Digital image analysis of optical sections of the cornea for the diagnosis of early keratoconus

A set of mathematical techniques and algorithms for the diagnosis of early keratoconus has been developed. The methods are based on the use of original features obtained by recognition of transverse optical sections of the cornea in images done with the camera based on Scheimpflug principle used in the Pentacam device (Oculus, Germany). In particular, a method is proposed for calculating the radius of curvature of the corneal borders, based on their approximation by a regression spline with a penalty function. The use of the stroma brightness in a digital image is proposed as a characterizing feature of the structure — an indicator of the light scattering intensity by the cornea material, or the brightness index of the cornea. A classifier is constructed that ensures the separability of classes 0 (norm) and 1 (early keratoconus) in a two-dimensional feature space: the position of the junction of the regular surface of the cornea and the ectasia zone, the brightness index of the cornea.