Fundus image blood vessel segmentation method based on coupled neural network and line connector

Abstract

The invention discloses a fundus image blood vessel segmentation method based on a coupled neural network and a line connector and belongs to the technical field of medical image and digital image processing. The method comprises the following steps that a simplified pulse coupling neural network model is proposed, and a basic structure of a blood vessel in an image is obtained by utilizing the similarity of adjacent neurons; a new denoising method is provided, most noise points are removed through pixel connectivity, and meanwhile complete blood vessel edges are reserved, a problem of blood vessel breakage in the segmentation process is solved through the line connector so that a complete blood vessel structure is presented, and blood vessel recognition accuracy is improved. A test is carried out on two public retina databases of DRIVE and STARE, and a result shows that compared with an existing method, the method is advantaged in that excellent scores are obtained on indexes such asaverage accuracy and sensitivity, and the good response time is achieved.

Description

technical field [0001] The invention relates to the field of medical image processing, in particular to a fundus image blood vessel segmentation method based on a coupled neural network and a line connector. Background technique [0002] Changes in the structural characteristics of fundus images can reflect the existence of certain physical diseases to a certain extent. Blood vessels are the most important structure in fundus images. The diameter, color and curvature of blood vessels are closely related to the existence of diseases. Vascular and coronary artery disease and retinal lesions in infants often cause changes in the shape of retinal blood vessels. Therefore, the analysis of fundus images is crucial for the diagnosis of ophthalmic problems and other diseases such as diabetes and hypertension. Due to the time-consuming and labor-intensive manual analysis, an excellent medical image-aided diagnosis system is now needed to pre-process the image, such as automatically ...

AI Technical Summary

Problems solved by technology

[0004] Due to the complex structure of the fundus retinal image, there are still problems such as uneven illumination, weak contrast, and noise interference, which lead to the inability of general image segmentation algorithms to segment retinal blood vessels. Finding an effective blood vessel segmentation method is a common concern of scholars today.

Images