Computer aided screening of retinopathy of prematurity — A multiscale Gabor filter approach

Non invasive imaging of retina offers easy evaluation and early diagnosis of various systemic diseases like diabetic retinopathy, arteriosclerosis, hypertension and retinopathy of prematurity (ROP). Treatment of ROP, one of the leading causes of childhood blindness is warranted based on the diagnosis of Plus disease which is characterized by high degree of tortuosity and large vessel width for blood vessels in the retina. Here we present a semi-automated computer based method for the screening of Plus disease. This method uses multiscale Gabor filters for the detection of retinal blood vessels and for obtaining their orientations at each pixel. Morphological image processing followed by Gabor filtering helps in obtaining the final vessel map of the images. A 50 image database obtained from KIDROP Bangalore, India were classified as either plus, pre-plus or normal case by quantitatively measuring the amount of tortuosity and width from the vessel map of each images. An overall sensitivity of 1.0 and specificity 0.92 obtained for these images.