Research On Image Synthesis Of Aortic CT Enhanced Scan Based On Deep Learning

In recent years,the research on image synthesis has made considerable progress,but there are still deficiencies in the image synthesis problems in some medical fields,including the synthesis of enhanced scan images of aortic dissection.In the clinical diagnosis of aortic dissection,the diagnostic advantage of CTA images is much stronger than that of plain CT images.However,the enhanced scanning used in CTA images has certain drawbacks.Therefore,the use of deep learning technology to build a network model and synthesize plain CT images into CTA images is of great significance for medical imaging assisted diagnosis.This paper successively proposes a method for synthesizing aortic dissection enhanced scan image based on the channel attention mechanism and a method for synthesizing images based on aortic segmentation.The main contents are as follows:(1)Make an image synthesis data set.The subjects were collected CT plain scan images of the chest cavity and CTA images after enhanced scan,and the images were manually classified into two types of use cases with and without aortic dissection.Aiming at the problem of inconsistencies in the image spatial position caused by organ movement and human breathing,medical image tools are used to register and align the collected CT plain scan images and CTA images.In order to eliminate interference from other organs and make the model more targeted,multiple sets of aortic labels were made on the collected images,and multiple sets of paired CT and CTA images containing only the aortic region were obtained through the segmentation network(2)Combining the characteristics of aortic dissection,a model for synthesizing aortic dissection enhanced scan image based on the channel attention mechanism is constructed.Utilizing the feature of the Fcd Net module that can effectively obtain information,the proposed method can well capture information such as the edge of aortic dissection tear,and improve the quality of image synthesis.Finally,through comparative experiments,it is verified that the proposed method has a better effect on the synthesis of aortic dissection.(3)Constructed a synthesis method based on segmented images of aorta.This paper proposes an enhanced scan image synthesis method for aortic dissection based on cascade generation confrontation for the image after aortic segmentation.The core idea is to build a network of cascaded generators and dual discriminators based on the channel attention mechanism to further improve the quality of the composite image.In order to prove the effectiveness of each module in the model,this article conducted a corresponding ablation experiment.The final experimental results show that the method proposed in this paper has a better synthesis effect than other models.