High-speed acquisition MRI reconstruction method based on residual self-attention image enhancement

Abstract

The invention relates to an MRI accelerated acquisition method, in particular to a high-speed acquisition MRI reconstruction method based on residual self-attention image enhancement. According to themethod, the method comprises the steps of: adopting a generative adversarial network to construct a network structure of the invention, and embedding a residual self-attention-based feature enhancement module into the bottom of a U-NET contraction path; inputting the high-power under-sampling image into a generator; and extracting a high-level feature map through a U-NET contraction path, inputting the high-level feature map into a feature enhancement module to obtain a feature enhancement map, decoding the feature enhancement map through a U-NET expansion path, combining the decoded featureenhancement map with a feature map corresponding to the contraction path, fusing features of a contraction layer of a corresponding level during expansion, supplementing missing boundary information,accurately predicting edge information, and obtaining a reconstructed image. According to the method, more abstract and richer texture detail features of the image can be captured, local information and non-local information are fused to enhance global information amount, effective features are automatically selected by a network in the whole process, adaptive extraction and reconstruction of texture details of a key area can be realized, and a high-speed-acquisition MR image can be well reconstructed.

Description

technical field [0001] The invention relates to an MRI accelerated acquisition method, in particular to a high-speed MRI reconstruction method, in particular to a high-speed MRI reconstruction method based on residual self-attention image enhancement. Background technique [0002] As a repeatable, non-invasive and quantitative tissue measurement method, Magnetic Resonance Imaging (MRI) has become an important means of diagnosis and treatment of major diseases due to its good soft tissue resolution. However, its inherent signal acquisition time is too long, which also brings some difficulties to the application. On the one hand, too long data acquisition time will not only bring discomfort to the patient, but the inevitable movement of the patient's body also increases the possibility of heavy artifacts in the image - when the patient moves (such as heartbeat and gastrointestinal peristalsis, etc.) Often, artifacts in the image often lead to misdiagnosis or missed diagnosis;...

AI Technical Summary

Problems solved by technology

[0006] The present invention provides a high-speed MRI reconstruction method based on residual self-attention image enhancement in order to solve the problem of serious loss of texture details during high-speed MR image reconstruction

Images