Free Papers

New algorithms based on artificial intelligence (AI) to improve the accuracy of intraocular lens (IOL) power calculation

Free Paper Details

First Author: L.Gauthier-Fournet FRANCE

Co Author(s):    V. Bart   A. Barnoud   -. -   -. -           

Abstract Details

Purpose:

To present two new approaches to calculate the emmetropic power of an IOL based on multi-parametric analysis of the pre-operative characteristics. The first approach (AI1) directly predicts post-operative deviation from SRKT refractive target. The second approach predicts directly the effective lens position (ELP) based on pure optical formula.

Setting:

Private practice – Ophthalmology department, Clinique cote Basque Sud, Saint Jean de LUZ, France.

Methods:

A training dataset of 1335 eyes and second testing dataset of 1442 eyes were established with preoperative biometric data (IOL-Master 500) and refractive result measured one month after surgery. First ,we trained an AI algorithm to learn with several key variables (AL, KM, Age, ACD...) to predict post-operative deviation from SRKT refractive target to get new refractive targets.Secondly, we developed a pure optical formula linking the ELP to the measured variables. Based on this we trained a model to learn with key variables to predict this ELP. By inverting the formula, with the predicted ELP, we obtained new refractive targets.

Results:

The first AI algorithm (AI1) based on SRK-T is as good as BARRETT formula while the second (AI2) algorithm based on its own optical formula increases significatively the IOL calculation accuracy. The predicted refractive results of +/- 0.5 dioptre of emmetropia reach from 64% for SRK-T, 68% for BARRETT and 68% for AI1 to 70% for the new algorithm AI2. The prediction standard deviation goes from 0.56 for SRKT, 0.56 for BARRETT and 0.52 for AI1 to 0.50 for AI2. The mean absolute error goes from 0.44 for SRKT, 0.41 for BARRETT and 0.41 for AI1) to 0.39 for AI2.

Conclusions:

We developed two approaches based on Artificial Intelligence to improve IOL recommendation accuracy: one based on the SRK-T formula and one based on an optical formula. The second one is the more efficient formula over performing BARRETT formula.

Financial Disclosure:

... has significant investment interest in a competing company, ... is employed by a for-profit company with an interest in the subject of the presentation