Clinical Validation of the Automated Characterization of Cone Size and Center in Keratoconic Corneas

To evaluate an automated method for detecting the cone shape characteristics and to assess the cornea specialists' subjective variability of these measures using different maps.

Topographic images of the anterior and posterior surface of each eye were presented to 12 clinicians in two different types of map: tangential curvature and relative elevation to the best-fit sphere. They were asked to mark the cone center and its boundaries in the two maps without knowing that they belonged to the same patient. The results between the maps were compared to assess the subjective variability dependent on the map type and the automated method was compared against both estimations to assess its accuracy.

Considering the results of anterior and posterior surfaces, there was low agreement between the cone center estimations using different types of maps for 10 of the 12 cases (P < .05), whereas the comparison between the automated method and the two map estimations did not show differences in 11 of the 12 cases (P > .05). There was high variability, up to 55%, among clinicians' estimations of the cone area. The results of the automated method were within the range of the expert's estimations.

An objective, mathematically derived method of determining morphological dimensions of the cone was consistent with clinicians' evaluations. Although there was high variability among the experts' subjective estimates, which were highly influenced by the type of map, the objective method provided a reliable evaluation of the keratoconus shape independent of maps or color scale. [.