Posters

A novel scoring tool for the classification of early and mild keratoconus cases

Poster Details

First Author: J.Velázquez Blázquez SPAIN

Co Author(s):    F. Cavas   J. Bolarín   J. Alió              

Abstract Details

Purpose:

The creation of a novel prediction model stablished upon a set of geometrical, optical and demographical variables, with a double purpose: the classification of Keratoconus (KC) it its earliest phases; and finding out the probability of a correct/incorrect classification for each patient.

Setting:

Vissum Corporation, Alicante, Spain. Technical University of Cartagena, Spain. Technology Centre for IT and Communications (CENTIC), Murcia, Spain

Methods:

196 eyes of 196 subjects were selected (125 males; 63.8%; 71 females, 36.2%). Of these, 82 formed the “healthy” control group, while 114 suffered from KC according to the RETICS grading system (68 early KC, 46 mild KC). 27 different parameters were considered (demographic, clinical, pachymetric and geometric) in just a single eye from each patient. Ordinal logistic regression techniques were used with the data obtained, to program a web application (EMKLAS) that allowed the use of new individual’s data to perform real-time estimates.

Results:

The early and mild KC classifier proved attained good training accuracy figures, reaching a global value of 73% and a 95% confidence interval of 65%–79%. It was particularly precise in the validation by an independent sample for both the Control (79%) and Mild KC (80%) groups, whereas the value obtained for the Early KC group was remarkably lower (69%). The variables that the statistical model finally included were Age, Gender, CDVA, Q8mm and Posterior Minimum Thickness Point Deviation.

Conclusions:

This web application permits a quick, unbiased and quantitative evaluation of Early and Mild KC condition, in detection and classification terms, and can be a helpful aid for ophthalmology professionals in their diagnose procedure of the disease.

Financial Disclosure:

None