Early keratoconus detection by patient-specific 3D modelling and geometric parameters análisis

Abstract

The purpose of this paper is to assess if the corneal geometric 
modelling analysis could be useful for detection of early keratoconus 
(early KC) with normal visual acuity by discriminating from those 
normal cases. The study included a total of 286 sub-jects (149 
early KC eyes associated to normal vision with mean age of 36.14 
± 10.28 years, and 137 normal eyes with mean age of 34.38 ± 7.02 
years). A reconstruction from the raw data of Sirius ScheimpflugPlacido corneal tomography and a posterior analysis of the virtual 
3D custom model were performed. The morphogeometric variables 
extracted from the corneal model were statistically analysed for 
both studied groups. Finally, receiver operator characteristic (ROC) 
curves were established to determine their predictive values and 
accuracy parameters. Thirteen of the fourteen morphogeometric 
measurements reached significant differences between groups (P < 
.05). Among the efficiency discrimination by ROC curve, six of the 
modelled variables obtained an area under the ROC curve over 0.7, 
these are: sagittal plane apex area, anterior corneal surface area, 
sagittal plane area in minimum thickness point, net deviation from 
centre of mass XY and total corneal volume, where the posterior 
apex deviation had the greatest area under the ROC curve (area: 
0.856, sensitivity: 79.3%, specificity: 78.5%). The analysis of corneal 
geometric custom modelling demonstrates to be a new and useful 
tool for the practice of refractive surgery by providing excellent 
accuracy to detect early corneal deformation in KC patients with 
normal visual acuity.