Kago-Eye2 software for semi-automated segmentation of subfoveal choroid of optical coherence tomographic images.

Abstract

To determine the capabilities of the "Kago-Eye2" software to semi-automatically segment the choroid in the optical coherence tomographic (OCT) images. A cross-sectional, prospective study of 44 healthy volunteers. The Kago-eye2 software was developed to detect the border between Choriocapillaris and Sattler's layer (C-S) and between Sattler's layer and Haller's layer (S-H). The intra- and inter-grader agreements were determined for the segmentations made with semi-automated and manual analysis using the Kago-Eye2 software. The inter-method agreements were determined for two independent graders. Forty-four right eyes of 44 heathy volunteers (22 men) with a mean age of 35.0 ± 8.8 years were studied. The intra-grader agreement of the C-S border was 0.97 for grader 1 and 0.892 for grader 2 for the manual segmentation, and 0.908 for grader 1 and 0.842 for grader 2 for the Kago-Eye2 segmentation. For the S-H border, the intra-grader agreement was 0.96 for grader 1 and 0.981 for grader 2 for manual segmentation and 0.855 for grader 1 and 0.839 for grader 2 with the Kago-Eye2. For the C-S and S-H border, the inter-grader agreement was 0.548 and 0.902 for manual segmentation and 0.947 and 0.833 for the Kago-Eye2. The inter-method agreement was 0.565 for the C-S border and 0.759 for the S-H border. The Kago-Eye2 software can segment the layers of the subfoveal choroid with good reproducibility and repeatability. We conclude that the Kago-Eye2 software can be used for semi-automatic segmentation of the choroidal layers.