Estimation of the internal astigmatism using algorithm based on big data analysis

Session Details

Session Title: Toric IOL Calculations & Alignment

Session Date/Time: Sunday 23/09/2018 | 08:00-10:00

Paper Time: 08:52

Venue: Room A5

First Author: : L.Gauthier FRANCE

Co Author(s): :    D. Smadja   V. Bart                 

Abstract Details

Purpose:

To evaluate a new approach that estimate the internal ocular astigmatism using a dynamic algorithm based on big data analysis from thousands of biometric ocular data.

Setting:

Private practice – Ophthalmology department, Clinique cote Basque Sud, Saint Jean de Luz, France. Ophthalmology Department, Shaare Zedek Medical Center, Jerusalem, Israel

Methods:

Anterior corneal astigmatism and manifest cylinder adjusted to the corneal plane were extracted from a database of 7274 phakic and presbyopic eyes under 60 years with clear lens. A computation of the internal astigmatism was performed by a vectorial substraction between the manifest cylinder adjusted to the corneal plane and the anterior astigmatism based on patient data history. Then, the K-Nearest Neighbour algorithm was taught to estimate the internal astigmatism according to same computation method, but on the most similar population (eyes that express the closest biometric data), what is called the “nearest neighbours”.

Results:

In most of anterior corneal astigmatism presented to the algorithm , a minimum of 100 “nearest neighbours cases” were found within 0.5 diopter of corneal anterior astigmatism or 5 degrees of similarity. The predicted internal astigmatism was found against the rule in most cases (0.63 against the rule for 1 diopter anterior with the rule astigmatism). However, in with the rule corneal anterior astigmatism greater than 2.5 diopters, the internal astigmatism was most likely with the rule. The magnitude and axis of internal astigmatism were found to vary depending on anterior corneal astigmatism but also on gender and initial keratometry.

Conclusions:

We proposed a new automated method, based on big data analysis, for estimating internal astigmatism using the most available and general biometric ocular data. Although this method still needs to be validated in pseudophakic eyes and compared to direct measurement of posterior surface, it might provide an easy and very accessible way to estimate the internal astigmatism based on observation and not on a dogmatic formula.

Financial Disclosure:

... has significant investment interest in a company producing, developing or supplying product or procedure presented, ... research is funded, fully or partially, by a company producing, developing or supplying the product or procedure presented