Özet:

Abstract Retinal damage caused due to complications of diabetes is known as Diabetic Retinopathy (DR). In this case, the vision is obscured due to the damage of retinal tinny blood vessels of the retina. These tinny blood vessels may cause leakage which affect the vision and can lead to complete blindness. Identification of these new retinal vessels and their structure is essential for analysis of DR. Automatic blood vessels segmentation plays a significant role to assist subsequent automatic methodologies that aid to such analysis. In literature most of the people have used computationally hungry a strong preprocessing steps followed by a simple thresholding and post processing, But in our proposed technique we utilize an arrangement of  light pre-processing which consists of Contrast Limited Adaptive Histogram Equalization (CLAHE) for contrast enhancement, a difference image of green channel from its Gaussian blur filtered image to remove local noise or geometrical object, Modified Iterative Self Organizing Data Analysis Technique (MISODATA) for segmentation of vessel and non-vessel pixels based on global and local thresholding, and a strong  post processing using region properties (area, eccentricity) to eliminate the unwanted region/segment, non-vessel pixels and noise that never been used to reject misclassified foreground pixels. The strategy is tested on the publically accessible DRIVE (Digital Retinal Images for Vessel Extraction) and STARE (STructured Analysis of the REtina) databases. The performance of proposed technique is assessed comprehensively and the acquired accuracy, robustness, low complexity and high efficiency and very less computational time that make the method an efficient tool for automatic retinal image analysis. Proposed technique perform well as compared to the existing strategies on the online available databases in term of accuracy, sensitivity, specificity, false positive rate, true positive rate and area under receiver operating characteristic (ROC) curve.

Anahtar Kelimeler: