Abstract
Purpose :
Optic disc blurring is an indicator of papilledema and optic neuritis. There have not been many substantial explorations of quantifying optic disc margin clarity. We analyzed fundoscopic images exhibiting various degrees of blurring to measure the clarity of optic disc margins.
Methods :
An established, publicly available dataset of fundoscopic images of normal optic discs, pseudopapilledema, and papilledema was used for analysis and 100 representative images were selected from each of the three categories for analysis. Analysis included calculation of the rolling rates of change in average radial intensities as well as the rolling magnitudes of difference in average radial intensities.
Results :
Fundoscopic images of normal optic discs showed greater magnitudes of rates of change in average radial intensities at optic disc margins than images of papilledema (p < 10-29). This difference also applied between normal optic discs and pseudopapilledema as well as pseudopapilledema and papilledema, respectively (p < 10-4 and 10-17, respectively). Decision tree machine learning modeling was performed on the results of analysis, which showed that fundoscopic images with a rate of change in radial intensities greater than -0.005 were 100% papilledema while images with a rate of change less than -0.006 and maximum difference in radial intensities of greater than 0.013 were 97.3% normal optic discs or pseudopapilledema. Overall, the rate of change in radial intensities of fundoscopic images was 98.6% sensitive for papilledema.
Conclusions :
This study found that the rate of change in average radial intensities of fundoscopic images is a sensitive metric for determining the clarity of optic disc margins and differentiating papilledema from pseudopapilledema and normal optic discs.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.