Horizontal thickness profile and corneal deformation parameters of Corvis ST for detecting keratoconus

Session Details

Session Title: Corneal Biomechanics

Session Date/Time: Monday 09/10/2017 | 08:00-10:30

Paper Time: 09:51

Venue: Room 2.1

First Author: : B.Lopes BRAZIL

Co Author(s): :    M. Salomao   R. Vinciguerra   P. Vinciguerra   R. Ambrosio   J. Mello Amaral        

Abstract Details

Purpose:

Purpose: To evaluate ability of horizontal pachymetric profile data combined with corneal deformation parameters (CDP) obtained with Corvis ST to distinguish keratoconic from normal corneas.

Setting:

Department of Ophthalmology, Federal University of São Paulo,São Paulo, Brazil

Methods:

In a retrospective study, 1 eye randomly selected from 139 patients with clinically detected bilateral keratoconus (KC), from 335 patients with normal corneas and the eye with normal topography from very asymmetric cases (NT-VAE) were enrolled. All patients were examined by a fellowship trained corneal specialist (RA and PV) and had Corvis ST exams (Oculus, Wetzlar, Germany). Data from 4mm horizontal pachymetric profile passing through the apex was assessed. With this data, some artificial intelligence models were built to improve detection of KC. The models were also fitted with the data of CDP to address the same task. Performance in distinguishing keratoconus from normal was assessed by ROC curves. Performances of the models were assessed by the leave-one-out cross validation method

Results:

The area under the ROC curve (AUC) of pachymetry at the apex was 0.904. The best model with only pachymetric data achieved an AUC of 0.935, the best model with only CDP data achieved AUC of 0.973 which had no statistical significant different from CBI (AUC = 0.983, p=0.2) that includes pachymetry data and CDP in its calculations. The best model with both data had the best AUC 0.991, this bordered on statistical significant difference from CBI (p=0.05).

Conclusions:

This is possible to distinguish keratoconic from normal corneas exclusively based on corneal deformation response. The CDP-SVM model could be used in combination with geometric data (i.e. Horizontal thickness profile) to improve accuracy for the diagnosis of ectatic corneal diseases and to characterize ectasia susceptibility.

Financial Disclosure:

receives consulting fees, retainer, or contract payments from a company producing, developing or supplying the product or procedure presented