E-POSTER DETAIL | 11th Asia Pacific Tele Ophthalmology Society Symposium

Title
A Comparative Evaluation of Traditional and Novel Artificial Intelligence-Based Intraocular Lens Power Calculation Formulas in Eyes with Normal Axial Length

Authors
Zhiyun Xiao, Yannan Zhu, Xiaojian Yin

Presenting
Zhiyun Xiao

PURPOSE:
Purpose: To compare the predictive accuracy of traditional and artificial intelligence (AI)-based intraocular lens (IOL) power calculation formulas in cataract surgery.

METHODS:
Methods: A retrospective study analyzed 97 eyes undergoing cataract surgery. Four formulas were evaluated: Barrett Universal II, Hill-RBF (data supplemented by literature review), Nallasamy, and Karmona. Prediction error (PE) and absolute PE were calculated by comparing predicted and actual postoperative refraction.

RESULTS:
Results: Barrett Universal II and Karmona showed the best overall performance, with mean numerical errors of 0.05±0.50D and 0.07±0.48D, and mean absolute errors of 0.36D each. Approximately 85% of eyes achieved a PE within ±0.50D with these formulas. Hill-RBF yielded the lowest mean absolute error (0.33D). Nallasamy exhibited a myopic bias and wider error dispersion (<50% within ±0.50D). Statistical analysis indicated significant differences among formulas (p<0.05).

RESULTS:

Results: Barrett Universal II and Karmona showed the best overall performance, with mean numerical errors of 0.05±0.50D and 0.07±0.48D, and mean absolute errors of 0.36D each. Approximately 85% of eyes achieved a PE within ±0.50D with these formulas. Hill-RBF yielded the lowest mean absolute error (0.33D). Nallasamy exhibited a myopic bias and wider error dispersion (<50% within ±0.50D). Statistical analysis indicated significant differences among formulas (p<0.05).

CONCLUSIONS:
Conclusion: The Barrett Universal II and AI-based Karmona formulas demonstrated high and comparable accuracy, supporting the clinical integration of AI-driven IOL calculations. Further validation in diverse populations is warranted.