Agreement of corneal epithelial profiles produced by automated segmentation of SD-OCT images having different optical resolutions

OBJECTIVE: Using a custom-developed segmentation algorithm, agreement of corneal epithelial thickness profile measurements between two spectral domain optical coherence tomography (SD-OCT) systems was assessed. METHODS: Eighteen left eyes (18 subjects; mean±standard deviation; age, 23.2±0.9 years) were imaged twice on nonconsecutive days by a custom-built ultra-high resolution OCT (UHR-OCT) system and a commercial RTVue OCT system. A segmentation algorithm based on axial gradient information and a shortest path search was developed to measure corneal epithelial thickness profiles from the SD-OCT images. RESULTS: There was good correlation between the automated and manual segmentation positions of the epithelium. The epithelial thickness differences between automated and manual segmentations by the UHR-OCT and RTVue OCT systems were 0.4±0.3 μm and 1.1±0.5 μm, respectively. The intraclass correlation coefficients (ICCs) and coefficients of repeatability for repeated UHR-OCT measurements of epithelial thickness were 0.90±0.05 μm and 1.44±0.01 μm, respectively. For RTVue OCT, the ICCs and coefficients of repeatability were 0.88±0.04 μm and 2.16±0.01 μm, respectively. The epithelial thickness measured by RTVue OCT was 1.1±0.2 μm greater than that by UHR-OCT (P<0.05, paired t test), and ICC for the agreement between the 2 systems was 0.85±0.06. The average 95% limit of agreement was -1.67 to 3.79 μm. CONCLUSIONS: For images of different qualities from both SD-OCT systems, the custom-developed automated segmentation of the corneal epithelium had good repeatability for thickness profile measurements. The accurate and precise algorithm also demonstrated good agreement of epithelial thickness profile measurements between the two SD-OCT systems.