Bowman's topography for improved detection of early ectasia

The aim of this study was to evaluate whether OCT topography of the Bowman's layer and artificial intelligence (AI) can result in better diagnosis of forme fruste (FFKC) and clinical keratoconus (KC). Normal (n = 221), FFKC (n = 72) and KC (n = 116) corneas were included. Some of the FFKC and KC patients had the fellow eye (VAE-NT) with normal topography (n = 30). The Scheimpflug and OCT scans of the cornea were analyzed. The curvature and surface aberrations (ray tracing) of the anterior corneal surface [air-epithelium (A-E) interface in OCT] and epithelium-Bowman's layer (E-B) interface (in OCT only) were calculated. Four random forest models were constructed: (1) Scheimpflug only; (2) OCT A-E only; (3) OCT E-B only; (4) OCT A-E and E-B combined. For normal eyes, both Scheimpflug and OCT (A-E and E-B combined) performed equally in identifying these eyes (P = .23). However, OCT A-E and E-B showed that most VAE-NT eyes were topographically similar to normal eyes and did not warrant a separate classification based on topography alone. For identifying FFKC eyes, OCT A-E and E-B combined performed significantly better than Scheimpflug (P = .006). For KC eyes, both Scheimpflug and OCT performed equally (P = 1.0). Thus, OCT Topography of Bowman's layer significantly improved the detection of FFKC eyes.