Edge enhancement for retinal vasculature caliber evaluation in prediction of cardiovascular disease

Abstract

At the moment, the gold standard in detection of coronary disease is coronary angiography. Recent studies show that retinal arteriolar and venular caliber change is associated with many cardiovascular risk factors. Thus, investigating blood column diameters inside retinal vessels during cardiac cycle (ECG gating) could help cardiologists in the prediction of cardiovascular disease. In this paper, we demonstrated an improved image preprocessing step for computer aided diagnosis algorithms that could be useful in measuring the width of the blood column. The algorithm consists of three main steps. First, higher resolution images at 21 megapixels were captured along with an inbuilt 570nm red-free filter to maximize the blood signal. Second, the image was decomposed using a Gabor filter at four different directions (0, 45, 90 and 135 degree) and maximum responses were obtained over the four directions. Finally, the size of vessels is calculated by performing the sliding linear regression filter (SLRF) on vessels slices. The results suggest the use of edge enhancement with SLRF decrease the variability by 2-3 times and might lead to more deterministic vessel widths with less variation.