A method for sparse mri reconstruction based on a combination of convolutional neural networks and iterative methods

Abstract

The invention discloses a sparse MRI reconstruction method based on the combination of convolutional neural network and iterative method. The method first prepares a data set, including training data and test data, the training data is used for training the network, and the test data is used for testing For the trained network, each set of data contains a set of samples and labels. The samples are low-quality data with noise and artifacts obtained by dividing the highly down-sampled k-space data into low-frequency data and high-frequency data, and performing zero-fill reconstruction respectively. The high-frequency image and low-frequency image of , the label is the high-quality MR image without noise and artifacts corresponding to the low-quality image. Two networks with the same structure are trained using low-frequency data and high-frequency data respectively, one is used to reconstruct high-frequency k-space data, and the other is used to reconstruct low-frequency k-space data, and the addition of the two reconstruction results is the final reconstruction result required. The invention utilizes less k-space data, has faster reconstruction speed and higher image quality.

Description

technical field [0001] The present invention relates to image processing, in particular to a sparse MRI reconstruction method based on the combination of convolutional neural network and iterative method. Background technique [0002] Magnetic resonance imaging (Magnetic Resonance Imaging, MRI) is realized by the action of high-frequency magnetic field outside the body, which generates signals from the material in the body radiating energy to the surrounding environment. The imaging process is similar to image reconstruction and CT. Compared with CT, its main advantages are: Ionizing radiation has no radioactive or biological damage to brain tissue. It can directly make tomographic images of cross-section, sagittal plane, coronal plane and various oblique planes, without artifacts such as ray hardening in CT images. It shows that the pathological process of the disease is more extensive and the structure is clearer than that of CT. Isodense lesions that are completely norm...

AI Technical Summary

Problems solved by technology

[0003] Purpose of the invention: The present invention aims at the problem of slow imaging in the prior art, and provides a sparse MRI reconstruction method based on the combination of convolutional neural network and iterative method. With the support of compressed sensing theory, the present invention combines deep learning and proposes In order to use the convolutional neural network to learn the sparse representation of the image, (the sparse representation of the image is the key point of the compressed sensing theory for MRI reconstruction of downsampling, the higher the sparseness of the image, the better the noise and artifacts in the image can be The removal, the structure in the image can be better restored

Images