Article Title: Machine Learning Models for Predicting Long-Term Visual Acuity in Highly Myopic Eyes
Key Takeaway Points:
High myopia, being one of the top 3 causes of irreversible blindness in Asia, Europe, and North America, is estimated to affect nearly 1 out of every 10 people by 2050. This study aimed to utilize machine learning models for the prediction of long-term visual acuity (VA) in highly myopic eyes. Researchers performed a retrospective observational study utilizing EMR data from 2011 to 2021 at a single center in Tokyo.
Algorithms were developed for the analysis of data from the 967 patients in the study. The primary outcome was best-corrected visual acuity (BCVA) at 3 and 5 years, with the secondary outcome being the risk of developing visual impairment (VI) at 5 years. The support vector machines model performed best for the prediction of BCVA at 3 years, with the random forest model performing optimally for BCVA prediction at 5 years. Logistic regression performed best for predicting the risk of visual impairment at 5 years. Poor baseline BCVA, prior myopic macular neovascularization, older age, and category 4 pathologic myopia were highlighted as having the greatest importance for the models’ predictions; these were associated with increased risk of VI at 5 years.
Since the study used data from a single center without external validation, it remains unclear whether the results can be generalized to other populations. That said, the author’s machine learning models performed quite well, and the findings suggest that artificial intelligence may be used to predict both BCVA and the risk of developing VI in a patient with high myopia. The findings from this study may allow clinicians to triage which patients with high myopia should be closely monitored and could lead to new management approaches.
Publication Date: October 26, 2023
Reference: Wang Y, Du R, Xie S, et al. Machine Learning Models for Predicting Long-Term Visual Acuity in Highly Myopic Eyes. JAMA Ophthalmol. 2023;141(12):1117–1124. doi:10.1001/jamaophthalmol.2023.4786
Summary by: Daniel Lamoureux
Comments