Rapid magnetic resonance image reconstruction method based on undersampling

Abstract

The invention relates to a rapid magnetic resonance image reconstruction method based on undersampling, and the method comprises the following steps: step 1, collecting a large amount of magnetic resonance data, and taking the data as a data set of a deep reconstruction network model; step 2, performing data enhancement on the training data by using all mask matrixes possibly existing in the under-scanning process, and expanding the data set acquired in the step 1; step 3, performing normalization processing on the data set expanded in the step 2; step 4, establishing a double-domain parallel reconstruction network; and step 5, training the double-domain parallel reconstruction network established in the step 4 by using the data subjected to normalization processing in the step 3 to obtain a trained double-domain parallel reconstruction network, performing normalization preprocessing on an acquired pre-scanning image, sending the pre-scanning image into the trained network for forward propagation, and outputting the pre-scanning image to obtain a reconstructed image. According to the method, the network has the capability of performing high-definition reconstruction of the nuclear magnetic resonance image by using the undersampled data, and artifacts caused by the condition violating the sampling theorem can be eliminated.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a reconstruction method of a magnetic resonance image, in particular to a fast magnetic resonance image reconstruction method based on undersampling. Background technique [0002] At present, magnetic resonance imaging technology has become an indispensable means of examination in medical diagnosis, and its imaging of the brain and other soft tissues can achieve good diagnostic results, and its current clinical application is very important and common. However, the current MRI still faces the disadvantages of long acquisition time and poor patient comfort. Long acquisition time and long appointment time severely limit the application and popularization of MRI. [0003] Compressed sensing technology (CS) [David Donoho. Compressed sensing. IEEE Transactions on Information Theory, 52(4): 1289–1306, 2006.], which appeared around 2006, is an important breakthrough that greatly ...

AI Technical Summary

Problems solved by technology

This method can be regarded as a pioneer in the sequential processing of the frequency domain and the image domain, but it has not been well applied and effective in MRI accelerated reconstruction due to the computational complexity.

[0007] The previous machine learning-based methods usually only focus on the learning and reconstruction of magnetic resonance images in the image space, or use a simple codec network, or through the cascade of codec networks, and through the intermediate layer for k-space data. network, but the optimization for undersampling is still only optimized in the image domain

The GrappaNet method published in 2020 [Sriram, A., Zbontar, J., Murrell, T., Zitnick, C.L., Defazio, A., & Sodickson, D.K. (2020). GrappaNet: Combining parallel imaging with deep learning for multi-coilMRI reconstruction.In Proceedings of the IEEE / CVF Conference on ComputerVisionand Pattern Recognition,pages 14315-14322.], involving network models optimized and reconstructed in the frequency domain and image domain, respectively, the input k-space data will be respectively optimized through the frequency domain network and The image domain optimizes the network, and then cascades to achieve a certain reconstruction effect, but it still does not use the feature relationship between the frequency domain and image domain reconstruction network, which can further improve the reconstruction effect

[0008] To sum up, the previous methods are all reconstruction methods only in the image domain or frequency domain, or the method of reconstruction optimization in the frequency domain and image domain successively, the reconstruction effect is still limited, and it is difficult to achieve higher acceleration magnification, there are still some deficiencies

Images