Retinal blood vessel image segmentation method based on double-channel U-shaped improved Transform network

Abstract

The invention discloses a retinal blood vessel image segmentation method based on a dual-channel U-shaped improved Transform network, and belongs to the field of medical image segmentation. The method comprises the following steps: preprocessing a retinal blood vessel image to obtain an image with uniform hue; according to the method, data are enhanced through a series of methods such as zooming, rotating, cutting and splicing, an original data set is expanded, and then a preprocessed retinal blood vessel image is input into a double-channel U-shaped Transform network for training so as to obtain a model capable of segmenting the retinal blood vessel image. The network is composed of two channels, global and local features of an image are extracted by using up-sampling of a plurality of Transform structures, and the features of the two channels are fused after up-sampling to obtain an image segmentation result. According to the method, on the basis of an original Transform structure, a gating mechanism is added, axial feature extraction is carried out, the calculation complexity is reduced, and finally, a cross entropy loss function and an Adam optimizer are adopted to iterate network model parameters so as to output an accurate retinal blood vessel image segmentation result.

Description

technical field [0001] The invention relates to the field of medical image segmentation, in particular to a retinal blood vessel image segmentation method based on a dual-channel U-shaped improved Transformer network. Background technique [0002] The retina, retinal arteriovenous vessels (referred to as retinal vessels, Retinal Vascular), the optic nerve (Optic Nerve) and the macula (Macula lutea) constitute the main structures of the fundus. Doctors usually use dot mydriatic drugs and fundus imaging equipment to obtain images of the fundus through the dilated pupil. With the development of medical imaging technology, the use of OCT (Optical Coherence Tomography, optical coherence tomography) to perform tomographic scanning of the fundus has become an important means of diagnosing fundus diseases. Retinal blood vessels are the only deep-layered tiny blood vessels in human tissues that can be observed in non-surgical conditions. According to the changes in their diameter, d...

AI Technical Summary

Problems solved by technology

[0006] The main technical problem to be solved by the present invention is to provide a CNN based on dual-channel U Segmentation method of retinal blood vessel image based on improved Transformer network

Images