values were 0.99, 0.98, 0.96, and 0.97, respectively. The mean Cohen's kappa coefficient for the multiclass classification task was 0.84.ConclusionsThe proposed DL model may identify OCT scans as normal and ME. In addition, it may classify its cause among three major exudative retinal diseases with high accuracy and reliability.Translational relevanceOur DL approach can optimize the efficiency and timeliness of appropriate etiological diagnosis of ME, thus improving patient access and clinical decision making. It could be useful in places with a shortage of specialists and for readers that evaluate OCT scans remotely.
| S-EPMC9526369 | biostudies-literature
Project description:PurposeTo evaluate the repeatability and reproducibility of retinal and choroidal thickness measured with Swept source Optical Coherence Tomography (SS-OCT) in eyes with Diabetic Macular Edema (DME).Methods42 DME eyes were imaged using SS-OCT standard Macular scanning protocols. Retinal and choroidal thickness were measured in the Total macular circle (TMC) and foveal central subfield (FCS) using device-integrated specific software. The coefficient of repeatability (CR) and intraclass correlation coefficient (ICC) were determined as a measure of repeatability and relative reliability within graders. Reproducibility was assessed using Bland-Altman plots and 95% limits of agreement (LoA) were determined as a measure of interobserver variability.ResultsIntragrader CR of retinal and choroidal thickness were 8.37 and 12.20 microns for TMC and 22.24 and 32.40 microns for FCS, and intergrader 95% LoA were 7.37-8.69 and -27.2-27.71 microns for TMC and -34.21-41.93 and -30.46-24.84 for FCS, respectively. Retinal and choroidal thickness showed very good intraobserver reliability for both TMC and FCS (ICC 0.99, LoA 0.98-0.99 in all cases). Intraobserver and interobserver variability for retinal and choroidal thickness was not significantly different for TMC (p = 0.98 and p = 0.90, p = 0.98 and p = 0.91) or FCS (p = 0.97 and p = 0.85, p = 0.78 and p = 0.73), respectively.ConclusionsRetinal and choroidal thickness in DME eyes can be quantified with good reliability, repeatability and reproducibility using new OCT devices that incorporate swept source technology. The technical advantages of this technology may provide new insights in the understanding of the choroidal changes related with DME.
| S-EPMC6062033 | biostudies-literature