Detection of neovascularization near the optic disk due to diabetic retinopathy

摘要

We propose a technique for detection of neovascularization near the optic disk due to diabetic retinopathy. Images of the retinal fundus are analyzed using a measure of angular spread of the Fourier power spectrum of the gradient magnitude of the original images using the horizontal and vertical Prewitt operators. The entropy of the angular spread of the Fourier power spectrum and spatial variance are adopted to distinguish normal optic disks from those affected by neovascularization. The two-sided Kolmogorov-Smirnov nonparametric test is used to evaluate the significance of the difference of entropy between normal and abnormal optic disks. Based on the computed measures, we employ a linear classifier to discriminate normal from abnormal optic disks. The proposed method was able to classify a small set of five normal and five neovascularization cases with 100% accuracy.