Photo Credit: sdigital

The following is a summary of “Accuracy of 7 artificial intelligence based intraocular lens power calculation formulas in extremely long Caucasian eyes,” published in the November 2024 issue of Ophthalmology by Stopyra et al. 

Researchers conducted a retrospective study to compare the accuracy of 7 AI-based Interocular lens (IOL) power calculation formulas in extremely long eyes.  

They analyzed patients with highly myopic eyes who underwent uneventful phacoemulsification surgery. Preoperatively, IOL power was calculated using the Barrett Universal II, Holladay 2, SRK/T, and 3 months post-surgery, refraction measurements were obtained. The IOL power calculations were then performed using various formulas, including Hill-RBF 3.0, Kane, PEARL-DGS, Ladas Super Formula AI (LSF AI), Hoffer QST, Karmona, and Zhu-Lu. The accuracy of the predictions was evaluated by root mean square absolute error (RMSAE), median absolute error (MedAE), and the percentage of eyes with prediction error (PE) within ±0.50 D.  

The results showed that 48 eyes with axial lengths greater than 30.00 mm were analyzed. The Hill-RBF 3.0 demonstrated the lowest RMSAE of 0.788, with statistical significance only compared to Karmona (0.956, P =0.021). For MedAE, both Hoffer QST (0.442) and Hill-RBF (0.490) outperformed LSF AI (0.800), with significant differences (P =0.013, P =0.008, respectively). Hill-RBF 3.0, Kane, and Hoffer QST achieved the highest percentage of eyes with PE within ±0.50 D, at 54.17% each. Statistical significance was noted for both Hill-RBF and Kane when compared to LSF AI (27.08%) and Karmona (39.58%), as well as Hoffer QST vs LSF AI.  

Investigators concluded that all formulas showed similar trueness, with Hill-RBF 3.0 outperforming Karmona RMSAE and LSF AI being less accurate than Hoffer QST and Hill-RBF 3.0 (MedAE). 

Source: ajo.com/article/S0002-9394(24)00519-1/abstract