Image synthesis method and device for fusing ROI, electronic equipment and medium

Abstract

The invention relates to an image synthesis method and device for fusing an ROI, electronic equipment and a medium. The method comprises the following steps: acquiring diseased and disease-free original images with corresponding category information; using a weak supervision positioning network to obtain a region of interest of the diseased original image to obtain a diseased mask image; obtaining a similar diseased image corresponding to the disease-free original image through similarity calculation, and obtaining a disease-free mask image; shielding the diseased original image and the disease-free original image by using the diseased mask image and the disease-free mask image respectively to obtain images shielded by masks; and designing a condition-based generation antagonism network model, and training the network model by using diseased and disease-free mask images, the images shielded by the masks and corresponding category information, thereby obtaining a synthetic medical image. According to the method, the region of interest in the medical image is acquired, and the antagonism network model is generated based on the conditions, so that the medical image of a specified type can be synthesized efficiently, and the image quality is improved.

Description

technical field [0001] The invention relates to the field of image synthesis in computer vision, in particular to an image synthesis method, device, electronic equipment and medium for merging ROI regions. Background technique [0002] In recent years, disease diagnosis methods based on deep learning have been widely used in the medical field. These deep learning network-based models can process medical images in large quantities, reduce the time for doctors to read images, and assist doctors in diagnosis. However, deep learning models generally rely on a large amount of training data. The available, high-quality, and labeled medical image data are obviously insufficient, and medical images are uneven. Medical images with diseases are very scarce, and the number of images without diseases is serious. The mismatch restricts the performance of medical diagnostic models. At present, Generative Adversarial Networks (GAN network for short) is used to solve the problem of insuff...

AI Technical Summary

Problems solved by technology

However, most studies use the entire image as input and consider global information. This method is more suitable for diseases with large lesion areas. If the lesion area is concentrated in a local area, it will lead to a waste of computer resources. Moreover, most methods synthesize images of a category label in the model, resulting in poor attention to the synthesized images, insufficient migration and generation capabilities of key parts, and insufficient quality of generated images, which limit the ability of image synthesis and medical diagnosis. effective assistance

Images