Tomographic Detection of Keratoconus by Combining Anterior, Posterior, and Pachymetric Versions of The Cone Location and Magnitude Index (CLMI)

Abstract

 

Purpose:

 

To test the ability of the CLMI (Cone Location and Magnitude Index) algorithm, when applied to other tomographic maps, to separate keratoconic (KCN) and normal (NL) subjects.

 

Methods:

 

Dual Scheimpflug-Placido tomography data were analyzed from three independent datasets, 1 development (97 KCN, 84 NL) and 2 validation (36 KCN, 67 Nl; and 32 KCN, 93 NL). The CLMI algorithm was applied to anterior/posterior curvature (axial and instantaneous) and posterior elevation to locate the steepest and highest circular regions, 2.0 mm and 0.5 mm diameter, respectively. A modified algorithm, which searches for minimal regions, was applied to pachymetry maps to locate the thinnest 1.0 mm region. T-tests were used to determine differences (Bonferroni p<0.0014). Stepwise logistic regressions were used to determine the strongest predictors of separation.

 

Results:

 

All parameters (anterior axial CLMI and region, anterior tangential CLMI and region, posterior axial CLMI and region, posterior tangential CLMI and region, posterior elevation CLMI and region, pachymetry CLMI and region, and posterior BFS region) were significantly different between groups (p<0.0001). Two anterior parameters were the strongest predictors of the presence of KCN in the development set, but complete separation was only achieved when all parameters were used. The all parameter logistic model subsequently produced near complete separation for validation set 1 and complete separation for validation set 2.

 

Conclusions:

 

Anterior CLMI parameters are the strongest predictors of the presence of keratoconus. However, the addition of posterior and pachymetric parameters dramatically improve separation of populations, which is important for screening.

 

Keywords: keratoconus • topography