Strategies for formula constant optimisation for intraocular lens power calculation

Area of Science:

• Ophthalmology
• Optical Engineering
• Computational Science

Background:

• Investigating modern nonlinear iterative strategies for formula constant optimization.
• Applying these strategies to a large dataset of theoretical-optical lens power calculations.

Purpose of the Study:

• To demonstrate the application and results of nonlinear iterative strategies for optimizing intraocular lens (IOL) power calculation formulae.
• To minimize prediction errors (PE) in refractive outcomes after IOL implantation.

Main Methods:

• Implemented nonlinear iterative optimization algorithms (Levenberg-Marquardt, interior point) for various statistical metrics of prediction error (PE).
• Optimized six metrics: root mean squared (SoSPE), mean absolute (SoAPE), mean (MPE), standard deviation (SDPE), median (MEDPE), and 90% confidence interval (CLPE) of PE.
• Applied methods to five established IOL power calculation formulae (SRKT, Hoffer Q, Holladay 1, Haigis, Castrop) using a dataset of 888 eyes.

Main Results:

• Nonlinear iterative algorithms demonstrated fast and stable convergence across all formulae and optimization metrics.
• Optimization successfully minimized all evaluated statistical metrics of prediction error (PE).
• Optimizing for standard deviation (SDPE) and confidence interval (CLPE) minimized these metrics but introduced systematic offsets in mean and median PE.

Conclusions:

• Nonlinear iterative techniques can effectively minimize any statistical metric of any target parameter, such as prediction error (PE).
• These optimization strategies represent a significant advancement for refining the evaluation of IOL power calculation formulae.
• The findings highlight the potential for enhanced accuracy in refractive outcomes through advanced computational optimization.