The Research Of Vessel Segmentation Techniques In Retinal Images

Retinal vasculature is an important part of systemic microcirculation system. The change of the structure characteristics of retinal image is closely related to diabetes, hypertension, coronary artery disease cerebral vascular sclerosis and other cardiovascular diseases. Diabetes is one of the global non-communicable diseases. Diabetic retinopathy is one of the most common retinal vascular diseases caused by diabetes, which leads to blindness. By extracting the retinal vessels and analyzing their characteristics and relevant parameters, for example, blood vessel diameter and tortuosity etc, it is largely possible to provide a way on the prediction of diabetic retinopathy to the greatest exent, so as to implement preventive intervention and take the medicine treatment.The structure characteristics of retinal image are complicated and individually different. The automatic detection of retinal blood vessels is commonly affected by some external conditions and pathological changes. Therefore, it is still an important task to improve the accuracy of the extraction of blood vessels.This paper presents a method based on multi-scale line detection and local entropy threshold for automatic detection and extraction of blood vessels in retinal images. The novelty is that it not only can extract blood vessels with center reflection and two adjacent blood vessels, but also can extract more small capillaries. Firstly, extracting the rich green component of the retinal fundus images and image preprocessing, including shade correction, noise reduction and CLAHE, etc. Secondly, based on the morphological structure characteristics of blood vessels, the proposed method is based on the fact that by changing the length of a basic line detector. The line detectors at varying scales are achieved; Finally, based on the gray gradient co-occurrence matrix, entropy threshold is applied for well keeping the spatial structure of vascular tree segments. The performance of the proposed method was evaluated on three publicly available datasets: DIARETDB0 DRIVE and HRF. The experimental result demonstrated the high accuracy of the blood vessel segmentation by the proposed method. Other advantages of the proposed method include its efficiency with fast processing time and the ability extracting of small blood vessels.Tortuosity is one of the primary symptoms used for tracking retinal disease. Retinal vasculature in the normal image is straight or with a slight bend. Diabetes can lead to blood vessels bent, and the tortuosity becomes more serious with the deepening of the illness. This paper presents a method based on the ratio of arc length and chord length to calculate the tortuosity of blood vessels. The method is simple and efficient.