Using artificial intelligence in the interpretation of corneal topography for laser vision correction

Abstract

METHODS:In this retrospective cohort study, we analyzed the tomography of 611 eyes of 4 groups of patients using manual interpretation and Hamed's Interpreter as well. RESULTS:There is a statistically significant difference between group 2 and group 1 regarding the inter eye differences in thinnest location (P-value 0.021) and also manifest refraction spherical equivalent (P-value 0.011). The mean of both was significantly high in group 1 (patients with postoperative ectasia) 17.0 ± 7.87 and -5.56 ± 2.16 respectively. There is a statistically significant difference between group 3 and group 1 regarding percent tissue altered (P-value <0.001) and residual stromal thickness (P-value <0.001). The mean of percent tissue altered was significantly higher among patients who had post-laser keratorefractive surgery ectasia group (37.23 ± 5.18) while the mean of residual stromal thickness was significantly low among this group (328.25 ± 41.6). In respect to group 4, the mean of the Inter eye score was 3.38 ± 1.04, and the mean of relative thickness map was -9.2 ± 0.596. The shape of the thickness profile map curve was a quick slope in 61.5% of eyes and normal in 38.5% of eyes in group 4. Some ectasia risk factors were missed during manual interpretation of topography that led to post LVC ectasia. CONCLUSIONS:Developing an artificial intelligence system that can interpret corneal tomography will alleviate the human errors of manual interpretation. INTRODUCTION:Detection of corneas with ectasia possibility comprises a cornerstone of preoperative screening for corneal laser vision correction. There are many investigation options like abnormal anterior and posterior corneal curvature 1 , localized corneal thinning 2 , focally decreased corneal epithelial thickness, 3 and corneal biomechanical instability. 4,5 However, no method gives absolute sensitivity and specificity, we still need to refine the existing diagnostic tools.Inter-eye asymmetry is a famous feature of keratoconus. 6,7 research studies have investigated the inter-eye asymmetry in keratoconus patients, 7-10 because normal corneas are almost symmetric. [11][12][13][14] The investigators demonstrate the significant overlap between healthy corneas and corneas with subclinical keratoconus, 1,15 also they demonstrate the increasing sensitivity by using artificial intelligence in interpreting and Using artificial intelligence in the interpretation of corneal topography for laser vision correction. EJO(MOC) 2021;4:176-188