Evaluation Of Prediction Accuracy Of Hill-Rbf 3.0 Formula

Comparison of the accuracy of Hill-RBF 3.0 intraocular lens power calculation formula with Barrett Universal II and SRK/T formulas.

The study was conducted at Ege University Faculty of Medicine Department of Ophthalmology.

Demographic data and preoperative optical biometry measurements including keratometry, anterior chamber depth, axial length measurements of patients who underwent an uneventful cataract surgery were recorded, and spherical equivalents of refraction (SE) at the postoperative 3rd month were evaluated retrospectively. SRK/T formula was used to determine the power of the intraocular lens (IOL).

The estimated postoperative refraction values with the IOL used in the cases were re-evaluated using the Hill-RBF 3.0 and Barrett Universal II formulas. The mean absolute errors and SE values of the refractive deviations were calculated. Refractive deviations of all formulas were compared and p < 0.05 was considered as significant.

450 eyes of 341 subjects were included in the study. Mean postoperative refractive deviation of SE was 0.08 ± 0.65 [(-2.15) – (2.38)] D in the SRK/T formula, 0.00 ± 0.66 [(-2.50) - 2.95] D in the Hill-RBF 3.0 formula  and 0.04 ± 0.66 [(-2.37) – (2.96)] D in Barrett Universal II. There was no significant difference between the three formulas (p=0.18, ANOVA test). Mean absolute error of SE was 0.51 ± 0.41 (0-2.38) D, 0.5 ± 0.43 (0-2.95) D and 0.5 ± 0.43 (0.01-2.96) D for SRK/T, Hill RBF 3.0 and Barrett Universal II, respectively. There was no statistically significant difference between the groups (p=0.989, ANOVA test).

Hill-RBF 3.0, a new generation formula based on artificial intelligence, was not inferior to SRK/T and Barrett Universal II formulas.