Intraocular Lens Power Calculation-Comparing Big Data Approaches to Established Formulas

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Data Science

Background:

• Traditional intraocular lens (IOL) power calculation formulas have limitations in predicting postoperative refractive outcomes.
• Accurate refractive prediction is crucial for successful cataract surgery outcomes.

Purpose of the Study:

• To compare the predictive performance of traditional IOL power calculation formulas against advanced regression models.
• To evaluate classical linear models, regression splines, and random forest regression for predicting postoperative refraction.

Main Methods:

• A retrospective analysis of 886 eyes from 631 patients undergoing cataract surgery with monofocal aspherical IOL implantation.
• Biometric data from optical biometry (IOLMaster 700) were used.
• Formula constants were optimized, and regression models were trained on datasets stratified by axial length (AL).
• Performance was assessed using mean absolute error (MAE), root mean squared error (RMSE), and prediction error variance.

Main Results:

• Regression models demonstrated superior in-sample predictive error compared to traditional IOL formulas.
• A regression spline model incorporating nonlinear effects of covariates (R2, AL, CCT) achieved the lowest out-of-sample prediction error (MAE = 0.279, RMSE = 0.359).
• This advanced model outperformed all traditional formulas and the Castrop formula.
• Random forest regression showed high in-sample accuracy but poor out-of-sample generalizability due to overfitting.

Conclusions:

• Regression models, especially those utilizing regression splines for nonlinear effects, offer a promising alternative to traditional IOL formulas.
• While linear and random forest models reduce in-sample error, their clinical utility is limited by out-of-sample performance.
• Future research should focus on enhancing generalizability and integrating machine learning for improved refractive outcomes, particularly in atypical eyes.