April 02, 2025
Expectations after cataract surgery can be high. Patients anticipate better vision and typically expect to decrease their dependency on glasses. With these expectations, there is a constant call for improvement in intraocular lens (IOL) power calculation accuracy.
"A patient's quality of life can be tremendously improved if they no longer require glasses for all daily tasks," says Ruti Sella, M.D., an ophthalmologist and researcher at Mayo Clinic in Rochester, Minnesota. To achieve improved accuracy in an increased number of patients — while also improving accuracy in patients with less common eye measurements — the focus shifts to improving the prediction accuracy of existing formulas by adopting new technologies.
"Harnessing the use of AI to build better formulas that can learn and develop with time and additional data is a key step in moving forward," says Dr. Sella.
Dr. Sella's recent research, published in the December 2024 issue of Eye, analyzed currently available AI-driven IOL calculation formulas, together with established formulas. Different formulas were compared using current guidelines and novel analysis tools.
AI-based formulas typically improve over time, especially when more high-quality data is available to fine-tune the algorithms. "It's important to stop from time to time and assess the options we have today for IOL calculations," says Dr. Sella. "This helps us understand which option currently provides the best accuracy overall and for specific types of eyes as well."
To understand which formula showed the highest accuracy, researchers at Mayo Clinic, Moran Eye Center in Utah and Rabin Medical Center in Israel completed an in-depth analysis. This included 404 eyes of 404 patients who underwent uneventful cataract surgery. "We compared the refractive goal in those eyes to the actual outcome," says Dr. Sella. "This was based on a set of different mathematical formulas that are widely available for surgeons to use."
To perform this comparison, meticulous statistics and metrics were applied. "Our study shows that overall, AI-based formulas for IOL calculations perform very well," says Dr. Sella. "Specifically, the Nallasamy formula — which integrates a machine learning algorithm — exhibited superior accuracy in predicting postoperative refraction in our group of analyzed eyes."
Researchers also validated the use of a new readily available and user-friendly online tool for evaluating IOL power calculation formulas. "Eyetemis is available online at no cost and provides beautiful graphs to go along with your comparisons, which is valuable for research purposes or for any surgeon evaluating their own results," says Dr. Sella.
"We utilized a robust database from the Moran Eye Center along with the expertise in IOL calculation research projects of my former research group in Israel," says Dr. Sella. "With Mayo Clinic overseeing and guiding the process to support the study and publication, it's an excellent example of the importance of collaboration across institutions."
Work is currently underway on several other projects evaluating the accuracy of available and upcoming formulas in eyes with cataracts, which can be more challenging for IOL calculations.
"For example, we are looking into new formulas for calculations in short eyes where each measurement error translates into a more significant refractive error," says Dr. Sella. "We're also looking into eyes postrefractive surgery and eyes with corneal diseases like keratoconus where direct measurement of the posterior cornea may play an important role."
Mayo Clinic's Ophthalmology Parametric Universal Search (OPUS) database also provides a unique opportunity to detect less represented eyes and improve outcomes. OPUS is a powerful AI-bioinformatic system that draws from 25 different databases, all of which are supported by Mayo Clinic's Center for Digital Health.
"Our hope is to utilize all of our resources to provide patients with more predictable and accurate refractive results after surgery to meet their wishes and decrease glasses dependency," says Dr. Sella.
For more information
Reitblat O, et al. Accuracy assessment of artificial intelligence IOL calculation formulae: Utilizing the heteroscedastic statistics and the Eyetemis Analysis Tool. Eye. 2024;38:3578.
Refer a patient to Mayo Clinic.