Skip to main content
. 2023 Mar 28;11:1168327. doi: 10.3389/fcell.2023.1168327

TABLE 5.

Research summary of artificial intelligence in age-related macular degeneration.

Year Country or region Authors Task Dataset (disease images) AI algorithm Output
2022 Korea Han et al. (2022) Diagnosis 4,749 images (2,624 images) VGG-16, VGG-19, ResNet Accuracy = 0.874
2021 America Tak et al. (2021) Classification 420 images (420 images) Convolutional neural networks Accuracy = 0.88
2021 Taiwan Chou et al. (2021) Diagnosis 699 images (491 images) EfficientNet-B3 Accuracy = 0.8367, Sensitivity = 0.8076, Specificity = 0.8472, AUC = 0.8857
2020 Korea Heo et al. (2020) Diagnosis 399 images (399 images) VGG16 Accuracy = 0.9086
2022 America Ganjdanesh et al. (2022) Prediction 30,000 images (30,000 images) ResNet-18 Accuracy = 0.905, AUC = 0.762
2022 China Song et al. (2022) Prediction 671 images (671 images) Classified convolution neural network, complete convolution neural network Accuracy = 0.930, Dice coefficients = 0.873, Sensitivity = 0.873, Specificity = 0.922
2022 Taiwan Yeh et al. (2022) Prediction 698 images (698 images) Deep convolution neural network AUC = 0.989, Accuracy = 0.936, Sensitivity = 0.933, Specificity = 0.938
2020 America Yan et al. (2020) Prediction 31,262 images, 52 related mutated genes (31,262 images) Convolutional neural networks AUC = 0.85
2022 Austria Holomcik et al. (2022) Division 9,268 images (9,268 images) U-Net F1 score = 0.65, Accuracy = 0.75, Recall = 0.72
2022 China He et al. (2022) Detection UCSD dataset, Duke dataset (46,421 images) ResNet-50, Local outlier factor UCSD: Accuracy = 0.9987
Duke: Accuracy = 0.9756