Longitudinal Refractive Evolution In Children With Aphakia Post Surgery: A Machine Learning Analysis
Published 2026 - 44th Congress of the ESCRS
Reference: FP15.07 | Type: Free paper | DOI: 10.82333/pr74-jt50
Authors: Dr Zhixiang Hua* 1 , Dr Lin Wang 1 , Dr Yang Shen 1 , Dr Jin Yang 1 , A/Prof. Xingtao Zhou 1
1Department of Ophthalmology,Eye & ENT Hospital of Fudan University,Shanghai,China
Purpose
To evaluate the feasibility of machine learning algorithms in predicting refractive evolution and to identify critical developmental transition points in children with aphakia, utilizing a large dataset of optometric records.
Setting
Eye and ENT Hospital of Fudan University, Shanghai, China.
Methods
This present study was conducted on 476 patients (764 eyes) who underwent congenital cataract surgery at the Eye and ENT Hospital of Fudan University between 2005 and 2025, involving a total of 3170 follow-up visits. The evolution and trends of refractive components were analyzed. Nine ML algorithms, including Gradient Boosting Decision Tree (GBDT), Random Forest, and Support Vector Machine, were trained using an 80/20 train-test split and 5-fold cross-validation. Predictive performance was compared across age groups, and a segmented threshold sensitivity analysis was applied to determine the optimal developmental transition point.
Results
Both spherical equivalent and sphere power in aphakic eyes exhibited significant non-linear declines with age. The refractive shift rate progressively decelerated from -1.38 D/year in the 0 to 1 year group to -0.33 D/year in the 4 to 5 year group. Among the models evaluated, GBDT demonstrated superior predictive performance, significantly outperforming traditional linear and logarithmic models. The predictive performance of GBDT was markedly higher in children under 4 years compared to those aged 4 and older. Segmented threshold analysis confirmed 4 years as the pivotal inflection point, characterized by the maximum performance disparity (ΔR2) between age cohorts.
Conclusions
Machine learning is a feasible and robust tool for predicting refractive evolution in pediatric aphakia. The transition point at age 4 serves as a critical node for clinical assessment and intervention. These findings provide a data-driven basis that may help to optimize the timing of clinical interventions and support the implementation of management strategies in two stages.