A MRI Reconstruction Method Based on Deep Learning and Convex Set Projection

Abstract

The invention discloses a magnetic resonance reconstruction method based on deep learning and convex set projection, and relates to the field of magnetic resonance technology, including: S1: Constructing based on overlapping structures and shared data of multiple convolutional neural network modules and multiple convex set projection layers network, the shared data includes collected K-space data and coil sensitivity information, and the highlighted projection layer is obtained based on the shared data; S2: After the network is constructed, all network parameters are trained through the back propagation process, and the network parameters are Perform verification; S3: Determine the structure and operation characteristics of the network according to the verified network parameters, input known test set data, perform forward propagation of the network, obtain unknown mapping data, and complete magnetic resonance reconstruction. The invention solves the problem that the current magnetic resonance reconstruction technology based on deep learning can only support single-channel magnetic resonance data, but cannot process multi-channel magnetic resonance data.

Description

technical field [0001] The invention relates to the technical field of magnetic resonance, in particular to a magnetic resonance reconstruction method based on deep learning and convex set projection. Background technique [0002] Magnetic resonance imaging is a technique that uses the nuclear magnetic resonance phenomenon of hydrogen protons for imaging. Nuclei containing a single number of protons in the human body, such as the ubiquitous hydrogen nucleus, have spin motions for the protons. The spin motion of charged atomic nuclei is physically similar to individual small magnets, and the directionality distribution of these small magnets is random in the absence of external conditions. When the human body is placed in an external magnetic field, these small magnets will rearrange according to the magnetic force lines of the external magnetic field, specifically in two directions parallel to or antiparallel to the external magnetic field magnetic force lines, and the abov...

AI Technical Summary

Problems solved by technology

However, the above approximate transformation is difficult to describe the complex micro-microbial structure, which leads to blurred or mosaic effect in the reconstructed image

Third, compressive sensing defines MRI reconstruction as a nonlinear optimization problem, thus resulting in long reconstruction times

However, this technology currently only supports single-channel MRI data and cannot handle multi-channel MRI data

Images