Validation of reliability, repeatability and consistency of three-dimensional choroidal vascular index.

ABSTRACT

This study aimed to investigate the reliability, repeatability and consistency of choroidal vascularity index (CVI) measurements provided by an artificial intelligence-based software in swept-source optical coherence tomography (SS-OCT) in normal subject, and to evaluate the influencing factors for 3D-CVI. Repeatability of 3D-CVI by SS-OCT was evaluated based on different scanning modes including Macular Cubes (3 mm × 3 mm, 6 mm × 6 mm, 9 mm × 9 mm) and Optic Nerve Head 6 mm × 6 mm. Intraclass Correlation Coefficient (ICC) was used to estimate the repeatability and reproducibility of five repeated measurement by SS-OCT. Consistency of CVI between SS-OCT and spectral-domain optical coherence tomography (SD-OCT) was measured and compared in a pilot study of ten eyes and agreement between SS-OCT and SD-OCT was evaluated by Bland-Altman analysis and Deming regression. The influencing factors for 3D-CVI including age, gender, axial length and spherical equivalent on CVI was further investigated in a prospective study of 125 eyes of 125 healthy subjects. ICC between different measurements by SS-OCT was 0.934 (95% CI 0.812-0.956) indicating good repeatability. Intraclass correlation coefficient between CVI measure by SS-OCT and SD-OCT was 0.887 (95% CI 0.796-0.938, P value < 0.001). The mean difference between 3D-CVI measured by SS-OCT and SD-OCT 0.133. CVI measured with SS-OCTA showed stronger correlations with axial length and age but not correlated with gender. There is good agreement between CVIs obtained from the built-in software that requires less timing in manual quantification. Studies investigating choroidal vascularity can be standardized by the AI-based CVI analyze software.