Table 2.
Characteristics of ophthalmic imaging studies.
| Study | Model | Prospective? | Test set | Population | Test datasets | Type of internal validation | External validation | Reference standard | AI vs clinician? | Imaging modality | Pathology |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Abramoff et al. 2016 | AlexNet/VGG | No | 1748 | Photographs | Messidor-2 | NR | No | Expert consensus | No | Retinal fundus photography | Referable DR |
| Abramoff et al. 201814 | AlexNet/VGG | Yes | 819 | Patients | Prospective cohort from 10 primary care practice sites in USA | NR | Yes | Expert consensus | No | Retinal fundus photography | More than mild DR |
| Ahn et al. 2018 | (a) Inception-v3; (b) customised CNN | No | (a) 464; (b) 464 | Images | Kim’s Eye Hospital, Korea | Random split | No | Expert consensus | No | Retinal fundus photography | Early and advanced glaucoma |
| Ahn et al. 2019 | ResNet50 | No | 219 | Photographs | Kim’s Eye Hospital, Korea | Random split | No | Expert consensus | No | Retinal fundus photographs | Pseudopapilloedema |
| Al-Aswad et al. 201946 | Pegasus (ResNet50) | No | 110 | Photographs | Singapore Malay Eye Study | Random split | No | Existing diagnosis from source data | Yes | Retinal fundus photographs | Glaucoma |
| Alqudah et al. 201922 | AOCT-NET | No | 1250 | Scans | Farsiu Ophthalmology 2013 | Hold-out method | Yes | NR | No | OCT | (a) AMD; (b) DME |
| Arcadu et al. 2019 | Inception-v3 | No | (a) 1237; (b) 1798 | Images | RISE/RIDE trials | Random split | No | Expert consensus | No | Retinal fundus photography | (a) DME—central subfield thickness >400 µm; (b) DME—central fovea thickness >400 µm |
| Asaoka et al. 2016 | Deep feed-forward neural network with stacked denoising autoencoder | No | 279 | Eyes | University of Tokyo Hospital, Tokyo | Random split | No | Other imaging technique | No | Visual Fields | Preperimetric open-angle glaucoma |
| Asaoka et al. 2019 | Customised CNN | No | 196 | Images | University of Tokyo Hospital, Tokyo | Random split | No | Expert consensus | No | OCT | Early open-angle glaucoma |
| Asaoka et al. 201923 | ResNet50 | No | (a) 205; (b) 171 | Scans | (a) Iinan Hospital; (b) Hiroshiuma University | NR | Yes | Expert consensus | No | OCT | Glaucoma |
| Bellemo et al. 201915 | VGG/ResNet | Yes | 3093 | Eyes | Kitwe Central Hospital Eye Unit, Zambia | NA | Yes | Expert consensus | No | Retinal fundus photography | (a) Referable DR; (b) vision-threatening DR; (c) DME |
| Bhatia et al. 201924 | VGG-16 | No | (a) 4686; (b) 384; (c) 148; (d) 100; (e) 135; (f) 135; (g) 148; (h) 100 | Scans | (a) Shiley Eye Institute of the UCDS; (b) Devers Eye Institute; (c) Noor Eye Hospital; (d) Ophthalmica Ophthalmology Greece; (e) Cardiff University; (f) Cardiff University; (g) Noor Eye Hospital; (h) Ophthalmica Ophthalmology Greece | NA | Yes | (a) Expert consensus; (b) NR; (c) NR; (d) NR; (e) Expert consensus + further imaging; (f) expert consensus + further imaging; (g) NR; (h) NR | No | OCT | (a) Abnormal scan; (b–f) AMD; (g–h) DME |
| Brown et al. 201847 | Inception-v1 and U-Net | No | 100 | Photographs | i-ROP | Hold-out method | No | Expert consensus | Yes | Retinal fundus photography | Plus disease in ROP |
| Burlina et al. 201749 | DCNN | No | 5664 | Images | AREDS 4 dataset | NR | No | Expert consensus | Yes | Retinal fundus photography | AMD-AREDS 4 step |
| Burlina et al. 201848 | ResNet50 | No | 5000 | Images | AREDS | Random split | No | Reading centre grader | No | Retinal fundus photographs | Referable AMD |
| Burlina et al. 201850 | AlexNet | No | 13,480 | Photographs | NIH AREDS | NR | No | Reading centre grader | Yes | Retinal fundus photography | Referable AMD |
| Burlina et al. 2018 | ResNet50 | No | (a) 6654; (b) 58,978 | Images | (a) AREDS 9 dataset; (b) AREDS 4 dataset | NR | No | Reading centre grader | Yes | Retinal fundus photography | (a) AMD-AREDS 4 step; (b) AMD-AREDS 9 step |
| Chan et al. 201825 | AlexNet, VGGNet, GoogleNet | No | 4096 | Images | SERI | NR | Yes | Reading centre grader | No | OCT | DME |
| Choi et al. 2017 | VGG-19 | No | (a) 3000; (b) 3000 | Photographs | STARE database | Random split | No | Expert consensus | No | Retinal fundus photographs | (a) DR; (b) AMD |
| Christopher et al. 201816 | (a) VGG-16; (b) Inception-v3; (c) ResNet50 | Yes | 1482 | Images | ADAGES and DIGS | Random split | No | Expert consensus | No | Retinal fundus photography | Glaucomatous optic neuropathy |
| Das et al. 2019 | VGG-16 | No | 1000 | Images | UCSD | Hold-out method | No | Expert consensus | No | OCT | DME |
| De Fauw et al. 201851 | (a) U-Net (b) customised CNN | No | (a) 997; (b) 116 | (a) Scans (Topcon device); (b) scans (Spectralis device) | Moorfields, London | Random split | No | Follow up | Yes | OCT | Urgent referral eye disease |
| ElTanboly et al. 2016 | Deep fusion classification network (DFCN) | No | 12 | OCT scans | Hold-out method | No | NR | No | OCT | Early DR | |
| Gargeya et al. 201726 | CNN | No | (a) 15,000 (b) 1748; (c) 463 | Photographs | (a) EyePACS-1; (b) Messidor-2; (c) E-Opthma | Random split | Yes | Expert consensus | No | Retinal fundus photography | DR |
| Gomez-Valverde et al. 201952 | VGG-19 | No | 494 | Photographs | ESPERANZA | Random split | No | Expert consensus | Yes | Retinal fundus photographs | Glaucoma suspect or glaucoma |
| Grassman et al. 201827 | Ensemble:random forest | No | (a) 12,019; (b) 5555 | Images | (a) AREDS dataset; (b) KORA dataset | Random split | Yes | Reading centre grader | No | Retinal fundus photography | AMD-AREDS 9 step |
| Gulshan et al. 201917 | Inception-v3 | Yes | 3049 | Photographs | Prospective | NA | Yes | Expert consensus | Yes | Retinal fundus photographs | Referable DR |
| Gulshan et al. 201628 | Inception-v3 | No | (a) 8788; (b) 1745 | Photographs | (a) EyePACS-1; (b) Messidor-2 | Random split | Yes | Reading centre grader | Yes | Retinal fundus photography | Referable DR |
| Hwang et al. 201929 | (a) ResNet50; (b) VGG-16; (c) Inception-v3; (d) ResNet50; (e) VGG-16; (f) Inception-v3 | No | (a–c) 3872; (d–f) 750 | Images | (a–c) Department of Ophthalmology of Taipei Veterans General Hospital; (d–f) External validation | Random split | Yes | Expert consensus | Yes | OCT | AMD-AREDS 4 step |
| Jammal et al. 201953 | ResNet34 | No | 490 | Images | Randomly drawn from test sample | No | Reading centre grader | Yes | Retinal fundus photographs | Glaucomatous optic neuropathy | |
| Kanagasingham et al. 201821 | DCNN | Yes | 398 | Patients | Primary Care Practice, Midland, Western Australia | NA | Yes | Reading centre grader | No | Retinal fundus photography | Referable DR |
| Karri et al. 2017 | GoogLeNet | No | 21 | Scans | Duke University | Random split | No | NR | No | OCT | (a) DME; (b) dry AMD |
| Keel et al. 201818 | Inception-v3 | Yes | 93 | Images | St Vincent’s Hospital Melbourne and University Hospital Geelong, Barwon Health | NA | Yes | Reading centre grader | No | Retinal fundus photography | Referable DR |
| Keel et al. 201930 | CNN | No | 86,202 | Photographs | Melbourne Collaborative Cohort Study | Hold-out method | Yes | Expert consensus | No | Retinal fundus photographs | Neovascular AMD |
| Kermany et al. 201854 | Inception-v3 | No | (a) 1000; (b–d) 500 | Scans | Shiley Eye Institute of the University of California San Diego, the California Retinal Research Foundation, Medical Centre Ophthalmology Associates, the Shanghai First People’s Hospital, and Beijing Tongren Eye Centre | Random split | No | Consensus involving experts and non-experts | Yes | OCT | (a) Choroidal neovascularisation vs DME vs drusen vs normal; (b) choroidal neovascularisation; (c) DME; (d) AMD |
| Krause et al. 201831 | CNN | No | 1958 | Images | EyePACS-2 | Hold-out method | Yes | Expert consensus | No | Retinal fundus photographs | Referable DR |
| Lee et al. 2017 | VGG-16 | No | 2151 | Scans | Random split | No | Routine clinical notes | No | OCT | AMD | |
| Lee et al. 2019 | CNN | No | 200 | Photographs | Seoul National University Hospital | Hold-out method | No | Other imaging technique | No | Retinal fundus photographs | Glaucoma |
| Li et al. 2018108 | Inception-v3 | No | 8000 | Scans | Guangdong (China) | Random split | No | Expert graders | No | Retinal fundus photography | Glaucomatous optic neuropathy |
| Li et al. 201955 | VGG-16 | No | 1000 | Images | Shiley Eye Institute of the University of California San Diego, the California Retinal Research Foundation, Medical Centre Ophthalmology Associates, the Shanghai First People’s Hospital, and Beijing Tongren Eye Centre | Random split | No | Expert consensus | No | OCT | Choroidal neovascularisation vs DME vs drusen Vs normal |
| Li et al. 2019 | OCT-NET | No | 859 | Scans | Wenzhou Medical University | Random split | No | Expert graders | No | OCT | Early DR |
| Li et al. 201933 | Inception-v3 | No | 800 | Images | Messidor-2 | Random split | Yes | Reading centre grader | No | Retinal fundus photographs | Referable DR |
| Li et al. 2019 | ResNet50 | No | 1635 | Images | Shanghai Zhongshan Hospital and the Shanghai First People’s Hospital | Random split | No | Reading centre grader | Yes | OCT | DME |
| Lin et al. 2019109 | CC-Cruiser | Yes—multicentre RCT | 350 | Images | Multicentre RCT | NA | NA | Expert consensus | Yes | Slit-lamp photography | Childhood cataracts |
| Li F et al. 2018 | VGG-15 | No | 300 | Images | NR | Random split | No | NR | No | Visual Fields | Glaucoma |
| Li Z et al. 201833 | CNN | No | 35,201 | Photographs | NIEHS, SiMES, AusDiab | Random split | Yes | Reading centre grader | No | Retinal fundus photographs | Referable DR |
| Liu et al. 201835 | ResNet50 | No | (a) 754; (b) 30 | Photographs | (a) NR; (b) HRF | Random split | Yes | Reading centre grader | Yes | Retinal fundus photographs | Glaucomatous optic discs |
| Liu et al. 201934 | CNN | No | (a) 28,569; (b) 20,466; (c) 12,718; (d) 9305; (e) 29,676; (f) 7877 | Photographs | (a) Local Validation (Chinese Glaucoma Study Alliance); (b) Beijing Tongren Hospital; (c) Peking University Third Hospital; (d) Harbin Medical University First Hospital; (e) Handan Eye Study; (f) Hamilton Glaucoma Centre | Random split | Yes | Consensus involving experts and non-experts | No | Retinal fundus photographs | Glaucomatous optic neuropathy |
| Long et al. 201756 | DCNN | No | 57 | Images | Multihospital clinical trial | Hold-out method | No | Expert consensus | Yes | Ocular images | Congenital Cataracts |
| MacCormick et al. 201936 | DenseNet | No | (a) 130; (b) 159 | Images | (a) ORIGA; (b) RIM-ONE | Random split | Yes | (a) NR; (b) expert consensus | No | Retinal fundus photography | Glaucomatous optic discs |
| Maetshke et al. 2019 | 3D CNN | No | 110 | OCT scans | Fivefold cross validation | Random split | No | Follow up | No | OCT | Glaucomatous optic neuropathy |
| Matsuba et al. 201857 | DCNN | No | 111 | Images | Tsukazaki Hospital | NR | No | Expert consensus + further imaging | Yes | Retinal fundus photography (optos) | Exudative AMD |
| Medeiros et al. 2019 | ResNet34 | No | 6292 | Images | Duke University | Random split | No | Follow up | No | Retinal fundus photography | Glaucomatous optic neuropathy |
| Motozawa et al. 2019 | CNN | No | 382 | Images | Kobe City Medical Centre | Random split | No | Routine clinical notes | No | OCT | AMD |
| Muhammad et al. 2017 | AlexNet | No | 102 | Images | NR | NR | No | Expert consensus | No | OCT | Glaucoma suspect or glaucoma |
| Nagasato et al. 2019 | VGG-16 | No | 466 | Images | NR | K-fold cross validation | No | NR | No | Retinal fundus photography (optos) | Retinal vein occlusion |
| Nagasato et al. 201958 | DNN | No | 322 | Scans | Tsukazaki Hospital and Tokushima University Hospital | K-fold cross validation | No | Expert graders | Yes | OCT | Retinal vein occlusion |
| Nagasawa et al. 2019 | VGG-16 | No | 378 | Images | Tsukazaki Hospital and Tokushima University Hospital | K-fold cross validation | No | Expert graders | No | Retinal fundus photography (optos) | Proliferative diabetic retinopathy |
| Ohsugi et al. 2017 | DCNN | No | 166 | Images | Tsukazaki Hospital | Random split | No | Expert consensus | No | Retinal fundus photography (optos) | Rhegmatogenous retinal detachment |
| Peng et al. 201959 | Inception-v3 | No | 900 | Images | AREDS | Random split | No | Reading centre grader | Yes | Retinal fundus photography | Age-related macular degeneration-AREDS 4 step |
| Perdomo et al. 2019 | OCT-NET | No | 2816 | Images | SERI-CUHK data set | Random split | No | Expert graders | No | OCT | DME |
| Phan et al. 2019 | DenseNet201 | No | 828 | Images | Yamanashi Koseiren Hospital | No | Expert consensus + further imaging | No | Retinal fundus photography | Glaucoma | |
| Phene et al. 201937 | Inception-v3 | No | (a) 1205; (b) 9642; (c) 346 | Images | (a) EyePACS, Inoveon, the United Kingdom Biobank, the Age-Related Eye Disease Study, and Sankara Nethralaya; (b) Atlanta Veterans Affairs (VA) Eye Clinic; (c) Dr. Shroff’s Charity Eye Hospital, New Delhi, India | Random split | Yes | Reading centre grader | Yes | Retinal fundus photographs | Glaucomatous optic neuropathy |
| Prahs et al. 2017 | GoogLeNet | No | 5358 | Images | Heidelberg Eye Explorer, Heidelberg Engineering | Random split | No | Expert graders | No | OCT | Injection vs No injection for AMD |
| Raju et al. 2017 | CNN | No | 53,126 | Images | EyePACS-1 | Random split | No | NR | No | Retinal fundus photography | Referable DR |
| Ramachandran et al. 201838 | Visiona intelligent diabetic retinopathy screening platform | No | (a) 485; (b) 1200 | Photographs | (a) ODEMS; (b) Messidor | NA | Yes | Expert graders | No | Retinal fundus photographs | Referable DR |
| Raumviboonsuk et al. 201939 | Inception-v4 | No | (a–c) 25,348; (d) 24,332 | Images | National screening program for DR in Thailand | NA | Yes | Expert consensus | Yes | Retinal fundus photography | (a) Moderate non-proliferative DR or worse; (b) severe non-proliferative DR or worse; (c) proliferative DR; (d) referable DME |
| Redd et al. 2018 | Inception-v1 and U-Net | No | 4861 | Images | Multicentre i-ROP study | NR | No | Expert graders + further imaging | No | Retinal fundus photography | Plus disease in ROP |
| Rogers et al. 201945 | Pegasus (ResNet50) | No | 94 | Photographs | EODAT | NA | Yes | Reading centre grader | Yes | Retinal fundus photographs | Glaucomatous optic neuropathy |
| Sandhu et al. 201819 | Deep fusion SNCAE | Yes | 160 | Scans | University of Waikato | NA | No | Clinical diagnosis | No | Retinal fundus photographs | Non-proliferative DR |
| Sayres et al. 201940 | Inception-v4 | No | 2000 | Images | EyePACS-2 | NA | Yes | Expert consensus | Yes | Retinal fundus photographs | Referable DR |
| Shibata et al. 201860 | (a) ResNet; (b) VGG-16 | No | 110 | Images | Matsue Red Cross Hospital | Random split | No | Expert consensus | Yes | Retinal fundus photography | Glaucoma |
| Stevenson et al. 2019 | Inception-v3 | No | (a) 2333; (b) 2283; (c) 2105 | Photographs | Publicly available databases | Random split | No | Existing diagnosis from source data | No | Retinal fundus photographs | (a) Glaucoma; (b) DR; (c) AMD |
| Ting et al. 201741 | VGGNet | No | (a) 71,896; (b) 15,798; (c) 3052; (d) 4512; (e) 1936; (f) 1052; (g) 1968; (h) 2302; (i) 1172; (j) 1254; (k) 7706; (l) 35,948; (m) 35,948 | Images | (a) Singapore National Diabetic Retinopathy Screening Program 2014–2015; (b) Guangdong (China); (c) Singapore Malay Eye Study; (d) Singapore Indian Eye Study; (e) Singapore Chinese Eye Study; (f) Beijing Eye Study; (g) African American Eye Disease Study; (h) Royal Victoria Eye and Ear Hospital; (i) Mexican; (j) Chinese University of Hong Kong, (k, l) Singapore National Diabetic Retinopathy Screening Program 2014–2015 | Random split | Yes | Expert consensus | No | Retinal fundus photography | Referable DR |
| Ting et al. 201942 | VGGNet | No | 85,902 | Images | Combined eight datasets | NA | Yes | Consensus involving experts and non-experts | No | Retinal fundus photography | (a) Any DR; (b) referable DR; (c) vision-threatening DR |
| Treder et al. 2017 | Inception-v3 | No | 100 | Scans | NR | Hold-out method | No | NR | No | OCT | Exudative AMD |
| van Grinsven et al. 201644 | (a) Ses CNN 60; (b) NSesCNN170 | No | 1200 | Images | Messidor | Random split | Yes | Existing diagnosis from source data | Yes | Retinal fundus photographs | Retinal haemorrhage |
| Verbraak et al. 201943 | AlexNet/VGG | No | 1293 | Images | Netherlands Star-SHL | NA | Yes | Expert consensus | No | Retinal fundus photography | (a) DR-vision-threatening; (b) DR- more than mild |
| Xu et al. 2017 | CNN | No | 200 | Photographs | Kaggle | Random split | No | Existing diagnosis from source data | No | Retinal fundus photographs | DR |
| Yang et al. 2019 | VGGNet | No | 500 | Photographs | Intelligent Ophthalmology Database of Zhejiang Society for Mathematical Medicine in China | Hold-out method | No | Expert consensus | No | Retinal fundus photographs | Referable DR |
| Yoo et al. 2019 | VGG-19 | No | 900 | Scans | Project Macula | Random split | No | NR | No | (a) OCT; (b) retinal fundus photographs | AMD |
| Zhang et al. 201961 | VGG-16 | No | 1742 | Images | Telemed-R screening | Random split | No | Expert consensus | Yes | Retinal fundus photographs | ROP |
| Zheng et al. 201920 | Inception-v3 | Yes | 102 | Scans | Joint Shantou International Eye Centre of Shantou University and the Chinese University of Hong Kong (JSIEC) | Hold-out method | No | NR | No | OCT | Glaucomatous optic neuropathy |