Retinal arteriovenous classification method and device based on topological structure estimation

Abstract

The invention discloses a retinal arteriovenous classification method based on topological structure estimation. The method comprises the steps of performing pixel-level segmentation on a retinal vessel image through a U-net deep learning network to obtain vessel segmentation result information; separately carrying out topology estimation on the blood vessel segmentation result information throughan advantage set theory, establishing a complete topology tree set, and obtaining topology information; carrying out arteriovenous classification on the myopia disc area blood vessels through the advantage set theory, and obtaining a myopia disc area blood vessel arteriovenous classification result; and processing the topological information and the arteriovenous classification result of the myopia disc region blood vessels according to a label propagation algorithm to obtain global arteriovenous classification. The invention further discloses a retinal arteriovenous classification device based on topological structure estimation. According to the invention, the topology of retinal vessels is continuously added into arteriovenous classification work, and the arteriovenous problem of vessels in a far optic disc area in fundus color photograph can be effectively solved through topology information guidance.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method and device for classifying retinal arteries and veins based on topology estimation. Background technique [0002] Vascular diseases are one of the most common public health problems in the world. The most common related diseases include diabetes, arteriosclerosis, cardiovascular disease, and hypertension. Sudden manifestations pose a serious threat to the life safety of patients. Therefore, early screening and diagnosis are of great significance to the life, health and safety of the public. However, due to the complexity and invasiveness of the current diagnostic methods for vascular diseases, it is impossible to achieve a large-scale general survey. Retinal blood vessels, as the only vascular tissue available in the human body under non-invasive conditions, have been proved in clinical studies that their morphological changes are closely related to blood vess...

AI Technical Summary

Problems solved by technology

At present, most of the common retinal vessel classification methods are used to distinguish arteries and veins through pixel-level classification in fundus color photos. However, due to the influence of light and structure itself, the difference between retinal vessels in the far optic disc area is relatively weak. Existing algorithms are difficult to achieve accurate classification

Images