Combining ROPtool measurements of vascular tortuosity and width to quantify plus disease in retinopathy of prematurity.
BACKGROUND: ROPtool is a computer program that is used to measure retinal vascular tortuosity and width. Our aim was to determine the best method for combining ROPtool's tortuosity and width values into an overall measure of plus disease. METHODS: ROPtool was used to measure vessel tortuosity and width in 184 quadrants of 46 RetCam images of premature infants. Five methods for combining tortuosity and width were assessed: the plus index and sum of adjusted indices provide tortuosity and width values using different scales; tortuosity-weighted plus increases the contribution of width as tortuosity increases; the vessel area index is used to average vessel area per unit length; and regression analysis establishes a mathematical equation to fit expert diagnosis. For each method, receiver operating characteristic curves were generated with a 3-expert consensus as the reference standard. Optimal cut points for plus disease were selected and then tested on a second set of 184 quadrants of 46 different images. RESULTS: When applied to the second data set, areas under the receiver operating characteristic curves were 0.857 for plus index, 0.918 for sum of adjusted indices, 0.941 for tortuosity-weighted plus, 0.868 for vessel area index, and 0.934 for regression analysis methods. The highest sensitivities and specificities were for tortuosity-weighted plus (85% for both) and regression analysis methods (85% and 86%, respectively). CONCLUSIONS: When combining ROPtool's measures of width and tortuosity, sum of adjusted indices, tortuosity-weighted plus, and regression analysis methods had high overall accuracy. ROPtool can now generate a clinically meaningful quantitative description of vascular abnormality to aid in plus disease diagnosis.
Cabrera, MT; Freedman, SF; Kiely, AE; Chiang, MF; Wallace, DK
Volume / Issue
Start / End Page
Electronic International Standard Serial Number (EISSN)
Digital Object Identifier (DOI)