Purpose
Our aim was to determine if computer-assisted quantitative analysis of retinal vascular dilation and tortuosity using ROPtool software can predict the need for laser treatment in retinopathy of prematurity (ROP).
Methods
Video indirect ophthalmoscopy recordings were done of the eyes of premature infants undergoing routine, clinically indicated ROP exams (1/1/2005-8/31/2012). Retrospectively, a high-quality still image of the posterior pole of one eye was selected from each exam for each infant. Infants were included if they had at least one exam prior to 35 weeks postmenstrual age and at least 2 exams with high-quality images prior to vessel maturation or prior to the exam determining need for laser treatment. All eligible infants who had laser were included, and twice as many infants without laser were randomly selected as controls. ROPtool was used to calculate retinal vessel dilation, tortuosity, and Sum of Adjusted Indices (SAI, combining dilation and tortuosity) for each image. For each of the vessel characteristics (dilation, tortuosity, SAI), the following were calculated: maximum value from any one vessel across all exams (Max), largest increase per week in Max (Change in Max), highest mean value when all vessels were averaged from any one exam (Highest_Mean), and largest increase per week in Highest_Mean (Change in Highest_Mean). These parameters were compared between infants who received laser and those who did not. P values were Bonferroni corrected, with a critical p value of 0.002.
Results
Medians for maximum tortuosity indexes were 9.923 and 6.477 for laser (n=27) and no laser (n=55), respectively (p<0.001). Medians for highest mean tortuosity indexes were 5.145 and 3.617 for laser and no laser, respectively (p<0.001). Parameters involving dilation and SAI were not significantly different between eyes that received laser and those that did not (Table). In multivariable analysis, the variable with the most significant association with need for laser treatment was highest mean tortuosity (p<0.001).
Conclusions
ROPtool analysis of tortuosity from indirect ophthalmoscopy still images is able to predict need for laser treatment in retinopathy of prematurity. Highest mean tortuosity of all vessels averaged in an image is the best predictor of the need for laser treatment.
Keywords: 706 retinopathy of prematurity •
550 imaging/image analysis: clinical •
688 retina