Topography-based screening for previous laser correction of hyperopia

A screening tool based on corneal topography, for the detection of previous hyperopic laser correction is described. A total of 312 topographies were randomly selected: 251 from unoperated corneas and 61 from corneas with previous LASIK to correct hyperopia. All topographies were performed using an Orbscan II unit. LASIK surgeries were performed using a Technolas 217C excimer laser and a Hansatome microkeratome. The algorithms use two criteria: DE and DC. DE is the sum of differences between the cornea elevation and its "best-fit-sphere" for a central region minus the same sum for a mid-periphery region of the cornea. DC is the difference between the mean curvatures in each region. To classify between operated and unoperated corneas, we used the support vector machine (SVM) learning algorithm. Each criterion allows useful classification of the topographic data but curvature-based detection is more accurate with 2.4% false positive and 3.3% false negative.