Skip to main content

Table 5 Summary of DL methods using FP and OCT to detect eye disease

From: Application of machine learning in ophthalmic imaging modalities

AuthorsYearImaging ModalitiesAimData setsDL techniquesPerformance
Arcadu F et al. [111]2019FPDiabetic macular thickening detectionLocal:17,997 FPsInception-v3AUC:0.97 (central subfield thickness ≥ 250 μm)0.91 (central foveal thickness ≥ 250 μm)0.94 (central subfield thickness ≥ 400 μm)0.96 (central foveal thickness ≥ 400 μm)
Nagasawa T et al. [112]2019FPTreatment-naïve proliferative diabetic retinopathy detectionLocal:132 FPsVGG-16Sensitivity: 94.7%Specificity: 97.2%AUC: 0.969
Phan S et al. [113]2019FPGlaucoma detectionLocal:3312 FPsVGG-19ResNet-152DenseNet-201AUCs of 0.9 or more (3 DCNNs)
Nagasato D et al. [114]2019FPBranch retinal vein occlusion detectionLocal:466 FPsVGG-16SVMSensitivity: 94.0%Specificity: 97.0%positive predictive value (PPV): 96.5%negative predictive value (NPV): 93.2%AUC: 97.6%
Burlina PM et al. [115]2019FPTo develop DL techniques for synthesizing high-resolution realistic fundus imagesLocal:133,821 FPsGANAUC:0.9706 (model trained on real data) 0.9235 (model trained on synthetic data)
Girard F et al. [116]2019FPJoint segmentation and classification of retinal arteries and veinsPublic:DRIVE, 40 FPsMESSIDOR, 1200 FPsCNNAccuracy: 94.8% Sensitivity: 93.7% Specificity: 92.9%
Coyner AS et al. [117]2018FPImage quality assessment of fundus images in ROPLocal: 6043 FPsVGG-19 DCNNAccuracy: 89.1% AUC: 0.964
Keel S et al. [118]2018FPDetection of referable diabetic retinopathy and glaucomaPublic:LabelMe, 114,906 FPs (referable DR) Sensitivity:90% (glaucomatous optic neuropathy) 96% (referable DR)
Sayres R et al. [119]2018FPAssist grading for DRPublic: EyePACS, 1796 FPsInception v-4Sensitivity:79.4% (unassisted) 87.5% (grades only) 88.7% (grades plus heatmap)
Peng Y et al. [120]2018FPAutomated classification of AMD severityPublic: AREDS, 59302 FPsDeepSeeNet (Inception v-3)Accuracy: 0.671 AUC: 0.94 (large drusen) 0.93 (pigmentary abnormalities) 0.97 (late AMD)
Guo Y et al. [121]2018FPRetinal vessel detectionPublic: DRIVE, 20 FPs STARE, 20 FPsMultiple DCNNsAccuracy: 95.97% (DRIVE training dataset) 96.13% (DRIVE testing dataset) 95.39% (STARE dataset) AUC: 0,9726 (DRIVE training dataset) 0.9737 (DRIVE testing dataset) 0.9539 (STARE dataset)
Khojasteh P et al. [122]2018FPDetection of exudates, microaneurysms and hemorrhagesPublic: DIARETDB1, 75 FPs e-Ophtha, 209 FPsCNNAccuracy: 97.3% (DIARETDB1 dataset) 86.6% (e-Ophtha) Sensitivity: 0.96 (exudates) 0.84 (hemorrhages) 0.85 (microaneurysms)
Gargeya R et al. [123]2017FPAutomated identification of DRPublic: EyePACS, 75,137 FPs MESSIDOR 2, 1748 E-Ophtha, 463 FPsDCNNSensitivity: 94% Specificity: 98% AUC: 0.97
Burlina PM et al. [63]2017FPAutomated grading of AMDPublic: AREDS, more than 130,000 FPsDCNNAccuracy: 88.4% (SD, 0.5%)-91.6% (SD, 0.1%) AUC: 0.94 (SD, 0.5%)-0.96 (SD, 0.1%)
Ordóñez PF et al. [124]2017FPTo improve the accuracy of microaneurysms detectionPublic: Kaggle, 88,702 FPs Messidor, 1200 FPs DiaRerDB1, 89 FPsStandard CNNVGG CNNSensitivity > 91% Specificity > 93% AUC > 93%
Takahashi H et al. [58]2017FPImproving staging of DRLocal: 9939 FPsGoogleNet DCNNPrevalence and bias-adjusted Fleiss’kappa (PABAK): 0.64 (modified Davis grading) 0.37 (real prognosis grading)
Abbas Q et al. [125]2017FPAutomatic recognition of severity level of DRLocal: 750 FPsDCNNSensitivity: 92.18% Specificity: 94.50% AUC: 0.924
Pfister M et al. [126]2019OCTAutomated segmentation of dermal fillers in OCT imagesLocal: 100 OCT volume data setsCNN (U-net-like architecture)Accuracy: 0.9938
Fu H et al. [127]2019OCTAutomated angle-closure detectionLocal: 4135 anterior segment OCT imagesCNNSensitivity: 0.79 ± 0.037 Specificity: 0.87 ± 0.009 AUC: 0.90
Masood S et al. [128]2019OCTAutomatic choroid layer segmentation from OCT imagesLocal: 525 OCT imagesCNN (Cifar-10 model)Accuracy: 97%
Dos Santos VA et al. [129]2019OCTSegmentation of cornea OCT scansLocal: 20,160 OCT imagesCNNAccuracy: 99.56%
Asaoka R et al. [130]2019OCTDiagnosis early-onset glaucoma from OCT imagesLocal: 4316 OCT imagesCNNAUC: 93.7%
Lu W et al. [131]2018OCTClassification of multi-categorical abnormalities from OCT imagesLocal: 60,407 OCT imagesResNetAccuracy: 0.959 AUC: 0.984
Schlegl T et al. [132]2018OCTDetection of macular fluid in OCT imagesLocal: 1200 OCT scansCNNIntraretinal cystoid fluid detection: Accuracy: 0.91 AUC: 0.94 Subretinal fluid detection: Accuracy: 0.61 AUC: 0.92
Prahs P et al. [133]2018OCTEvaluation of treatment indication with anti-vascular endothelial growth factor medicationsLocal: 183,402 OCT scansGoogleNet inception DCNNAccuracy: 95.5% Sensitivity: 90.1% Specificity: 96.2% AUC: 0.968
Shah A et al. [134]2018OCTRetinal layer segmentation in OCT imagesLocal: 3000 OCT scansCNNAverage computation time: 12.3 s
Chan GCY et al. [135]2018OCTAutomated diabetic macular edema classificationPublic: Singapore Eye Research Institute, 14,720 OCT scansAlexNet, VGG, GoogleNetAccuracy: 93.75%
Muhammad H et al. [136]2017OCTClassification of glaucoma suspectsLocal:102 OCT scansCNN, Random forestAccuracy: 93.1% (retinal nerve fiber layer)
Lee CS et al. [81]2017OCTSegmentation of macular edema in OCTLocal:1289 OCT imagesU-Net CNNcross-validated Dice coefficient: 0.911
Lee CS et al. [137]2017OCTClassification of normal and AMD OCT imagesPublic:Electronic medical records, 101,002 OCT imagesVGG-16Accuracy: 87.63% AUC: 92.78%
  1. DL = deep learning; FP = fundus photography; OCT = optical coherence tomography; CNN = convolution neural network; DCNN = deep convolution neural network; DR = diabetic retinopathy; AMD = age-related macular degeneration; AUC = area under the curve
\