Medical image cross-modal synthesis system and method based on multi-source confrontation strategy

Abstract

The invention discloses a medical image cross-modal synthesis system and a medical image cross-modal synthesis method based on a multi-source adversarial strategy, which can accurately describe key features of a target domain modal image and improve the quality of synthesizing the target domain modal image, thereby obtaining the target domain modal image better meeting actual requirements. The system comprises a multi-source input fusion network unit which is provided with a first convolution layer and a feature cascade layer; the generation neural network unit comprises a coding compression sub-network and a decoding expansion sub-network; the identification neural network unit is provided with a fourth convolutional layer and a second down-sampling layer, extracts a first advanced feature map of the first target domain modal image and a second advanced feature map of the real modal image, and identifies the first advanced feature map and the second advanced feature map; the difference between the actual recognition value and the expected recognition value is obtained and fed back to the neural network generation unit; and outputting the second target domain modal image when the difference is smaller than the threshold value.

Description

technical field [0001] The present invention relates to the field of big data / AI technology, in particular to a medical image cross-modal synthesis system and method based on a multi-source confrontation strategy. Background technique [0002] At present, medical imaging includes many imaging modalities, such as X-ray, computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET) and so on. Different imaging modalities can provide multi-level information for clinical diagnosis from different angles, thereby significantly reducing the rate of misdiagnosis and missed diagnosis. However, in the actual clinical diagnosis process, due to factors such as high cost and radiation hazards, the image data of some modalities is relatively scarce. In related technologies, cross-modal synthesis of medical images is an effective solution to provide missing modal data, but the current single-source domain modal images can provide limited physiological char...

AI Technical Summary

Problems solved by technology

However, in the actual clinical diagnosis process, due to factors such as high cost and radiation hazards, the image data of some modalities is relatively scarce.

In related technologies, cross-modal synthesis of medical images is an effective solution to provide missing modal data, but the current single-source domain modal images can provide limited physiological characteristics information of subjects, and the synthesized target domain modal images are often There is a certain gap with the real target domain modal image, and it is difficult to meet the use requirements in actual clinical diagnosis and intervention

Images