Accuracy of 7 artificial intelligence-based intraocular lens power calculation formulas in medium-long eyes: 2-center study

Abstract

Objective: To compare accuracy of 7 artificial intelligence (AI)-based intraocular lens (IOL) power calculation formulas in medium-long eyes DESIGN: Retrospective observational study.

Methods: The data of patients with eyes with an axial length of 24.5 mm to 25.99 mm, who underwent phacoemulsification between May 2018 and September 2023 were analyzed. Inclusion criteria were complete biometric and refractive data. Exclusion criteria were intraoperative or postoperative complications, previous eye surgery, or corneal diseases, postoperative best-corrected visual acuity less than 0.8 and corneal astigmatism greater than 2.0 D. Prior to cataract surgery, IOL power was calculated using a Zeiss IOLMaster 700 (Carl Zeiss Meditec AG, Jena, Germany). The power of the implanted IOL was randomly selected from the outcomes of one of the following formulas, i.e., SRK/T, Holladay 2, or Barrett Universal II. Three months after phacoemulsification, refraction was measured. Postsurgery, IOL power calculations were performed using the following formulas: Hill-RBF 3.0, Kane, PEARL-DGS, Ladas Super Formula AI (LSF AI), Hoffer QST, Karmona, and Nallasamy. The main outcome measures used were SD, since prediction error (PE) was zeroed, and the percentage of eyes with PE was within ±0.50 D.

Results: One hundred eighty-four eyes with axial lengths of between 24.50 mm and 25.97 mm, were studied. The Karmona formula obtained the lowest SD (0.322) just before Hill-RBF 3.0 (0.324) and Pearl-DGS (0.334), however, without statistical significance (p > 0.05). The highest percentage of eyes with PE within ±0.50 D was achieved by Karmona (89.67%) ahead of Hill-RBF 3.0, LSF AI and Pearl-DGS (all equally 87.50 each) without statistical significance (p > 0.490).

Conclusions: All studied AI-based formulas provided highly accurate outcomes in medium-long eyes.