Purpose:

To evaluate a novel hybridized intraocular lens (IOL) formula that integrates artificial intelligence (AI) and to compare it with two existing IOL formulae.

Methods:

A total of 200 consecutive eyes from a single surgeon (DRH) were evaluated. Each eye underwent an uncomplicated cataract surgery and received a single type of lens. Three formulas were evaluated: Ladas Super Formula (LSF), Barrett Universal II, and a novel hybrid formula. This novel formula was constructed by adjusting the original LSF with AI-trained algorithms based on standard input parameters. The formula was trained with 100 of the 200 randomly selected eyes. The remaining 100 eyes were used to evaluate the performance of this new formula. Each formula was evaluated by calculating the mean absolute error (MAE) and the number of eyes that were within 0.50 D of the predicted refraction.

Results:

The original LSF and the Barrett Universal II formula were comparable with 76% and 72% of eyes being less than 0.50 D from predicted refraction and their MAE being 0.37 D and 0.38 D (P>0.05), respectively. The novel AI-trained hybrid formula calculated 80% of eyes to be less than 0.50 D from predicted refraction. It also decreased the MAE to 0.32 D which was statistically significant when compared to the MAE of the Barrett Universal II formula (P<0.05).>

Conclusions:

This novel formula demonstrates an improvement in both the number of eyes within 0.50 D of predicted refraction as well as decreased MAE. We also demonstrate that this method can be trained using as few as 100 eyes to create a statistically significant improvement in IOL formulae. Future studies will provide comparisons between traditional IOL formulae and this novel methodology using additional input parameters.