Repeatability and Reproducibility of Automated Quantitative Conjunctival Hyperemia Analysis Software (Preprint)

BACKGROUND The degree of conjunctival hyperemia reflects the severity of various ocular disease. Most of clinical grading systems rely on individual physicians and it is difficult to compare between clinical examinations. We developed automated grading software to objectively assess conjunctival hyperemia. OBJECTIVE To assess the repeatability and reproducibility of newly developed automated quantitative conjunctival hyperemia analysis software. METHODS Eighty images of twenty glaucoma patients who were using at least one anti-glaucoma medication and complained of conjunctival hyperemia were collected. All patients’ severity of conjunctival hyperemia was evaluated by newly developed automated conjunctival hyperemia grading software into 4 levels: blood vessel, strong red area, weak red area and non-red area. Automatic registration was applied for improving the software repeatability and reproducibility. RESULTS Spearman correlation coefficient values(rs) values of 0.884, 0.982 and 0.932 for blood vessel area, strong red area and weak red area of two measurements of the same image, respectively. Spearman correlation coefficient (rs) values of 0.988, 0.992 and 0.995 for blood vessel area, strong red area and weak red area of two images of the same eye, respectively. The clinical grading had the best correlation with the summation of blood vessel and strong red area, following with strong red area. CONCLUSIONS The proposed conjunctival hyperemia analysis software provides an automatically quantitative objective assessment of conjunctival hyperemia with excellent repeatability and reproducibility including excellent correlation with clinical grading by ophthalmologists. This algorithm including the automatic region of interest (ROI) registration between two images system give precise results in comparing the conjunctival hyperemic change in clinical practice.