Distinction between Early Keratoconus with Normal Vision and Normal Cornea based on Pattern Recognition Analysis

Abstract

Aim: To find the profile that differentiates most normal corneas from early keratoconus with normal vision. Materials and methods:Multicentric, comparative study including a total of 995 eyes and divided into two groups: 625 eyes suffering from early keratoconus but with normal vision [spectacle corrected distance visual acuity (CDVA) of 0.9 decimal or better] and 370 normal control eyes with same normal vision level. To ascertain the main differences that would allow the identification of the keratoconic eyes from normals, a pattern recognition analysis was performed combining two statistical methods: Principal component analysis (PCA) and discriminant analysis. Visual and refractive parameters, corneal topography, aberrometry, and PCA were evaluated in both groups. Results:The application of the PCA with Varimax rotation offered a total of five factors which explains the 85.51% of the total variability. Discriminant analysis indicated that factors 1 and 3 were at the greatest discriminating capacity. From a total of 318 cases, the newly identified abnormal pattern profile allowed the recognition of 275, which presents a sensitivity and specificity of 71.6 and 97.3% respectively. Conclusion:In eyes with normal CDVA, those factors related to the nonorthogonal shape irregularity of the cornea and the refractive power are the ones that showed more discriminating capabilities between normal and early keratoconic eyes.Clinical significance: Principal component analysis allows to correctly discriminate between normal and mild keratoconus patients; additionally, this method is not restricted to a particular corneal topography technology and is available to any normally equipped ophthalmology office.