Forme Fruste Keratoconus Detection by Pattern Analysis of Corneal, Epithelial, and Stromal Thickness Maps with Optical Coherence Tomography

Abstract

 

Purpose

 

To screen forme fruste keratoconus (FFK) by analyzing corneal, epithelial and stromal thickness map patterns with optical coherence tomography (OCT).

 

Methods

 

A 26,000 Hz Fourier-domain OCT system (RTVue CAM, Optovue, Inc.) with 5 µm axial resolution was used. A “Pachymetry+Cpwr” scan pattern (6 mm scan diameter, 8 radials, 1024 axial-scans each, repeat 5 times) centered on the pupil was used to image the cornea of normal and FFK eyes. A computer algorithm was developed to automatically calculate the corneal, epithelial and stromal thickness maps. The pattern map was defined as the thickness map divided by the average thickness of the map. The normal subjects were divided into the training and evaluation groups. The OCT maps of the training group were averaged and normalized to serve as the normal average pattern maps. The corneal, epithelial and stromal thickness map pattern standard deviation (PSD) values were calculated by root-mean-square of the difference between the individual pattern maps and the normal average pattern maps. Diagnostic accuracy was evaluated by the area under the receiver operating characteristic curve (AROC) for distinguishing FFK eyes from normal eyes (evaluation group).

 

Results

 

From the normal group, 108 eyes of 57 subjects were assigned for training (defining the normal pattern maps) and 42 eyes of 22 subjects were used for AROC evaluation. The PSD values for stromal, corneal, and epithelial thickness maps were all significantly (p<0.001) higher in FFK eyes, compared to the normal evaluation group (Table 1). The AROC values were 0.988 (pachymetry map PSD), 1.00 (epithelial thickness map PSD), and 0.967 (stromal thickness map PSD).

 

Conclusions

 

High-resolution Fourier-domain OCT is able to map corneal, epithelial, and stromal thicknesses. Characteristic corneal and its sub-layer thickness changes in FFK could be detected with very high accuracy using PSD variables. These novel diagnostic variables may be useful in the detection of early keratoconus.

 

 

Table 1. Pattern Standard Deviations of Corneal Maps

 

Keywords: 550 imaging/image analysis: clinical • 574 keratoconus