Search engine for discovering works of Art, research articles, and books related to Art and Culture
ShareThis
Javascript must be enabled to continue!

Predicting Cycloplegic Spherical Equivalent in Children and Adolescents: A Comparative Machine Learning Approach with SHAP-Based Model Interpretation

View through CrossRef
Abstract Objective: To develop and validate machine learning models for predicting cycloplegic spherical equivalent refraction (SER), myopia status, and the need for cycloplegic dilation using non-cycloplegic clinical parameters in a pediatric population. Methods: A total of 1,224 eyes from 612 children (aged four to 17 years) were included. Data were collected from two clinical centers with harmonized 27-field standards. Three prediction tasks were established: (1) the need for cycloplegic refraction (based on absolute spherical equivalent shift >0.5 D), (2) cycloplegic myopia status (SER ≤ -0.5 D), and (3) the exact post-cycloplegic spherical equivalent in children and adolescents. Feature engineering incorporated absolute values, fellow-eye features, interocular differences, and derived biometric indices. Multiple algorithms spanning tree-based ensembles (XGBoost, LightGBM, Random Forest, Extra Trees, Hist Gradient Boosting), neural networks (MLP variants), and classical classifiers were evaluated. Models were trained on 70% of the data and validated on 30% of the data, with stratified sampling. Performance was assessed using accuracy, AUC, RMSE, and R², with bootstrap 95% confidence intervals and SHAP analysis for interpretability. Results: For predicting the need for cycloplegic refraction, XGBoost achieved the highest accuracy of 0.818 (AUC = 0.780), with precision 0.696 and recall 0.464. For myopia status classification, LightGBM achieved accuracy 0.908 and AUC 0.934, with precision 0.906 and recall 0.974. For post-cycloplegic SER regression, the MLP-linear regressor yielded the lowest RMSE of 0.630 D, MAE 0.424 D, and R² = 0.881. SHAP analyses showed that the need for cycloplegic refraction was driven by axial length, while myopia status was dominated by non-cycloplegic SER. Conclusions: Machine learning models using non-cycloplegic parameters can accurately predict cycloplegic refractive outcomes and identify children who need dilation. This approach may offer a scalable, resource-efficient solution for pediatric refractive screening and epidemiological research where cycloplegic examination is impractical.
Title: Predicting Cycloplegic Spherical Equivalent in Children and Adolescents: A Comparative Machine Learning Approach with SHAP-Based Model Interpretation
Description:
Abstract Objective: To develop and validate machine learning models for predicting cycloplegic spherical equivalent refraction (SER), myopia status, and the need for cycloplegic dilation using non-cycloplegic clinical parameters in a pediatric population.
Methods: A total of 1,224 eyes from 612 children (aged four to 17 years) were included.
Data were collected from two clinical centers with harmonized 27-field standards.
Three prediction tasks were established: (1) the need for cycloplegic refraction (based on absolute spherical equivalent shift >0.
5 D), (2) cycloplegic myopia status (SER ≤ -0.
5 D), and (3) the exact post-cycloplegic spherical equivalent in children and adolescents.
Feature engineering incorporated absolute values, fellow-eye features, interocular differences, and derived biometric indices.
Multiple algorithms spanning tree-based ensembles (XGBoost, LightGBM, Random Forest, Extra Trees, Hist Gradient Boosting), neural networks (MLP variants), and classical classifiers were evaluated.
Models were trained on 70% of the data and validated on 30% of the data, with stratified sampling.
Performance was assessed using accuracy, AUC, RMSE, and R², with bootstrap 95% confidence intervals and SHAP analysis for interpretability.
Results: For predicting the need for cycloplegic refraction, XGBoost achieved the highest accuracy of 0.
818 (AUC = 0.
780), with precision 0.
696 and recall 0.
464.
For myopia status classification, LightGBM achieved accuracy 0.
908 and AUC 0.
934, with precision 0.
906 and recall 0.
974.
For post-cycloplegic SER regression, the MLP-linear regressor yielded the lowest RMSE of 0.
630 D, MAE 0.
424 D, and R² = 0.
881.
SHAP analyses showed that the need for cycloplegic refraction was driven by axial length, while myopia status was dominated by non-cycloplegic SER.
Conclusions: Machine learning models using non-cycloplegic parameters can accurately predict cycloplegic refractive outcomes and identify children who need dilation.
This approach may offer a scalable, resource-efficient solution for pediatric refractive screening and epidemiological research where cycloplegic examination is impractical.

Related Results

Primerjalna književnost na prelomu tisočletja
Primerjalna književnost na prelomu tisočletja
In a comprehensive and at times critical manner, this volume seeks to shed light on the development of events in Western (i.e., European and North American) comparative literature ...
Prediction of Cycloplegic Refractive Error based on Non-Cycloplegic Measurements in Chinese Children aged 3 to 17 years
Prediction of Cycloplegic Refractive Error based on Non-Cycloplegic Measurements in Chinese Children aged 3 to 17 years
Abstract Introduction Cycloplegic refraction remains the gold standard for pediatric refractive assessment, yet its implementation faces challenges in large-scale studies. ...
Frequency of Common Chromosomal Abnormalities in Patients with Idiopathic Acquired Aplastic Anemia
Frequency of Common Chromosomal Abnormalities in Patients with Idiopathic Acquired Aplastic Anemia
Objective: To determine the frequency of common chromosomal aberrations in local population idiopathic determine the frequency of common chromosomal aberrations in local population...
Pediatric cycloplegic refraction
Pediatric cycloplegic refraction
Background: Cycloplegic refraction is an essential part of the pediatric ophthalmic assessment and is the cornerstone of strabismus evaluation. This narrative review aimed to ascer...
Correlation between ocular biological parameters and cycloplegic refractive shift in pediatric myopia
Correlation between ocular biological parameters and cycloplegic refractive shift in pediatric myopia
AIM: To analyze the correlation between ocular biometric parameters and cycloplegic refractive shift (ΔSE) and to provide evidence for clinical decision-making regarding cycloplegi...
Vector SHAP Values for Machine Learning Time Series Forecasting
Vector SHAP Values for Machine Learning Time Series Forecasting
ABSTRACTWe propose a new vector SHapley Additive exPlanations (SHAP) to interpret machine learning models for forecasting time series using lags of predictor variables. Unlike the ...
Selection of Injectable Drug Product Composition using Machine Learning Models (Preprint)
Selection of Injectable Drug Product Composition using Machine Learning Models (Preprint)
BACKGROUND As of July 2020, a Web of Science search of “machine learning (ML)” nested within the search of “pharmacokinetics or pharmacodynamics” yielded over 100...
Myopia Prediction Using Machine Learning: An External Validation Study
Myopia Prediction Using Machine Learning: An External Validation Study
We previously developed machine learning (ML) models for predicting cycloplegic spherical equivalent refraction (SER) and myopia using non-cycloplegic data and following a standard...

Back to Top