Blood vessel segmentation method, device and equipment for retinal image and storage medium

Abstract

The invention discloses a blood vessel segmentation method for a retina image, and relates to the field of medical image processing.The method comprises the steps of obtaining a to-be-processed retinaimage; preprocessing the retina image to be processed to obtain a blood vessel enhanced image; based on a pre-trained blood vessel segmentation model, adopting a multi-channel cross attention mechanism comprising a channel weighting processing process and a space weighting processing process to process a to-be-processed retina image and an initial blood vessel enhanced image, and obtaining outputof the blood vessel segmentation model; and obtaining a blood vessel segmentation result according to the output of the blood vessel segmentation model. The invention further discloses a corresponding device, equipment and a storage medium, the blood vessel segmentation speed of the retina image is effectively increased, the detection effect on tiny blood vessels and blood vessels affected by focuses is improved, the design is simple and easy to implement, the model parameter quantity can be adjusted according to the task difficulty, high accuracy and specificity are achieved, and the model complexity is lower.

Description

technical field [0001] The present invention relates to the field of medical image processing, in particular to a blood vessel segmentation method, device, equipment and storage medium for retinal images. Background technique [0002] Cardiovascular disease is the non-infectious disease with the highest prevalence rate in my country, and has become the "first killer" that threatens the health of our people. It is commonly manifested as elevated blood pressure, heart attack or stroke. Cardiovascular disease includes high Blood pressure, coronary heart disease, cerebrovascular disease, peripheral vascular disease, etc. [0003] Cardiovascular disease and many systemic diseases are manifested in retinal blood vessels, and affect retinal fundus blood vessels in different ways. Retinal vascular changes are much earlier than symptomatic clinical lesions. For example, most cardiovascular diseases can observe pathological changes in retinal vascular morphology at an early stage, whi...

AI Technical Summary

Problems solved by technology

[0008] (1) There are tiny blood vessels in the retinal image, and their texture features are close to the retinal background, which is difficult to be identified. The existing technology is easy to lose blood vessel bifurcation points and small blood vessels

[0009] (2) The prior art generally assumes that the cross-section of blood vessels is circular, which is correct for normal blood vessels, but this assumption is not true for diseased blood vessels; and once there is a retinal image affected by pathological conditions, Such as retinal hemorrhage points, the state-of-the-art vessel segmentation is prone to results with high false positives

[0010] (3) The existing technology needs to manually design features with strong classification ability, which leads to great difficulties in designing features, and the model design based on deep learning is complex and has a large number of parameters, large amount of calculation, and expensive calculation

Images