Honeycomb lung focus segmentation method based on SAA-Unet network

Abstract

The invention discloses a honeycomb lung focus segmentation method based on an SAA-Unet network, and belongs to the technical field of image processing. The technical problem to be solved is to provide the improvement of the honeycomb lung focus segmentation method based on the SAA-Unet network. According to the technical scheme for solving the technical problem, U-Net is used as a basic network, feature information in a honeycomb lung lesion part is deeply excavated, the generalization ability of a main task is improved, and honeycomb lung lesion features are extracted more accurately; meanwhile, in order to improve the model segmentation accuracy, the problem of feature loss of the image in the convolution and deconvolution processes is solved by using a division attention module, and finally, the weight value of an important lesion region is improved by fusing high-level and low-level feature information by using an attention mechanism, so that the segmentation accuracy of the network model on the honeycomb lung lesion region is realized; the method is applied to honeycomb lung focus segmentation.

Description

technical field [0001] The invention discloses a method for segmenting cellular lung lesions based on a SAA-Unet network, belonging to the technical field of image processing. Background technique [0002] Cellular lung disease is a common disease that seriously threatens human health. It has the characteristics of long course of disease, high fatality rate, poor prognosis, and low survival rate, which leads to low survival rate and cure rate of patients with honeycomb lung disease. With the continuous development of computer computing power, computer-aided medical treatment has been accepted by medical workers for its accuracy and convenience, but how to accurately segment honeycomb lung from CT slices has become an important and difficult point. At the same time, due to the characteristics of irregular shape, uneven gray scale, complex texture, and large deformation, the existing deep learning-based segmentation methods, especially the convolutional neural network, are dif...

AI Technical Summary

Problems solved by technology

At the same time, due to the characteristics of irregular shape, uneven gray scale, complex texture, and large deformation, the existing deep learning-based segmentation methods, especially the convolutional neural network, are difficult to detect in the image of honeycomb lung. Segmentation is less accurate

[0003] At present, the evaluation of honeycomb lung mainly relies on the human visual judgment of radiologists. This subjective visual evaluation method relies heavily on the doctor's clinical experience and cognitive ability of this symptom, and can only qualitatively characterize the area of ​​​​the honeycomb lung lesion. Analysis, unable to achieve accurate quantitative analysis of the honeycomb lung lesion area

Images