ESCRS - PP17.01 - Predictive Factors For Visual Performance After Bi-Aspheric Ablation Profile For Presbyopic Corneal Treatment Using Amaris With Presbymax Module

Predictive Factors For Visual Performance After Bi-Aspheric Ablation Profile For Presbyopic Corneal Treatment Using Amaris With Presbymax Module

Published 2024 - 42nd Congress of the ESCRS

Reference: PP17.01 | Type: Free paper | DOI: 10.82333/yfwr-h074

Authors: Pierre BAUDU* 1 , Soodabeh Darci 2 , Samuel Arba Mosquera 3 , Franck Penin 4

1Clinique Avicennne,Le Port,Réunion, 2SCHWIND eye-tech-solutions,Kleinostheim,Germany, 3SCHWIND eye-tech-solutions,Kleinostheim,Germany;Recognized Research Group in Optical Diagnostic Techniques,University of Valladolid,Valladolid,Spain;Department of Ophthalmology and Sciences of Vision,University of Oviedo,Oviedo,Spain, 4SCHWIND France eye-tech-solutions,Dietwiller,France

Purpose

To investigate the predicting factors to estimate the changes in distance-corrected near visual acuity (DCNVA) and corrected distance visual acuity (CDVA) (both monocular and binocular) 6-month after bi-aspheric multifocal central presbyLASIK treatments for myopia and hyperopia with or without astigmatism, based on artificial intelligence (AI) Machine Learning (ML) models.

Setting

Private practice.

Methods

We analysed CDVA and DCNVA, both mono- and binocularly, as well as their pre- to postoperative changes.  We analysed Machine Learning correlations of those changes with the preoperative status and planned features (19 features).  We formally defined success as no loss in CDVA (change in CDVA >= 0, safety) and gain in DCNVA (change in DCNVA > 0, efficacy).

Results

A cohort of 716 eyes of 358 patients treated using PresbyMAX software were reviewed after 6-month follow-up was completed for developing the predictive models.  Changes in CDVA correlated to planned addition, planned OZ and age, changes in DCNVA correlated mainly to age and planned addition.  The predictions for monocular and binocular success resulted in 3% false positive and 3% false negatives monocularly and 2% false positive and 2% false negatives binocularly.

Conclusions

AI-ML predictive models based upon preoperative and plan data may help evaluating PresbyLASIK candidates.  However, a number of false negatives has been found.  Further refinements and probably more comprehensive predictive models shall be developed.  Predictive models combined with expectation management is essential to achieve patient satisfaction.  Further investigation is necessary to evaluate the overall benefit of this procedure.