Table 2

Summary table for the different DL systems in the detection of retinal diseases using OCT

DL systemsYearDiseaseOCT machinesTest imagesCNNAUCAccuracy (%)Sensitivity (%)Specificity (%)
Lee et al 13 322017Exudative AMDSpectralis20 613VGG-160.92887.6084.6091.50
Trader et al 332018Exudative AMDSpectralis100Inception-V30.980100NANA
Kermany et al 34 2018CNVSpectralis1000Inception-V3
1. Multiclass comparison0.99996.5097.8097.40
2. Limited model0.98893.4096.6094.00
3. Binary model
CNV vs normal1100100100
DMO vs normal0.99998.2096.8099.60
Drusen vs normal0.999999899.20
De Fauw et al 43 2018Urgent, semiurgent, routine and observation onlyTopcon997 patients1. Deep segmentation network using U-NetUrgent
Normal, CNV, macular oedema, FTMH, PTMH, CSR, VMT, GA, drusen, ERMSpectralis116 patients2. Deep classification network using a custom 29 CNN layers with 5 pooling layersUrgent referral
  • The diagnostic performance is not comparable between the different DL systems given the different data sets used in the individual study. AUC for specific conditions: CNV 0.993; macular oedema 0.990; normal 0.995; FTMH 1.00; PTMH 0.999; CSR 0.995; VMT 0.980; GA 0.990; drusen 0.967; and ERM 0.966.

  • AMD,age-related macular degeneration; AUC,area under the receiver operating characteristic curve; CNN,convolutional neural network; CNV,choroidal neovascularisation; CSR,central serous chorioretinopathy; DL,deep learning; DMO,diabetic macular oedema; ERM,epiretinal membrane; FTMH,full-thickness macula hole; GA,geographic atrophy; NA, not available; OCT,optical coherence tomography; PTMH,partial thickness macula hole; VMT, vitreomacular traction.