No-reference scanning correction method for echo plane imaging eddy current artifacts

Abstract

A no-reference scanning correction method for echo plane imaging eddy current artifacts comprises the following steps: firstly, training parameters of a neural network: randomly generating simulationsamples in batches by utilizing simulation software, and then introducing pre-compiled EPI sequences to sample the simulation samples to obtain K space data corresponding to each sample; and accordingto the phase characteristics of the eddy current effect, carrying out phase modification on the collected signal, and carrying out Fourier transform to obtain a simulated EPI image influenced by theeddy current effect, and storing the modified phase information as a label, and training the neural network model by taking the images as a training data set, and storing network parameters when the training error is reduced to be lower than a set threshold value, and after training of the neural network model is completed, inputting the truly acquired EPI image into the neural network model for reconstruction, thereby obtaining the EPI image without the eddy current artifact. According to the method, the EPI image eddy current artifacts are removed under the condition of no reference scanning, and assistance can be provided for clinical application of EPI.

Description

technical field [0001] The present application relates to an MRI image reconstruction method, in particular to a neural network-based non-reference scan correction method for eddy current artifacts in echo planar imaging. Background technique [0002] Magnetic Resonance Imaging (MRI) has no ionizing radiation and has multiple imaging parameters. It plays an important role in clinical diagnosis and scientific research. For example, published literature: E.Yacoub, N.Harel, and K.U "High-field fMRI unveils orientation columns in humans," Proceedings of the National Academy of Sciences of the United States of America, vol.105, no.30, pp.10607-10612, 2008; and J.Budde, G.Shajan, J. Hoffmann, K. and R. Pohmann,"Human imaging at 9.4T using T 2 * -,phase-,and susceptibility-weighted contrast,"Magnetic Resonance in Medicine,vol.65,no.2,pp.544-550,2011. [0003] Traditional magnetic resonance imaging requires multiple radio frequency pulse excitations to obtain complete K-space ...

AI Technical Summary

Problems solved by technology

However, since the K-space data needs to be sampled twice in the opposite direction, the echo chain length of this method is long or the time domain resolution is low (V.B.Xie, M.Lyu, and E.X.Wu, "EPINyquist ghost and geometric distortion correction by two-frame phaselabeling,"Magnetic Resonance in Medicine,vol.77,no.5,pp.1749-1761,2017)

Images