Evaluation of an artificial intelligence-based intraocular lens calculator: AI-based IOL-optimized formula

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Accurate intraocular lens (IOL) power calculation is crucial for achieving desired refractive outcomes after cataract surgery.
• Existing formulas like Barrett Universal II (BUII) and Kane have limitations, especially in specific eye types.
• Advancements in artificial intelligence offer potential for improved refractive prediction accuracy.

Purpose of the Study:

• To evaluate the accuracy of the ZEISS AI IOL Calculator for refractive prediction.
• To compare the performance of the ZEISS AI IOL Calculator against the Barrett Universal II (BUII) and Kane formulas.
• To assess performance in different axial length subgroups, including short and long eyes.

Main Methods:

• Retrospective case series analysis of 10,838 eyes.
• Calculation of refractive prediction errors (PEs) using ZEISS AI, BUII, and Kane formulas.
• Statistical comparison of standard deviation (SD), root-mean-square absolute error (RMSAE), mean absolute error (MAE), and percentage of eyes within specific PE diopter ranges.

Main Results:

• The ZEISS AI IOL Calculator demonstrated superior performance compared to BUII and Kane formulas across various metrics.
• BUII showed significantly greater SDs, RMSAEs, and MAEs than ZEISS AI in the overall group and short eyes.
• ZEISS AI achieved higher percentages of eyes within ±0.50 D of PEs, particularly in short eyes, compared to BUII and Kane.

Conclusions:

• The ZEISS AI IOL Calculator offers enhanced accuracy in refractive prediction.
• Its performance is notably superior to the BUII and Kane formulas, especially for eyes with shorter axial lengths.
• AI-based calculators represent a significant advancement in optimizing IOL selection for improved patient outcomes.