Table 1.
AI-based structure-function in best-corrected visual acuity (BCVA).
| Author/Ref. | Title | Disease | Technique | Prediction | Outcome measure | Outcome |
|---|---|---|---|---|---|---|
| Rohm et al. [27] | Predicting visual acuity by using machine learning in patients treated for neovascular age-related macular degeneration | Neovascular age-related macular degeneration |
• Five different machine-learning algorithms • Best performance by Lasso regression |
• logMAR visual acuity after 3- and 12 months |
• Mean Absolute Error (MAE) • Root Mean Sqaured Error (RMSE) |
• 3 Months = MAE: 0.11–014/RMSE: 0.18–0.2 • 12 Months = MAE: 0.16–0.2/RMSE: 0.2–0.22 |
| Schmidt-Erfurth et al. [28] | Machine learning to analyze the prognostic value of current imaging biomarkers in neovascular age-related macular degeneration | Neovascular age-related macular degeneration | • Random forest | • BCVA at Baseline and 3 months follow-up | • Accuracy (R2) |
• R2 = 0.21 baseline • R2 = 0.70 3 months |
| Gerenda et al. [29] | Computational image analysis for prognosis determination in DME | Diabetic macular edema | • Random forest | • BCVA at Baseline and 1-year follow-up | • Accuracy (R2) |
• R2 = 0.21 baseline • R2 = 0.23 1 year |
| Aslam et al. [30] | Use of a neural net to model the impact of optical coherence tomography abnormalities on vision in age-related macular degeneration | Neovascular age-related macular degeneration | • Scaled conjugate gradient backpropagation (supervised learning) | • BCVA | • Root Mean Sqaured Error (RMSE) | • 8.21 Letters |
| Pfau et al. [31] | Artificial intelligence in ophthalmology: guideline for physicians for the critical evaluation of studies | Neovascular age-related macular degeneration | • Nested cross validation | • BCVA (LogMAR) | • MAE | • 0.142 |
| Müller et al. [23] | Prediction of function in ABCA4-related retinopathy using ensemble machine learning | ABCA4-related Retinopathy |
• Ensemble machine learning algorithms • Three models (a) Retinal layer (b) All structural data (c) demographic data |
• BCVA • Divided into four categories from no to severe impairment |
• Area under the curve (ROC) |
(a) 88.64–92.25% (b) 90.23–93.68% (c) 87.26–91.44% |
| Sumaroka et al. [45] | Foveal therapy in blue cone monochromacy: predictions of visual potential from artificial intelligence | Blue Cone Monochromacy |
• Random forest (a) Layer thickness (b) Reflectivity |
• BCVA | • Root mean squared error (RMSE) |
(a) 0.159 (b) 0.167 |