Fundus image blood vessel segmentation method

Abstract

The invention provides a fundus image blood vessel segmentation method which comprises the following steps of S1, carrying out augmented operation on a fundus image data set, namely cutting an original image according to a certain rule to obtain an augmented data set, and dividing the data set into a training set and a test set; S2, carrying out blood vessel segmentation on each fundus image sample in the augmented training data set by using a segmentation network to obtain a black-white probability graph, and carrying out image processing to obtain a probability graph of segmented images in the training data set; S3, using a discrimination network to distinguish whether the probability graph is obtained by a segmentation network or a segmentation image in a data set, and obtaining a confidence graph with a numerical value range of 0-1 to describe the probability that the input image is judged to be true or false; and S4, inputting the color fundus image in the test set into the trained network, and splicing the obtained result images to obtain a complete segmentation image of the fundus blood vessel, wherein the step S2 and the step S3 perform fundus image blood vessel segmentation based on the generative adversarial network, and before the step S4, the segmented network in the step S2 and the judgment network in the step S3 are sequentially subjected to iterative training according to the designed loss function, and the performances of the two networks are respectively improved in a game mode. The method is based on the superiority of the generative adversarial idea in data generation, does not depend on selection of initial features, reduces the complex calculation steps, effectively improves the segmentation precision of fundus image blood vessels, and especially has the excellent performance on tiny blood vessels.

Description

technical field [0001] The invention relates to a non-medical purpose fundus image blood vessel segmentation method. Background technique [0002] As an important part of the eyeball structure, the retina has a very important connection with various systems of the whole body. The retina consists of the optic disc, macula, and retinal blood vessels. As the only deep microvascular system in the human body that can be directly observed by non-invasive means, retinal blood vessels can provide ophthalmologists with important information about changes in their structural state, such as arteriosclerosis, hypertension, diabetes, cardiovascular disease and old age. Macular degeneration, etc. will affect the blood vessels of the fundus, causing changes in the width and curvature of the arteries and veins. This information provides key material for eye-related scientific research. [0003] At present, fundus images are commonly used clinically to observe the status of retinal blood ...

AI Technical Summary

Problems solved by technology

However, the implementation steps of the above work are relatively complicated and rely on a complex parameter estimation process. The

Images