Abstract
Purpose: :
Microaneurysms, one of diabetic retinopathy's most telltale lesions, range from 25 to 125 microns. An early sign of diabetic retinopathy, microaneurysms have also been found in patients with pre–diabetes and are considered a surrogate measure of diabetic retinopathy severity. Digital retinal cameras with a range of resolutions for evaluating diabetic retinopathy are available. However, there has been no comprehensive study comparing the impact of digital resolution with microaneurysm visibility.
Methods: :
We compared the performance of three image resolutions similar to commercially available non–mydriatic cameras: 3008x2000 pixels, 1088x723 and 800x531. For a sufficiently large database of images to test resolution variables, we created computer–generated microaneurysms and added them to lesion free images. Most microaneurysms were 86 microns in diameter but ranged from 68 to 113 microns. Images were down–sampled and up–sampled to simulate photographs taken with different resolutions. Three retinal ophthalmologists reviewed 100 light pigment and 100 dark pigment images of each resolution.
Results: :
There was no statistically significant difference in ROC Az performance as a function of resolution (F = 0.324, p = 0.7312) or fundus pigment (t = 0.999, p = 0.3915) among observers.
Conclusions: :
Our results suggest digital resolution as low as 800x600 can be used to appreciate large microaneurysms. The acceptance of diabetic retinopathy telemedicine evaluation is partially related to sensitivity and cost. Lower resolution retinal camera systems are associated with lower cost, potentially removing one barrier. A study of resolution impact on smaller microaneursyms is ongoing. The impact on other types of diabetic lesion should also be studied.
Keywords: diabetic retinopathy • imaging/image analysis: clinical • retina