Automated keratoconus detection using height data of anterior and posterior corneal surfaces.

PURPOSE

To develop a keratoconus detection algorithm using the corneal topographic data of the anterior and posterior corneal surfaces.

METHODS

Topographic measurements of the cornea were made with a slit-scanning corneal topographer. We examined 120 subjects (165 eyes); keratoconus patients and keratoconus suspect patients comprised the keratoconus group, and post-photorefractive keratectomy patients, with-the-rule astigmatism patients, and controls without disease comprised the nonkeratoconus group. Two variables of the anterior corneal surface, two variables of the posterior corneal surface, and one corneal thickness variable were obtained by applying the Fourier harmonic decomposition formula. By performing a logistic regression analysis with a training set to differentiate the keratoconus group from the nonkeratoconus group, the Fourier-incorporated keratoconus detection Index (FKI) was created. The validity of the FKI was determined by using independent validation sets.

RESULTS

The FKI distinguished the keratoconus group from the nonkeratoconus group with 96.9% sensitivity and 95.4% specificity in the validation set.

CONCLUSIONS

A newly developed automated keratoconus classifier can be used to screen keratoconic patients. The index is based on information obtained by Fourier analysis from not only the anterior corneal surface but also from the posterior corneal surface and corneal thickness.