Current topography-based methods for keratoconus screening have limitations. Researchers at Doheny Eye Institute have been working to develop a method for capturing focal thinning that is characteristic of keratoconus using optical coherence tomography-derived pachymetry maps. A method based on fitting the data to a Gaussian waveform is showing promise.
"We know that topography does not screen out all eyes at risk for post-LASIK ectasia, and while corneal thickness offers another clue, it is not very specific, because thin corneas have been operated on without developing ectasia," said Dr. Huang, holder of the Charles C. Manger III, MD, Chair in Corneal Laser Surgery and an associate professor of ophthalmology and biomedical engineering, Doheny Eye Institute, University of Southern California, Los Angeles. "Looking at the pattern of focal cornea thinning to capture focal thinning that is characteristic of keratoconus may be a better approach to using pachymetric information.
"In a previous study, we developed a system using a general statistical description of an OCT pachymetry map to detect focal thinning based on a set of six parameters and demonstrated it had very good diagnostic accuracy for categorizing keratoconus compared with a topography-based method in a small body of test patients," he said. "This new approach to matching the focal thinning with a Gaussian waveform is much cleaner, because it involves fewer parameters. Now its clinical utility should be further evaluated in a larger study that should include eyes with forme fruste keratoconus."
Obtaining the map
The first step in obtaining the corneal pachymetry map is to create a reference map using eyes without keratoconus. Then, an inverse normalized pachymetry map is derived by dividing the reference map by the map of the eye under analysis.
"In an eye with keratoconus, this inverse normalized pachymetry map shows a peak, revealing the cone, whereas in a normal eye, the map has a uniform appearance," Dr. Huang said.
Next, the depth and width of focal thinning in each eye is quantified by fitting a rotationally symmetric, two-dimensional Gaussian waveform to the inverse normalized pachymetry map. The Gaussian waveform is centered at the peak, and the width is adjusted to maximize the cross-correlation product between the waveform and the map.
The technique was investigated in a retrospective study involving 15 eyes without keratoconus and 15 eyes with it. Comparisons of the average cone depth and average cone diameter showed statistically significant differences between the normal and keratoconus groups. For the eyes with keratoconus, the mean width of the cone diameter was 4.93 mm, and the depth measurements showed an average of about 30% focal thinning. For the normal eyes, mean width was about 7 mm, and the depth measurement indicated negligible thinning-about 3% to 4%.
"In this small initial sample, the technique demonstrated 100% sensitivity and 100% specificity in diagnosing keratoconus," Dr. Huang concluded.