Fast multi-channel magnetic resonance imaging method based on pcu-net network

Abstract

The invention discloses a fast multi-channel magnetic resonance imaging method based on a PCU-Net network, comprising three steps of data processing and division, construction and training optimization of the PCU-Net network, and multi-channel magnetic resonance image reconstruction. This method extends the complex U-shaped convolutional neural network to the training and prediction of multi-channel data, extracts the features between multiple channels through the multi-channel complex module, and trains the network based on the multi-channel complex mean square error for multi-channel undersampling data MRI image reconstruction, and speed up network convergence by looping in network parameters. Experimental results show that the method of the present invention can not only reconstruct multi-channel magnetic resonance images with high quality, but also quickly reconstruct multi-channel images based on optimized parameters after training to meet the needs of real-time online reconstruction.

Description

technical field [0001] The invention belongs to the field of magnetic resonance imaging, in particular to a fast multi-channel magnetic resonance imaging method based on a PCU-Net network. Background technique [0002] Magnetic resonance imaging (Magnetic Resonance Imaging, MRI) is a non-invasive medical imaging method without ionizing radiation, which has been widely used in clinical auxiliary diagnosis. However, in practical applications, the application of this technology is limited due to the disadvantages of slow MRI imaging and the tendency to produce motion artifacts. [0003] Parallel magnetic resonance imaging (pMRI) and compressed sensing (CS) are both important MRI acceleration methods. pMRI uses multiple parallel coils to receive spatial induction signals at the same time, and realizes spatial information according to the difference information of the spatial sensitivity of each coil obtained. The encoding can reduce the filling of k-space phase encoding lines, ...

AI Technical Summary

Problems solved by technology

[0009] The patents or articles based on deep learning published above mainly use the stacking and prior information of the convolutional modules in the neural network to study multi-channel magnetic resonance image reconstruction, while the method using the U-Net convolutional neural network is mainly used for research Based on single-channel data of real numbers, although Wang S et al. have used complex modules combined with convolutional neural networks to study multi-channel data based on complex numbers, the proposed convolutional network is only a cascade of simple convolution modules and does not use U- Net convolutional neural network, and there have not been any patents or articles on multi-channel magnetic resonance imaging based on complex U-Net (CU-Net) networks

Images