New coronal pneumonia CT image infected area segmentation method and system based on improved CE-Net

Abstract

The invention provides a new coronal pneumonia CT image infected area segmentation method and system based on improved CE-Net, and the method comprises the steps: firstly, adding an attention mechanism SE module in a coding stage, introducing global context information, enhancing a receptive field in a feature extraction stage, and increasing the weight of a target related feature channel, thereby improving the segmentation capability of a small target; and secondly, a feature aggregation module is introduced, a bilinear interpolation method is adopted, image features of different levels are fused, expression with higher discrimination capability is obtained, and the segmentation precision of the network is further improved. According to the method, the features of the new coronal pneumonia infection area in the CT image can be better captured on the COVID-19-CT-Scs data set, the good segmentation effect is achieved, and compared with an original CE-Net network and other segmentation algorithms, the overall segmentation effect is remarkably improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a method and system for segmenting infection areas in CT images of new coronary pneumonia based on improved CE-Net. Background technique [0002] Although existing deep learning algorithms have achieved good results in image processing of COVID-19, there is still little work related to segmenting COVID-19 infected areas in images, because there are several difficulties in segmenting infected areas from 2D CT images : 1) There are huge differences in the location, size, and shape of infection in different 2D CT images, which often lead to false negative detections. 2) The contrast between the infected area and the normal area is low. 3) The boundary of the infected area is usually blurred, making it difficult to obtain very accurate labels. Contents of the invention [0003] In order to make up for the gaps and deficiencies of the existing technology, the ...

AI Technical Summary

Problems solved by technology

[0002] Although existing deep learning algorithms have achieved good results in image processing of COVID-19, there is still little work related to segmenting COVID-19 infected areas in images, because there are several difficulties in segmenting infected areas from 2D CT images : 1) There are huge differences in the location, size, and shape of infection in different 2D CT images, which often lead to false negative detections

2) Low contrast between infected area and normal area

3) The boundary of the infected area is usually blurred, making it difficult to obtain very accurate labels

Images