Neural network magnetic resonance image reconstruction method based on double-domain alternating convolution

Abstract

The invention relates to a neural network magnetic resonance image reconstruction method based on double-domain alternating convolution. The neural network magnetic resonance image reconstruction method is characterized by comprising the following steps: the step 1, obtaining a K space data set; the step 2, generating under-sampling K space data; the step 3, establishing a coding and decoding neural network structure of double-domain alternating convolution; the step 4, training a coding and decoding neural network model of double-domain alternating convolution by using the under-sampled K space data generated in the step 2 and image domain data obtained by performing inverse Fourier transform on the K domain information in the step 1; and the step 5, reconstructing the undersampled magnetic resonance data by using the trained double-domain alternating coding and decoding neural network to obtain a magnetic resonance reconstructed image with relatively high definition. According to the method, accelerated reconstruction of magnetic resonance imaging is realized by using the small kernel convolutional neural network on the K domain, artifacts caused by breaking through the Nyquist sampling limit are eliminated, meanwhile, clear magnetic resonance imaging can be reconstructed, and the reconstruction precision is improved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a reconstruction method of an undersampled magnetic resonance image, in particular to a reconstruction method of a neural network magnetic resonance image based on double-domain alternating convolution. Background technique [0002] At present, magnetic resonance imaging can realize non-invasive visual inspection of human soft tissues, but due to its long scanning time and sensitivity to motion artifacts, patients have severe discomfort during use. The scan time is directly related to the sample size of the measurement data. In order to speed up the scan speed, under-sampled scans can be performed, but limited by the Nyquist sampling theorem, artifacts will appear when the sampling rate is lower than the Nyquist sampling rate . [0003] Lustig M proposed a technique for fast magnetic resonance imaging, Compressed Sensing MRI, using prior information about the underlying s...

AI Technical Summary

Problems solved by technology

By using k-space deep learning and image domain loss function, the fully data-driven k-space interpolation realizes the interpolation of undersampled k-space data, and then realizes the reconstruction of magnetic resonance imaging. Although the algorithm achieves end-to-end reconstruction, it is The undersampled K space data is discontinuous, so the algorithm directly performs convolution calculation in the K domain, which will cause artifacts in the reconstructed image

[0004] In summary, in the known reconstruction algorithms based on K-space magnetic resonance imaging, the effective convolution of the K-domain convolutional neural network has not been well solved, and several existing methods have a large amount of calculation and cannot be truly realized. End-to-end reconstruction, reconstruction results have artifacts and other shortcomings

Images