Image reconstruction method and device, computer device and storage medium

Abstract

The application relates to an image reconstruction method and device, computer equipment and a storage medium. A k-space data set of a target part is acquired, the k-space data set is collected in a k-space center complete sampling and k-space edge undersampling mode, a low spatial resolution T2 parameter map is reconstructed based on the k-space data set, and then a high spatial resolution T2 parameter map is reconstructed according to the k-space data set and the low spatial resolution T2 parameter map. The method improves the time resolution of magnetic resonance imaging and ensures the spatial resolution, so that the high time and spatial resolution of magnetic resonance dynamic imaging can be ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to an image reconstruction method and device, computer equipment and a storage medium. BACKGROUND

[0002] Magnetic resonance imaging (MRI) is an imaging technique that uses the signals generated by atomic nuclei in a strong magnetic field to produce images. This technique can obtain high-contrast clear images of the internal sample / tissue without damage and ionizing radiation, and has been widely used in medical diagnosis.

[0003] Generally, due to the constraints of some factors, the imaging speed of magnetic resonance is very slow, and the slower imaging speed greatly limits the image temporal resolution of magnetic resonance imaging in dynamic imaging of moving organs such as the heart, and also produces serious motion artifacts in the image, which reduces the image quality.

[0004] Therefore, how to ensure the high temporal and spatial resolution of magnetic resonance dynamic imaging has become a technical problem to be solved. SUMMARY

[0005] Therefore, it is necessary to provide an image reconstruction method, device, computer equipment and storage medium to solve the above technical problems, which can ensure the high temporal and spatial resolution of magnetic resonance dynamic imaging.

[0006] In a first aspect, an image reconstruction method is provided, which comprises:

[0007] Obtaining a k-space data set of a target part, and the acquisition mode of the k-space data set is k-space center full sampling and k-space edge undersampling;

[0008] Reconstructing a low spatial resolution T2 parameter map based on the k-space data set; the low spatial resolution refers to a spatial resolution less than a first preset value;

[0009] Reconstructing a high spatial resolution T2 parameter map according to the k-space data set and the low spatial resolution T2 parameter map; the high spatial resolution refers to a spatial resolution greater than a second preset value.

[0010] In one embodiment, the reconstruction of the high spatial resolution T2 parameter map according to the k-space data set and the low spatial resolution T2 parameter map comprises:

[0011] Inputting the k-space data set, the low spatial resolution T2 parameter map and a reconstruction parameter into a preset first reconstruction model to obtain the high spatial resolution T2 parameter map; the first reconstruction model includes a reconstruction regular constraint term of the low spatial resolution T2 parameter map.

[0012] In one of the embodiments, the obtaining process of the first reconstruction model comprises:

[0013] constructing a target function of the initial reconstruction model according to the k-space dataset, the reconstruction parameter and the sparsity parameter; the target function comprises a data fidelity term, a sparsity regular constraint term and a reconstruction regular constraint term; the reconstruction regular constraint term comprises a low spatial resolution T2 parameter map;

[0014] minimizing the target function, iteratively optimizing the initial reconstruction model until a preset iteration condition is met, and obtaining the first reconstruction model.

[0015] In one of the embodiments, the k-space dataset comprises N k-space data, the N k-space data have different T2 preparation times, and N is a positive integer.

[0016] In one of the embodiments, the low spatial resolution T2 parameter map is reconstructed based on the k-space dataset, comprising:

[0017] reconstructing at least one low spatial resolution T2 weighted image according to the k-space dataset;

[0018] reconstructing the low spatial resolution T2 parameter map based on the at least one low spatial resolution T2 weighted image.

[0019] In one of the embodiments, the low spatial resolution T2 parameter map is reconstructed based on the at least one low spatial resolution T2 weighted image, comprising:

[0020] fitting the signal intensity values of the pixels of the at least one low spatial resolution T2 weighted image to obtain a fitting value as the T2 value of each pixel;

[0021] an image formed by the pixels represented by the T2 values is determined as the low spatial resolution T2 parameter map.

[0022] In one of the embodiments, the low spatial resolution T2 parameter map is reconstructed based on the at least one low spatial resolution T2 weighted image, comprising:

[0023] inputting the k-space dataset and the reconstruction parameter into a preset second reconstruction model to obtain the low spatial resolution T2 parameter map.

[0024] In a second aspect, the embodiments of the present application provide an image reconstruction device, which comprises:

[0025] a data acquisition module configured to acquire a k-space dataset of a target part, the k-space dataset being acquired in a k-space center full sampling and k-space edge undersampling manner;

[0026] The first reconstruction module is configured to reconstruct a low spatial resolution T2 parameter map based on the k-space dataset; the low spatial resolution represents a spatial resolution less than a first preset value.

[0027] The second reconstruction module is configured to reconstruct a high spatial resolution T2 parameter map based on the k-space dataset and the low spatial resolution T2 parameter map; the high spatial resolution represents a spatial resolution greater than a second preset value.

[0028] In a third aspect, an embodiment of the present application provides a computer device, including a memory and a processor, the memory stores a computer program, and the processor implements the method steps of any one of the embodiments of the first aspect when executing the computer program.

[0029] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method steps of any one of the embodiments of the first aspect.

[0030] The image reconstruction method, device, computer device and storage medium provided by the embodiments of the present application acquire a k-space dataset of a target part, the k-space dataset is acquired in a manner of k-space center full sampling and k-space edge undersampling; and a low spatial resolution T2 parameter map is reconstructed based on the k-space dataset, and then a high spatial resolution T2 parameter map is reconstructed based on the k-space dataset and the low spatial resolution T2 parameter map. In the method, since the low spatial resolution T2 parameter map is reconstructed based on the k-space data, the k-space data can be acquired in an undersampling manner, that is, the k-space data is acquired in a manner of k-space center full sampling and k-space edge undersampling, so that the amount of data of the acquired k-space data is reduced, the scanning time is greatly shortened, the time resolution of the magnetic resonance imaging is improved, and after the low spatial resolution T2 parameter map is obtained, the high resolution T2 parameter map is further reconstructed based on the low spatial resolution T2 parameter map, and the spatial resolution is also ensured, so that the high time and spatial resolution of the magnetic resonance dynamic imaging can be ensured. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 An application environment diagram of an image reconstruction method provided in an embodiment;

[0032] Figure 2 A flowchart of an image reconstruction provided in an embodiment;

[0033] Figure 3 A flowchart of an image reconstruction provided in another embodiment;

[0034] Figure 4 A flowchart of an image reconstruction provided in another embodiment;

[0035] Figure 5 a flowchart of an image reconstruction process provided in another embodiment;

[0036] Figure 6 a flowchart of an image reconstruction process provided in another embodiment;

[0037] Figure 7 a structural block diagram of an image reconstruction apparatus provided in an embodiment;

[0038] Figure 8 an internal structural diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0039] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.

[0040] The image reconstruction method provided by the present application can be applied in an application environment as shown in the accompanying drawings. Figure 1 The internal structure of the computer device includes a processor for providing computing and control capabilities, a memory including a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database is used to store related data of the image reconstruction method process. The network interface is used to communicate with other external devices through network connection. The computer program is executed by the processor to implement an image reconstruction method.

[0041] Before the technical solutions of the embodiments of the present application are specifically introduced, the technical background or the technical evolution context based on which the embodiments of the present application are introduced. First, the T2 parameter map in the embodiments of the present application is explained: taking the heart as the target site, cardiac magnetic resonance imaging (CMR) can evaluate the degree of myocardial edema through T2 weighted sequence images. These images are reconstructed based on the prolongation of transverse relaxation time caused by edema. Due to the accumulation of water, the proportion of free water increases, and myocardial edema appears as high signal intensity on the T2 weighted image. Each pixel on the T2 weighted image represents different signal intensity, and the specific performance is that different signal intensity displays different gray values. T2 weighted images can be qualitatively evaluated, and T2 quantitative sequence (T2 mapping) is the "upgrade" of T2 weighted images, which is used for quantitative measurement, and reflects the T2 value of myocardial tissue through numerical value. If the myocardium has edema, the T2 value of the myocardial tissue will usually increase. Based on this, the image in which the value of each pixel is a specific T2 value can be obtained through the T2 mapping technology to serve as a diagnostic image, and this image is the T2 parameter map.

[0042] In actual application, in order to minimize the image blur caused by the movement of the heart, the acquisition window must be narrowed, preferably no more than 50 ms, but such a narrow acquisition window is not enough to cover a sufficient k-space range to achieve a certain spatial resolution, so it is also unrealistic to improve the image quality through a narrow acquisition window. On the other hand, the patient's discomfort is increased by the long imaging time in the magnetic resonance imaging process, and is easily affected by respiratory motion, resulting in motion artifacts, so that the k-space data cannot be collected too much, and it is common to collect 3, so when fitting the T2 weighted image to the T2 parameter map, the number of k-space data is also small, which will also lead to poor final image quality.

[0043] In the related art, the T2 parameter map is obtained by separately and independently reconstructing the collected multiple k-space data into corresponding T2 weighted images, and then performing image processing on the multiple T2 weighted images to obtain the T2 parameter map. In this method, the T2 parameter map is directly obtained from the T2 weighted image, that is, the spatial resolution of the T2 parameter map is positively correlated with the spatial resolution of the T2 weighted image. The T2 weighted image is reconstructed from the k-space data, and therefore, if a high spatial resolution T2 weighted image is to be obtained, full sampling or under sampling of a large amount of data is required during k-space data acquisition to ensure a high spatial resolution of the reconstructed T2 weighted image. However, full sampling or under sampling of a large amount of data during k-space data acquisition requires a large amount of data to be collected, which requires a longer time and reduces the time resolution. Conversely, performing the operation reduces the spatial resolution of the T2 parameter map. Therefore, how to ensure a high time and spatial resolution of magnetic resonance dynamic imaging is still a technical problem to be solved. Based on this, the embodiments of the present application provide an image reconstruction method, device, computer equipment and storage medium, which can ensure a high time and spatial resolution of magnetic resonance dynamic imaging. In addition, it should be noted that the applicant has made a lot of creative labor from the discovery of the above technical defects and the technical solutions described in the following embodiments.

[0044] The technical solutions of the present application and how the technical solutions solve the above technical problems will be described in detail below through embodiments and in combination with the drawings. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described again in some embodiments. It should be noted that when the image reconstruction method provided by the embodiments of the present application is described below, the execution subject is a computer device. In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments.

[0045] In one embodiment, as shown in Figure 2 An image reconstruction method is provided. The embodiment relates to the specific process that a computer device reconstructs a low spatial resolution T2 parameter map from k-space data of a target part, and then reconstructs a high spatial resolution T2 parameter map based on the k-space data set and the low spatial resolution T2 parameter map; the embodiment includes the following steps:

[0046] S101, acquiring a k-space data set of a target part, and the k-space data set is collected in a k-space center full sampling and k-space edge under sampling manner.

[0047] The data collected by the nuclear magnetic resonance scanning is k-space data, that is, frequency domain data, which is used to represent the space of spatial frequency in the magnetic resonance image. Fourier inverse transform of the k-space data can decode the spatial localization encoding information in the original data to obtain image domain data, that is, magnetic resonance image data. Reconstruction is the process of converting k-space data into image domain data. The k-space data points and image pixel points are in the Fourier transform and inverse transform relationship, and each data point in the k-space contains the information of the entire image. Therefore, the premise of reconstructing the T2 parameter map is to obtain the k-space data first.

[0048] In the embodiments of the present application, the target site can be a moving organ such as the heart, coronary artery, or other applicable organs, which are not limited in the embodiments of the present application. Taking the heart as an example, the k-space data set of the heart needs to be obtained. Optionally, the k-space data set includes N k-space data, and the N k-space data have different T2 preparation times, and N is a positive integer. For example, N is equal to 3, and the obtained k-space data set includes three k-space data, and the three k-space data have different T2 preparation times. For the heart, a heartbeat cycle includes diastole and systole, so the three k-space data of the heart can be collected one by one, the first k-space data is collected after the heart recovers, the second k-space data is collected when the heart beats again, and the third k-space data is collected after the heart recovers and beats again. The preparation time is the preparation time during the collection of one k-space data, which is different from the waiting time. For multiple k-space data, the preparation times of different k-space data are different. The specific preparation time can be determined according to the actual situation, which is not limited in the embodiments of the present application.

[0049] When the k-space data in the k-space data set is obtained, the collection method is k-space center full sampling and k-space edge undersampling.

[0050] When collecting k-space data, the sampling frequency of the undersampling mode does not satisfy the Nyquist sampling frequency, which can cause aliasing artifacts. The energy of an image is mainly concentrated in the center region of its k-space. If the center region of the k-space is undersampled, the aliasing artifacts will contain most of the low-frequency components. The edge of the k-space contains little energy of an image, so the undersampling of the edge does not cause serious aliasing artifacts. Therefore, the scanning time can be reduced by fully sampling the center region of the k-space and undersampling the edge of the k-space. The range of the fully sampled center region of the k-space and the degree of undersampling of the edge region other than the center region (different degrees represent different amounts of sampled data) can be determined according to the subsequent need of reconstructing a low spatial resolution T2 parameter map, which is not limited in the embodiments of the present application. Moreover, the data positions of the edge undersampling of different k-space data are also different. For example, although both are undersampling, the first k-space data collects data of 1, 3, and 7, and the second k-space data collects data of 2, 4, and 9, which are not limited in the embodiments of the present application.

[0051] It should be emphasized that the description of the collection mode in the embodiments of the present application is only an example of the collection mode. In the embodiments of the present application, the k-space data set can be obtained from a database of a computer device, downloaded from other online platforms, or collected according to a simulated model by simulating the heart beating through a specific software, which is not limited in the embodiments of the present application. That is, the k-space data obtained in the image reconstruction of the embodiments of the present application is not real-time collection, but is obtained by downloading or calling.

[0052] In S102, a low spatial resolution T2 parameter map is reconstructed based on the k-space data set. The low spatial resolution means a spatial resolution less than a first preset value.

[0053] After obtaining the k-space data set of the target part, a low spatial resolution T2 parameter map is reconstructed according to the k-space data set. The spatial resolution refers to the minimum detail that can be displayed in the image when the density resolution is greater than 10%. The density resolution refers to the minimum density difference that can be distinguished between tissues. The two are mutually restrictive. The spatial resolution is closely related to the pixel size, which is generally 1.5 times the width of the pixel. The smaller the pixel size and the more the number of pixels, the higher the spatial resolution and the clearer the image. Therefore, the degree of the low spatial resolution can be determined according to the preset critical value, for example, the first value of the preset critical value is 4, so that the spatial resolution less than 4 is the low spatial resolution. Of course, the first value is only an example, and the critical value of the low spatial resolution reconstructed in the actual application can be determined according to the specific situation.

[0054] For example, the manner of reconstructing the low spatial resolution T2 parameter map based on the k-space data set can be to use a pre-trained neural network model, take the acquired k-space data set as input, and then the result output by the neural network model is the low spatial resolution T2 parameter map. Alternatively, each k-space data in the k-space data set can be used to reconstruct a corresponding magnetic resonance image, and then the low spatial resolution T2 parameter map is reconstructed from the magnetic resonance image, and the like. The embodiments of the present application do not limit this.

[0055] S103, reconstructing a high spatial resolution T2 parameter map according to the k-space data set and the low spatial resolution T2 parameter map; the high spatial resolution represents a spatial resolution greater than a second preset value.

[0056] According to the acquired k-space data set and the reconstructed low spatial resolution T2 parameter map, a high spatial resolution T2 parameter map is further reconstructed.

[0057] Similarly, the degree of high spatial resolution can also be determined according to a preset threshold value. For example, the preset threshold value is a second value equal to 8, and then the spatial resolution greater than 8 is the high spatial resolution. Of course, the second value here is only an example, and in actual application, the threshold value of the reconstructed high spatial resolution can be determined according to specific circumstances. Moreover, it can be understood that the threshold value for dividing the low spatial resolution and the high spatial resolution can be the same value or different values, that is, the first preset value can be equal to the second preset value or not equal to the second preset value.

[0058] For example, an embodiment of reconstructing a high spatial resolution T2 parameter map according to a k-space data set and a low spatial resolution T2 parameter map can be to use a preset algorithm, such as bilinear difference or nearest neighbor difference. Specifically, a simple bilinear difference can be performed on the low spatial resolution T2 parameter map, then a hash algorithm is used to quickly divide the image blocks into different categories (buckets), for each category, four pre-trained filters are used for linear filtering, and then the results of different image blocks are fused to obtain the final high spatial resolution T2 parameter map.

[0059] Another embodiment can also be to train a preset neural network model with multiple k-space data sets and low spatial resolution T2 parameter maps, and corresponding high spatial resolution T2 parameter maps, so that the trained neural network model can output the corresponding high spatial resolution T2 parameter map according to the k-space data set and the low spatial resolution T2 parameter map.

[0060] The image reconstruction method provided in the embodiment is used for acquiring a k-space data set of a target part, the k-space data set is collected in a manner of complete sampling of a k-space center and under-sampling of a k-space edge, and a low spatial resolution T2 parameter map is reconstructed based on the k-space data set, and then a high spatial resolution T2 parameter map is reconstructed according to the k-space data set and the low spatial resolution T2 parameter map. In the method, since the low spatial resolution T2 parameter map is reconstructed based on the k-space data, the k-space data can be collected in an under-sampling manner, that is, collected in a manner of complete sampling of a k-space center and under-sampling of a k-space edge, so that the amount of k-space data collected is reduced, the scanning time is greatly shortened, the time resolution of magnetic resonance imaging is improved, and after the low spatial resolution T2 parameter map is obtained, the high spatial resolution T2 parameter map is further reconstructed according to the low spatial resolution T2 parameter map, and the spatial resolution is also ensured, so as to ensure the high time and spatial resolution of magnetic resonance dynamic imaging.

[0061] On the basis of the above-mentioned embodiment, the following provides an implementation manner of reconstructing a high spatial resolution T2 parameter map according to a k-space data set and a low spatial resolution T2 parameter map, the embodiment comprising: inputting the k-space data set, the low spatial resolution T2 parameter map and a reconstruction parameter into a preset first reconstruction model to obtain the high spatial resolution T2 parameter map; the first reconstruction model comprising a reconstruction regular constraint term of the low spatial resolution T2 parameter map.

[0062] The first reconstruction model can be the pre-trained neural network model mentioned above, or can be an algorithm model constructed according to a preset reconstruction algorithm. Taking the algorithm model as an example, the first reconstruction model comprises the reconstruction regular constraint term of the low spatial resolution T2 parameter map, the reconstruction regular constraint term is used for constraining and preventing overfitting from the low spatial resolution T2 parameter map to the high spatial resolution T2 parameter map, and the low spatial resolution T2 parameter map introduced as prior knowledge is used to estimate the actual result of the high spatial resolution T2 parameter, so that the transition from the low spatial resolution T2 parameter map to the high spatial resolution T2 parameter map is more accurate.

[0063] The reconstruction parameter refers to any parameter required for the reconstruction process, for example, a time modulation pattern, which is a characteristic representing the change of the modulation signal in amplitude, frequency and phase over time, and is a parameter preset before the scanning of the k-space data, and can be directly obtained for use in the process of reconstructing the image from the k-space data, i.e., it is a known value in the reconstruction process. For another example, the reconstruction parameter includes a coil sensitivity map. In a magnetic resonance device, the electromagnetic wave detected by a receiving coil is used as a magnetic resonance signal, so the coil sensitivity refers to the degree of response of the receiving coil to the input signal, and the higher the value, the stronger the ability to detect weak signals, and the map composed of the coil sensitivity values of the entire coil is the coil sensitivity map. The above two parameters are only examples, and the reconstruction parameter is not limited in the embodiments of the present application.

[0064] For the present embodiment, the k-space data set, the low spatial resolution T2 parameter map and the reconstruction parameter are all known values, so that these data are input into the preset first reconstruction model, and the output result is the high spatial resolution T2 parameter map. This makes it faster to obtain the high spatial resolution T2 parameter map from the low spatial resolution T2 parameter map.

[0065] Optionally, as shown in Figure 3 the acquisition process of the first reconstruction model includes the following steps:

[0066] S201, constructing a target function of an initial reconstruction model according to the k-space data set, the reconstruction parameter and the sparse parameter; the target function includes a data fidelity term, a sparse regular constraint term and a reconstruction regular constraint term; the reconstruction regular constraint term includes the low spatial resolution T2 parameter map.

[0067] The model can be simply understood as a function, so for the construction process of the above first reconstruction model, the target function is required as the iterative target. Then, the k-space data set, the reconstruction parameter and the sparse parameter are required to construct the target function. The k-space data set and the reconstruction parameter herein are consistent with the meanings in the foregoing embodiments, and the difference is that the k-space data set and the reconstruction parameter herein can be the data in the training set, i.e., a large amount of diversified training data is obtained as the training set. The sparse parameter refers to a parameter for measuring the spatial smoothness of the image, for example, a spatial difference or other sparse operators, etc.

[0068] The objective function includes a data fidelity term, a sparse regularization constraint term and a reconstruction regularization constraint term. The data fidelity term is used to ensure that the result conforms to the degradation process. The regularization terms are used to enhance the output. The sparse regularization constraint term and the reconstruction regularization constraint term are both used to enhance the output. The difference is that the sparse regularization constraint term is used to enhance the output from the spatial smoothness dimension, and the reconstruction regularization constraint term is used to enhance the output from the low spatial resolution T2 parameter map as prior knowledge to estimate the high spatial resolution T2 parameter and prevent overfitting dimension.

[0069] For example, the objective function is as follows: In the objective function, This part is the data fidelity term, wherein x i represents any k-space data reconstructed T2-weighted image in the k-space data set, and subscript i represents any one; T2 represents a high spatial resolution T2 parameter map; y i represents any k-space data in the k-space data set, which is actual data; p1 represents a reconstruction parameter time modulation mode; S represents a reconstruction parameter coil sensitivity map, and represents an inversion recovery signal model, wherein F (x i , T2) is equal to a*Exp (-t / b) + c, wherein a is an initial signal amplitude, b is a T2 value, and the offset c is optional. p1FSF (x i , T2) represents the k-space data inversely estimated according to the reconstruction parameter and the inversion signal model, and F represents a Fourier transform; therefore, the data fidelity term is used to minimize the difference between the actual k-space data and the estimated k-space data to ensure that the final output T2 parameter map meets the requirements. Wherein, is a sparse regularization constraint term, and λ i and β are regularization coefficients, which balance the error between the regularization term and the data fidelity term. When the regularization sparse variable is large, the solution tends to be smooth, and otherwise the edges of the solution are sharpened. T is a sparse parameter space difference. Wherein, γ|T2-T 2low |1 is a reconstruction regularization constraint term, and γ is also a regularization coefficient, which balances the error between the regularization term and the data fidelity term; T 2low represents the low spatial resolution T2 parameter map obtained in the foregoing. The sparse regularization constraint term and the reconstruction regularization constraint term can both promote the sparsity in some transform domain.

[0070] In S202, the objective function is minimized, the initial reconstruction model is iteratively optimized until the preset iteration condition is met, and the first reconstruction model is obtained.

[0071] After the objective function is constructed, the objective function is minimized, that is, the minimum value of the objective function is solved. For example, a preset iteration condition can be set as reaching a preset iteration number T or the iteration process converging. Then, the constructed initial reconstruction model is alternately iterated and optimized, and the update of the regularization term parameter is performed until the preset iteration condition is met, the training is completed, and the first reconstruction model is obtained.

[0072] From the above objective function, it can be seen that only x i and T2 are unknown parameters, and other data are known quantities. The output of the first reconstruction model constructed by the objective function is a T2 weighted image corresponding to each k-space data in the k-space data set and a final high spatial resolution T2 parameter map. Thus, the high spatial resolution T2 parameter map is obtained.

[0073] In the embodiment of the application, the k-space data set, the low spatial resolution T2 parameter map, and the reconstruction parameter are input into the preset first reconstruction model to obtain the high spatial resolution T2 parameter map, so that the high spatial resolution T2 parameter map can be quickly and accurately reconstructed from the low spatial resolution T2 parameter map. In addition, the data fidelity term, the sparse regularization constraint term, and the reconstruction regularization constraint term are included in the training objective function of the first reconstruction model. In the training process, the error minimization of the data output result and the sparsity and overfitting constraints are performed to ensure the correctness of the output result of the finally trained first reconstruction model.

[0074] The process of reconstructing the low spatial resolution T2 parameter map based on the k-space data set is described below through different embodiments.

[0075] As shown in FIG. 1, in one embodiment, the process of reconstructing the low spatial resolution T2 parameter map based on the k-space data set includes the following steps: Figure 4

[0076] S301, at least one low spatial resolution T2 weighted image is reconstructed according to the k-space data set.

[0077] The k-space data set includes a plurality of k-space data, for example, three k-space data. Then, at least one low spatial resolution T2 weighted image can be reconstructed based on each k-space data. Of course, one low spatial resolution T2 weighted image can also be reconstructed for each k-space data, which is not limited in the embodiment of the application.

[0078] ​The manner of reconstructing the low spatial resolution T2 weighted image according to the k-space data set can be that the image obtained after inverse Fourier transformation of the k-space data is the low spatial resolution T2 weighted image. It should be noted that in the embodiments of the present application, the image obtained after inverse Fourier transformation of the k-space data set is called a low spatial resolution T2 weighted image because the k-space data is acquired in the manner defined in the embodiments of the present application, that is, the k-space center region is fully sampled and the edge region is undersampled. In this manner, the amount of k-space data is small, and the spatial resolution of the reconstructed magnetic resonance image (T2 weighted image) is very low, so it is called a low spatial resolution T2 weighted image.

[0079] S302, reconstruct a low spatial resolution T2 parameter map based on the at least one low spatial resolution T2 weighted image.

[0080] After obtaining the low spatial resolution T2 weighted image, a low spatial resolution T2 parameter map is reconstructed according to the low spatial resolution T2 weighted image. Here, the difference between the T2 weighted image and the T2 parameter map mentioned above can be understood, that is, each pixel on the T2 weighted image represents the signal intensity of different tissues, which is specifically represented by different gray scales. The stronger the tissue signal intensity, the brighter the corresponding pixel point on the T2 weighted image; on the contrary, the weaker the tissue signal intensity, the darker the corresponding pixel point on the T2 weighted image. However, each pixel on the T2 parameter map represents a specific T2 value, not signal intensity.

[0081] Optionally, as shown in Figure 5 reconstructing the low spatial resolution T2 parameter map based on the low spatial resolution T2 weighted image includes:

[0082] S401, fitting the signal intensity values of the pixels of the at least one low spatial resolution T2 weighted image to obtain a fitting value as the T2 value of each pixel.

[0083] Taking an example in which the k-space data set includes three k-space data, and each k-space data corresponds to a reconstructed T2 weighted image, that is, fitting the signal intensity values of the pixels of the three low spatial resolution T2 weighted images to obtain a fitting value as the T2 value of each pixel.

[0084] Assuming that the pixel points of an image are 10*10, that is, 100, and three low spatial resolution T2-weighted images are T21, T22 and T23, then the signal intensity of the pixel points corresponding to the first row and the first column of the three images T21, T22 and T23 is fitted, for example, by a preset fitting function, and the fitting value obtained after fitting is determined as the T2 value, that is, the T2 value of the pixel points corresponding to the first row and the first column. In this way, the T2 values of the other 99 pixel points can be obtained, and the T2 values of the 100 pixel points are obtained comprehensively.

[0085] S402, the image formed by representing each pixel point by a T2 value is determined as a low spatial resolution T2 parameter map.

[0086] After obtaining the T2 values of each pixel point, an image formed by representing each pixel point by a T2 value is a low spatial resolution T2 parameter map.

[0087] In the embodiments of the present application, after the k-space data is reconstructed into a T2-weighted image, a low spatial resolution T2 parameter map is fitted based on the signal intensity values of each pixel point in the T2-weighted image. Since the signal intensity values of each pixel point in the T2-weighted image represent the characteristics of the target site tissue in T2 characteristics, fitting the T2 values of the corresponding pixel points based on the signal intensity values of each pixel point can make the low spatial resolution T2 parameter map obtained finally more accurate.

[0088] In another embodiment, a low spatial resolution T2 parameter map is reconstructed based on at least one low spatial resolution T2-weighted image, including: inputting a k-space data set and reconstruction parameters into a preset second reconstruction model to obtain a low spatial resolution T2 parameter map.

[0089] Similarly, the second reconstruction model can also be the pre-trained neural network model mentioned above, or an algorithm model constructed according to a preset reconstruction algorithm. The k-space data set and the reconstruction parameters in the second reconstruction model can be referred to the description in the foregoing embodiments, which will not be repeated here.

[0090] Taking an algorithm model as an example, the process of constructing the second reconstruction model also needs to construct an objective function. For example, the objective function of the second reconstruction model is: In the objective function, the first part is This part is a data fidelity term, where x i Refers to the T2-weighted image reconstructed by any k-space data in the k-space data set, and the subscript i represents any one; T2 refers to a high spatial resolution T2 parameter map; y irepresents any k-space data in the k-space data set, which is actual data; p1 represents a reconstruction parameter time modulation mode; S represents a reconstruction parameter coil sensitivity map, and Φ represents an inversion recovery signal model, where Φ(x i , T2) is equal to a*Exp(-t / b))+c, where a is an initial signal amplitude, b is a T2 value, and the offset c is optional. p1FSΦ(x i , T2) represents k-space data inversely estimated according to the reconstruction parameter and the inversion signal model, and F represents a Fourier transform; therefore, the data fidelity term is to minimize the difference between the actual k-space data and the estimated k-space data, so that the final output T2 parameter map meets the requirements. Wherein, is a sparse regularization constraint term, and λ i and β are regularization coefficients, which balance the error between the regularization term and the data fidelity term. When the regularization sparsity is large, the solution tends to be smooth, and otherwise the edges of the solution are sharpened. T is a sparse parameter space difference.

[0091] After constructing the objective function of the second reconstruction model, the objective function is minimized, and the initial reconstruction model of the second reconstruction model is iteratively optimized until a preset iteration condition is met, to obtain the second reconstruction model. In the second reconstruction model, only x i and x i are unknown parameters, and other data are known quantities, and the output of the second reconstruction model is a T2 weighted image corresponding to each k-space data in the k-space data set and a low spatial resolution T2 parameter map.

[0092] The second reconstruction model is used to obtain a low spatial resolution T2 parameter map from a k-space data set and a reconstruction parameter. Since the model is constructed in advance, and the error minimization and sparse constraint of the data output result are performed when the model is constructed, the constructed model can quickly and accurately reconstruct a low spatial resolution T2 parameter map from the k-space data set.

[0093] In addition, the embodiment of the present application also provides an image reconstruction method, as shown in Figure 6 , the embodiment includes:

[0094] S1, obtaining a k-space data set of a target part, the k-space data set is collected in a k-space center complete sampling and k-space edge undersampling manner; performing S2 or S5.

[0095] S2, reconstructing at least one low spatial resolution T2 weighted image according to the k-space data set; performing S3.

[0096] S3, fitting the signal intensity values of each pixel point of the at least one low spatial resolution T2 weighted image, and determining the fitting values as the T2 values of each pixel point; performing S4.

[0097] S4, determine the image composed of the pixel points represented by T2 values as a low spatial resolution T2 parameter map; perform S6.

[0098] S5, input the k-space data set and the reconstruction parameter into a preset second reconstruction model to obtain a low spatial resolution T2 parameter map; perform S6.

[0099] S6, construct a target function of an initial reconstruction model according to the k-space data set, the reconstruction parameter and a sparse parameter; the target function comprises a data fidelity term, a sparse regular constraint term and a reconstruction regular constraint term; the reconstruction regular constraint term comprises the low spatial resolution T2 parameter map; perform S7.

[0100] S7, minimize the target function, iteratively optimize the initial reconstruction model until a preset iteration condition is met to obtain a first reconstruction model; perform S8.

[0101] S8, input the k-space data set, the low spatial resolution T2 parameter map and the reconstruction parameter into a preset first reconstruction model to obtain a high spatial resolution T2 parameter map; the first reconstruction model comprises the reconstruction regular constraint term of the low spatial resolution T2 parameter map.

[0102] The image reconstruction method provided by the embodiment has similar implementation principles and technical effects to those of the above-described method embodiments, and thus will not be described here.

[0103] It should be understood that, although each step in the flowchart of the above-described embodiment is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless explicitly stated herein, the execution of these steps has no strict sequence limitation, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart of the above-described embodiment can comprise multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0104] In one embodiment, as shown in Figure 7 An image reconstruction apparatus is provided, which comprises a data acquisition module 10, a first reconstruction module 11 and a second reconstruction module 12, wherein:

[0105] The data acquisition module 10 is configured to acquire a k-space data set of a target part, and the k-space data set is acquired in a k-space center full sampling and k-space edge undersampling manner.

[0106] The first reconstruction module 11 is configured to reconstruct a low spatial resolution T2 parameter map based on the k-space dataset; the low spatial resolution represents a spatial resolution less than a first preset value.

[0107] The second reconstruction module 12 is configured to reconstruct a high spatial resolution T2 parameter map based on the k-space dataset and the low spatial resolution T2 parameter map; the high spatial resolution represents a spatial resolution greater than a second preset value.

[0108] In one embodiment, the data acquisition module 10 is configured to input the k-space dataset, the low spatial resolution T2 parameter map and the reconstruction parameter into a preset first reconstruction model to obtain the high spatial resolution T2 parameter map; the first reconstruction model includes a reconstruction regular constraint term of the low spatial resolution T2 parameter map.

[0109] In one embodiment, the device further includes:

[0110] The construction module is configured to construct a target function of an initial reconstruction model based on the k-space dataset, the reconstruction parameter and the sparsity parameter; the target function includes a data fidelity term, a sparsity regular constraint term and a reconstruction regular constraint term; the reconstruction regular constraint term includes the low spatial resolution T2 parameter map.

[0111] The optimization module is configured to minimize the target function, iteratively optimize the initial reconstruction model until a preset iteration condition is met, and obtain the first reconstruction model.

[0112] In one embodiment, the k-space dataset includes N k-space data, the N k-space data have different T2 preparation times, and N is a positive integer.

[0113] In one embodiment, the first reconstruction module 11 includes:

[0114] The first reconstruction unit is configured to reconstruct at least one low spatial resolution T2 weighted image based on the k-space dataset.

[0115] The second reconstruction unit is configured to reconstruct a low spatial resolution T2 parameter map based on the at least one low spatial resolution T2 weighted image.

[0116] In one embodiment, the second reconstruction unit is specifically configured to fit signal intensity values of each pixel point of the at least one low spatial resolution T2 weighted image, and determine a fitting value as a T2 value of each pixel point; an image formed by representing each pixel point by the T2 value is determined as the low spatial resolution T2 parameter map.

[0117] In one embodiment, the first reconstruction module 11 includes a third reconstruction unit configured to input the k-space dataset and the reconstruction parameter into a preset second reconstruction model to obtain the low spatial resolution T2 parameter map.

[0118] The specific limitations of the image reconstruction device can refer to the limitations of the image reconstruction method described above, which will not be repeated here. Each module in the above image reconstruction device can be realized by software, hardware and their combination. The above modules can be embedded in the processor in the computer device in hardware form or independent of the processor in the computer device, or stored in the memory in the computer device in software form, so that the processor calls and executes the operations corresponding to the above modules.

[0119] In one embodiment, a computer device is provided, which can be a terminal, and its internal structure diagram can be as shown in Figure 8 The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in wired or wireless mode. Wireless mode can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement an image reconstruction method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0120] Those skilled in the art can understand that Figure 8 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0121] In one embodiment, a computer device is provided, which includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the following steps:

[0122] Obtain a k-space data set of a target site, and the k-space data set is collected in a k-space center full sampling and k-space edge undersampling manner;

[0123] Reconstruct a low spatial resolution T2 parameter map based on the k-space data set; the low spatial resolution represents a spatial resolution less than a first preset value;

[0124] The high spatial resolution T2 parameter map is reconstructed according to the k-space data set and the low spatial resolution T2 parameter map; the high spatial resolution represents a spatial resolution greater than a second preset value.

[0125] The computer device provided in the above embodiment has similar implementation principles and technical effects to the above method embodiments, and thus details are not repeated here.

[0126] In one embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the following steps:

[0127] The k-space data set of the target part is acquired, and the k-space data set is acquired in a manner of k-space center full sampling and k-space edge undersampling;

[0128] The low spatial resolution T2 parameter map is reconstructed based on the k-space data set; the low spatial resolution represents a spatial resolution less than a first preset value;

[0129] The high spatial resolution T2 parameter map is reconstructed according to the k-space data set and the low spatial resolution T2 parameter map; the high spatial resolution represents a spatial resolution greater than a second preset value.

[0130] The computer readable storage medium provided in the above embodiment has similar implementation principles and technical effects to the above method embodiments, and thus details are not repeated here.

[0131] Those skilled in the art can understand that all or part of the processes in the above method embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, the computer program can include the processes of the above method embodiments. In the embodiments provided in the present application, any reference to a memory, storage, database or other medium can include at least one of a non-volatile memory and a volatile memory. The non-volatile memory can include a read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory or an optical memory. The volatile memory can include a random access memory (RAM) or an external cache memory. As an illustration but not limitation, the RAM can be in various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM).

[0132] Any combination of the technical features in the above embodiments can be made, and for the sake of brevity, not all possible combinations are described above, however, as long as the combination of the technical features does not exist in contradiction, it shall be considered within the scope of the present disclosure.

[0133] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it shall not be understood as a limitation on the patent scope of the present application. It shall be pointed out that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, and these shall be within the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method of image reconstruction, characterized by, The method comprises: acquiring a k-space data set of a target site, the k-space data set being acquired in a manner of full sampling of a k-space center and under-sampling of a k-space edge; the target site being a moving organ, and a plurality of k-space data of the k-space data set having different T2 preparation times; reconstructing a low spatial resolution T2 parameter map based on the k-space data set; the low spatial resolution representing a spatial resolution less than a first preset value; inputting the k-space data set, the low spatial resolution T2 parameter map and reconstruction parameters into a preset first reconstruction model to obtain a high spatial resolution T2 parameter map; the first reconstruction model including a reconstruction regular constraint term of the low spatial resolution T2 parameter map; the high spatial resolution representing a spatial resolution greater than a second preset value.

2. The method of claim 1, wherein, Different k-space data are different in data position of edge under-sampling.

3. The method of claim 1, wherein, The acquisition process of the first reconstruction model comprises: constructing an objective function of an initial reconstruction model according to the k-space data set, reconstruction parameters and sparse parameters; the objective function including a data fidelity term, a sparse regular constraint term and a reconstruction regular constraint term; the reconstruction regular constraint term including a low spatial resolution T2 parameter map; minimizing the objective function to iteratively optimize the initial reconstruction model until a preset iteration condition is met to obtain the first reconstruction model.

4. The method according to any one of claims 1 to 3, characterized in that, The k-space data set includes N k-space data, the N k-space data having different T2 preparation times, and N being a positive integer.

5. The method according to any one of claims 1 to 3, characterized in that, The reconstructing a low spatial resolution T2 parameter map based on the k-space data set comprises: reconstructing at least one low spatial resolution T2 weighted image according to the k-space data set; reconstructing the low spatial resolution T2 parameter map based on the at least one low spatial resolution T2 weighted image.

6. The method of claim 5, wherein, The reconstructing the low spatial resolution T2 parameter map based on the at least one low spatial resolution T2 weighted image comprises: fitting signal intensity values of each pixel point of the at least one low spatial resolution T2 weighted image to obtain a fitting value as a T2 value of each pixel point; determining an image formed by each pixel point expressed by the T2 value as the low spatial resolution T2 parameter map.

7. The method of claim 5, wherein, The reconstructing the low spatial resolution T2 parameter map based on the at least one low spatial resolution T2 weighted image comprises: inputting the k-space data set and reconstruction parameters into a preset second reconstruction model to obtain the low spatial resolution T2 parameter map.

8. An image reconstruction apparatus, characterized by comprising: The device comprises: a data acquisition module configured to acquire a k-space data set of a target site, the k-space data set being acquired in a manner of full sampling of a k-space center and under-sampling of a k-space edge; the target site being a moving organ, and a plurality of k-space data of the k-space data set having different T2 preparation times; a first reconstruction module configured to reconstruct a low spatial resolution T2 parameter map based on the k-space data set; A second reconstruction module configured to input the k-space dataset, the low spatial resolution T2 parameter map and a reconstruction parameter into a preset first reconstruction model to obtain a high spatial resolution T2 parameter map; the first reconstruction model comprises a reconstruction regular constraint term of the low spatial resolution T2 parameter map; the high spatial resolution represents a spatial resolution greater than a second preset value. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 7.

Images