阅读说明：本技术 磁共振图像处理方法、装置、存储介质和磁共振成像系统 (Magnetic resonance image processing method, magnetic resonance image processing device, storage medium and magnetic resonance imaging system ) 是由 李国斌 刘楠 黄小倩 廖术 于 2019-01-16 设计创作，主要内容包括：本发明实施例公开了一种磁共振图像处理方法、装置、存储介质和磁共振成像系统。该方法包括：获取待校正数据,并将待校正数据输入伪影校正模型,生成初始校正数据,其中,伪影校正模型基于神经网络模型预先训练获得,初始校正数据对应的k空间包括比待校正数据对应的k空间多的高频成分；分别对待校正数据和初始校正数据进行加权处理,生成加权结果,并对两个加权结果进行融合处理,生成目标校正k空间数据；重建目标校正k空间数据,生成伪影校正后的目标校正磁共振图像。通过上述技术方案,实现了在保持伪影纠正后的磁共振图像的分辨率和信噪比基本不变,也不增加扫描时间的情况下,获得伪影校正效果更好的磁共振图像。(The embodiment of the invention discloses a magnetic resonance image processing method, a magnetic resonance image processing device, a storage medium and a magnetic resonance imaging system. The method comprises the following steps: acquiring data to be corrected, inputting the data to be corrected into an artifact correction model, and generating initial correction data, wherein the artifact correction model is obtained by pre-training based on a neural network model, and a k space corresponding to the initial correction data comprises more high-frequency components than the k space corresponding to the data to be corrected; respectively carrying out weighting processing on the data to be corrected and the initial correction data to generate weighting results, and carrying out fusion processing on the two weighting results to generate target correction k-space data; reconstructing target correction k-space data to generate an artifact corrected target correction magnetic resonance image. By the technical scheme, the magnetic resonance image with better artifact correction effect is obtained under the condition that the resolution and the signal-to-noise ratio of the magnetic resonance image after artifact correction are basically kept unchanged and the scanning time is not increased.)

1. A magnetic resonance image processing method characterized by comprising:

acquiring data to be corrected, inputting the data to be corrected into an artifact correction model, and generating initial correction data, wherein the artifact correction model is obtained by pre-training based on a neural network model, and a k space corresponding to the initial correction data comprises more high-frequency components than a k space corresponding to the data to be corrected;

respectively carrying out weighting processing on the data to be corrected and the initial correction data to generate weighting results, and carrying out fusion processing on the two weighting results to generate target correction k-space data;

and reconstructing the target correction k-space data to generate an artifact corrected target correction magnetic resonance image.

2. The method of claim 1, wherein obtaining data to be corrected comprises:

acquiring original k-space data corresponding to the data to be corrected, wherein a k-space central area in the original k-space data is filled;

and carrying out numerical filling on the residual k space region except the k space central region in the original k space data to generate the data to be corrected.

3. The method according to claim 1, before performing weighting processing on the data to be corrected and the initial correction data respectively to generate weighting results, further comprising:

and generating a first weight matrix and a second weight matrix according to a preset weight value distribution rule.

4. The method according to claim 3, wherein the preset weight value distribution rule is that the sum of a first weight value in the first weight matrix and a second weight value in the second weight matrix corresponding to the same element in k-space data is 1,

and the weight value of each matrix element acting on a first set area in the k-space data in the first weight matrix is greater than the weight value of each matrix element acting on a second set area in the k-space data.

5. The method according to claim 4, wherein the first defined region is a k-space central region and the second defined region is a remaining k-space region excluding the k-space central region.

6. The method according to claim 3, wherein the data to be corrected and the initial correction data are weighted respectively, and generating the weighted result comprises:

carrying out weighting processing on the data to be corrected according to the first weight matrix to generate a first weighting result;

and carrying out weighting processing on the initial correction data according to the second weighting matrix to generate a second weighting result.

7. The method of claim 1, wherein the artifact correction model is pre-trained by:

obtaining at least two groups of model training data, wherein each group of model training data comprises input data and expected output data, and artifact components in the expected output data are less than artifact components in the input data;

training the set neural network model by taking the input data as training input data of the set neural network model and the expected output data as training constraint data of the set neural network model to obtain the artifact correction model;

wherein the k-space to which the desired output data corresponds contains more high frequency components than the k-space to which the input data corresponds.

8. A magnetic resonance image processing apparatus characterized by comprising:

the initial correction data generation module is used for acquiring data to be corrected, inputting the data to be corrected into an artifact correction model and generating initial correction data, wherein the artifact correction model is obtained by pre-training based on a neural network model, and a k space corresponding to the initial correction data comprises more high-frequency components than a k space corresponding to the data to be corrected;

the weighted fusion module is used for respectively carrying out weighted processing on the data to be corrected and the initial correction data to generate weighted results and carrying out fusion processing on the two weighted results to generate target correction k space data;

and the reconstruction module is used for reconstructing the target correction k-space data and generating a target correction magnetic resonance image after artifact correction.

9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the image processing method according to any one of claims 1 to 7.

10. A magnetic resonance imaging system, comprising:

an MRI scanner for scanning a subject positioned therein and generating data to be corrected in relation to the subject;

an image processor, communicatively coupled to the MRI scanner, programmed to:

acquiring the data to be corrected, and generating initial correction data according to the data to be corrected, wherein k space corresponding to the initial correction data to be corrected comprises more high-frequency components than k space corresponding to the data to be corrected;

respectively carrying out weighting processing on the data to be corrected and the initial correction data to generate weighting results, and carrying out fusion processing on the two weighting results to generate target correction k-space data;

and reconstructing the target correction k-space data to generate an artifact corrected target correction magnetic resonance image.

Technical Field

The embodiment of the invention relates to a medical image processing technology, in particular to a magnetic resonance image processing method, a magnetic resonance image processing device, a storage medium and a magnetic resonance imaging system.

Background

The magnetic resonance image taken by an imaging system, such as a Magnetic Resonance Imaging (MRI) system, may be represented as magnetic resonance image data in the spatial domain or as magnetic resonance image related data in k-space, i.e. the frequency domain. Sharp transitions in the magnetic resonance image, e.g. near the boundary of an organ, can be demonstrated in k-space using relatively high frequency components. However, limited sampling times or poor signal-to-noise ratios (SNR) may lead to undersampling of magnetic resonance image-related data in k-space (k-space data for short). This may lead to insufficient high frequency components in the magnetic resonance image data, resulting in a "fringe ringing" phenomenon, referred to as "truncation artifacts", in the reconstructed magnetic resonance image.

There are two main categories of current methods for reducing truncation artifacts in reconstructed magnetic resonance images. One type of method is to apply low-pass filtering to the acquired k-space data. However, this method filters out the high frequency data of k-space, which results in a blurred magnetic resonance image. In addition, when the Sinc function interpolation is performed on the reconstructed magnetic resonance image, the method can not effectively filter the truncation artifact strengthened by the Sinc function interpolation. Another type of method is to extrapolate k-space data or interpolate magnetic resonance image domain data. However, the method of interpolating values in magnetic resonance image domain data has a poor effect of suppressing the mosaic effect due to insufficient resolution; the constraint strength of the k-space data extrapolation method is not easy to control, if the constraint is too loose, a heavier truncation artifact cannot be effectively inhibited, and if the constraint is too heavy, the appearance of the magnetic resonance image can be modified, so that the details of the magnetic resonance image look unnatural.

Disclosure of Invention

The embodiment of the invention provides a magnetic resonance image processing method, a magnetic resonance image processing device, a storage medium and a magnetic resonance imaging system, which aim to obtain a magnetic resonance reconstructed image with better artifact correction effect under the condition of keeping the resolution and the signal-to-noise ratio of a magnetic resonance image after artifact correction basically unchanged and not increasing the scanning time.

In a first aspect, an embodiment of the present invention provides a magnetic resonance image processing method, including:

acquiring data to be corrected, inputting the data to be corrected into an artifact correction model, and generating initial correction data, wherein the artifact correction model is obtained by pre-training based on a neural network model, and a k space corresponding to the initial correction data comprises more high-frequency components than a k space corresponding to the data to be corrected;

respectively carrying out weighting processing on the data to be corrected and the initial correction data to generate weighting results, and carrying out fusion processing on the two weighting results to generate target correction k-space data;

and reconstructing the target correction k-space data to generate an artifact corrected target correction magnetic resonance image.

In a second aspect, an embodiment of the present invention further provides a magnetic resonance image processing apparatus, including:

the initial correction data generation module is used for acquiring data to be corrected, inputting the data to be corrected into an artifact correction model and generating initial correction data, wherein the artifact correction model is obtained by pre-training based on a neural network model, and a k space corresponding to the initial correction data comprises more high-frequency components than a k space corresponding to the data to be corrected;

the weighted fusion module is used for respectively carrying out weighted processing on the data to be corrected and the initial correction data to generate weighted results and carrying out fusion processing on the two weighted results to generate target correction k space data;

and the reconstruction module is used for reconstructing the target correction k-space data and generating a target correction magnetic resonance image after artifact correction.

In a third aspect, the embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the magnetic resonance image processing method provided by any of the embodiments of the present invention.

In a fourth aspect, an embodiment of the present invention further provides a magnetic resonance imaging system, including:

an MRI scanner for scanning a subject positioned therein and generating data to be corrected in relation to the subject;

an image processor, communicatively coupled to the MRI scanner, programmed to:

acquiring the data to be corrected, and generating initial correction data according to the data to be corrected, wherein k space corresponding to the initial correction data to be corrected comprises more high-frequency components than k space corresponding to the data to be corrected;

respectively carrying out weighting processing on the data to be corrected and the initial correction data to generate weighting results, and carrying out fusion processing on the two weighting results to generate target correction k-space data;

and reconstructing the target correction k-space data to generate an artifact corrected target correction magnetic resonance image.

According to the embodiment of the invention, the artifact correction model obtained based on the neural network model training is used for carrying out artifact correction on the data to be corrected containing the truncation artifact/Gibbs artifact to generate the initial correction data with the suppressed artifact, and the suppression degree of the truncation artifact is improved under the condition of not increasing the scanning time. The target correction magnetic resonance image after artifact correction is generated through the weighted fusion processing of the data to be corrected containing more artifact components and the initial correction data with the artifact effectively restrained, so that the target correction magnetic resonance image contains partial data in the data to be corrected and partial data in the initial correction data, and the magnetic resonance image with better artifact correction effect is obtained under the condition that the resolution and the signal-to-noise ratio of the magnetic resonance image after artifact correction are basically unchanged.

Drawings

Fig. 1 is a flowchart of a magnetic resonance image processing method according to a first embodiment of the present invention;

FIG. 2A is a schematic diagram of a k-space data distribution according to a first embodiment of the present invention;

FIG. 2B is a schematic view of another k-space data distribution in accordance with one embodiment of the present invention;

fig. 3A is a flowchart of a magnetic resonance image processing method according to a second embodiment of the present invention;

FIG. 3B is a diagram illustrating an artifact correction model according to a second embodiment of the present invention;

fig. 4 is a schematic structural diagram of a magnetic resonance image processing apparatus according to a third embodiment of the present invention;

fig. 5 is a schematic structural diagram of a magnetic resonance imaging system in the fifth embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.