Ai-Based Versus Non Ai Based Iol Power Formulas: A Comparative Analysis Of Refractive Outcomes Using The Escrs Online Calculator
Published 2026 - 44th Congress of the ESCRS
Reference: POCAT601 | Type: Free paper | DOI: 10.82333/e9z7-ay92
Authors: Dr Raj Vaishal Kenia* 1 , Dr Shweta Dhiman 1 , Dr Smruti ranjan Dethi 1 , Dr Santosh Mahapatra 1 , Dr Shraddha Shataparba 1
1Ophthalmology ,JPM ROTARY EYE HOSPITAL ,Cuttack,India
Purpose
To prospectively compare the refractive prediction accuracy of artificial intelligence (AI)–based and modern theoretical intraocular lens (IOL) power calculation formulas in adult cataract surgery, and to evaluate whether AI-driven formulas provide a clinically meaningful advantage when calculated using the standardized ESCRS online platform.
Setting
J.P.M. Rotary Eye Hospital & Research Institute, Cuttack, India.
Methods
This prospective comparative study included 195 eyes undergoing uncomplicated phacoemulsification with monofocal IOL implantation. Biometry was performed using the IOLMaster 700. Seven modern formulas—Barrett Universal II, Cooke K6, EVO, Hill-RBF 3.0, Hoffer QST, Kane, and PEARL-DGS—were evaluated via the ESCRS online calculator.
Prediction error (PE) and absolute prediction error (APE) at 1 month were analyzed. Primary outcomes were mean prediction error (MPE), median absolute error (MedAE), and percentage of eyes within ±0.50 D. Subgroup analysis was performed across axial length categories.
Results
Mean axial length was 23.01 ± 0.84 mm. All formulas demonstrated minimal systematic bias, with MPE ranging from +0.023 D to +0.038 D (p = 1.000).
Median absolute error ranged from 0.13 to 0.14 D across formulas. Numerically, Hill-RBF demonstrated the lowest mean absolute error (0.2549 D), closely followed by Kane (0.2571 D), while Hoffer QST showed the highest mean absolute error (0.2680 D); however, these differences were not statistically significant (p = 0.987).
Approximately 85–90% of eyes were within ±0.50 D and >95% within ±1.00 D across all formulas. No significant inter-formula differences were observed within short, normal, or long axial length subgroups.
Conclusions
AI-based and modern theoretical IOL formulas demonstrated statistically equivalent refractive performance using the ESCRS online calculator. Although Hill-RBF and Kane showed marginally lower absolute errors, the differences were clinically insignificant. Contemporary IOL formulas appear to have reached a performance plateau in routine adult cataract eyes.