Longtime Vision Function Prediction in Childhood Cataract Patients Based on Optical Coherence Tomography Images

Abstract

The results of visual prediction reflect the tendency and speed of visual development during a future period, based on which ophthalmologists and guardians can know the potential visual prognosis in advance, decide on an intervention plan, and contribute to visual development. In our study, we developed an intelligent system based on the features of optical coherence tomography images for long-term prediction of best corrected visual acuity (BCVA) 3 and 5 years in advance. Two hundred eyes of 132 patients were included. Six machine learning algorithms were applied. In the BCVA predictions, small errors within two lines of the visual chart were achieved. The mean absolute errors (MAEs) between the prediction results and ground truth were 0.1482–0.2117 logMAR for 3-year predictions and 0.1198–0.1845 logMAR for 5-year predictions; the root mean square errors (RMSEs) were 0.1916–0.2942 logMAR for 3-year predictions and 0.1692–0.2537 logMAR for 5-year predictions. This is the first study to predict post-therapeutic BCVAs in young children. This work establishes a reliable method to predict prognosis 5 years in advance. The application of our research contributes to the design of visual intervention plans and visual prognosis.

Machine Learning Models for Predicting Long-Term Visual Acuity in Highly Myopic Eyes

“…Our results were similar to those of Wei et al (MAE = 0.152, RMSE = 0.261), who aimed to predict the postoperative VA of highly myopic eyes after cataract surgery. Our model is comparable to another long-term prediction reported by Xiang et al (MAE = 0.148, RMSE = 0.192 for 3 years; MAE = 0.120, RMSE = 0.169 for 5 years), which predicted the posttherapeutic BCVA of childhood cataracts. Regarding the prediction of the risk of developing VI, our results (AUROC = 0.87) were comparable with those of Kawczynski et al (AUROC = 0.84), which predicted BCVA at 12 months of fewer than 59 letters based on the baseline OCT images in patients with neovascular age-related macular degeneration.…”

“…Some AI systems predicting VA in other eye diseases have been published. 16,17,24,25 Our models presented a better R 2 value compared with that of Liu et al 24 (R 2 = 0.6), who used an ensemble ML system to predict the outcomes in patients with diabetic macular edema after anti-VEGF injections given every 3 months. Our results were similar to those of Wei et al 17 (MAE = 0.152, RMSE = 0.261), who aimed to predict the Regarding the prediction of the risk of developing VI, our results (AUROC = 0.87) were comparable with those of Kawczynski et al 16 (AUROC = 0.84), which predicted BCVA at 12 months of fewer than 59 letters based on the baseline OCT images in patients with neovascular age-related macular degeneration.…”

“…Some AI systems predicting VA in other eye diseases have been published. 16,17,24,25 Our models presented a better R 2 value compared with that of Liu et al 24 (R 2 = 0.6), who used an ensemble ML system to predict the outcomes in patients with diabetic macular edema after anti-VEGF injections given every 3 months. Our results were similar to those of Wei et al 17 (MAE = 0.152, RMSE = 0.261), who aimed to predict the post- operative VA of highly myopic eyes after cataract surgery.…”

“…Some AI systems predicting VA in other eye diseases have been published . Our models presented a better R 2 value compared with that of Liu et al ( R 2 = 0.6), who used an ensemble ML system to predict the outcomes in patients with diabetic macular edema after anti-VEGF injections given every 3 months.…”