Magnetic resonance fingerprinting imaging method, imaging device, storage medium, magnetic resonance device

Abstract

The application provides a magnetic resonance fingerprint imaging method, an imaging device, a storage medium and a magnetic resonance device. The magnetic resonance fingerprint imaging method comprises: acquiring K-space data at each preset time point as original data; compressing the original data by using a transformation matrix to generate compressed data; performing image reconstruction on the compressed data to obtain an image time sequence; and generating a quantitative map according to the image time sequence. It can be seen that the application compresses data before image reconstruction, and the subsequent image reconstruction calculation process is based on the compressed data, thereby greatly reducing the calculation amount in the image reconstruction process, effectively improving the image reconstruction speed, and further improving the magnetic resonance fingerprint imaging efficiency.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance imaging technology, and in particular to a magnetic resonance fingerprint imaging method, imaging device, storage medium, and magnetic resonance device. Background Technology

[0002] Magnetic resonance fingerprinting (MRF) sequences involve three main processes: raw data acquisition, image reconstruction, and quantitative mapping. Since quantitative mapping is based on the time-series information of each voxel, MRF fingerprinting requires acquiring a large amount of time-point information and using the data acquired at each time point for image reconstruction, a process with a very high computational cost. To address this, a dictionary compression method based on singular value decomposition (SVD) has been proposed to reduce the computational burden of the quantitative process in MRF fingerprinting. The core idea is to apply SVD to the time dimension compression of the dictionary matrix, thereby compressing the original dictionary of size (D_entry x D_TR) to (D_entry x D_svd), where D_svd is much smaller than D_TR. Furthermore, the same transformation matrix is ​​also used to compress the image dimension data of the MRF fingerprint signal, reducing the dictionary's storage space and the computational burden of the dictionary matching process.

[0003] However, existing magnetic resonance fingerprinting methods still involve a large overall computational load and have not effectively addressed this issue. Therefore, a new processing method is urgently needed to solve the aforementioned technical problems. Summary of the Invention

[0004] The purpose of this invention is to provide a magnetic resonance fingerprint imaging method, imaging device, storage medium, and magnetic resonance device to solve at least one of the problems of how to reduce the computational load in the image reconstruction step and how to improve the speed of magnetic resonance fingerprint imaging.

[0005] To address the aforementioned technical problems, this invention provides a magnetic resonance fingerprint imaging method, comprising:

[0006] Collect K-space data at each preset time point as the raw data;

[0007] The original data is compressed using a transformation matrix to generate compressed data;

[0008] The compressed data is used to reconstruct images to obtain image time series.

[0009] A quantitative map is generated based on the image time series.

[0010] Optionally, in the magnetic resonance fingerprinting method, the time dimension of the original data is greater than the time dimension of the compressed data.

[0011] Optionally, in the magnetic resonance fingerprinting method, before compressing the original data using a transformation matrix to generate compressed data, the following steps are included:

[0012] The original data is reconstructed according to the sampling trajectory types in K-space, so that each preset time point in the reconstructed original data corresponds to a data segment of the corresponding type of sampling trajectory.

[0013] The K-space data collected at each preset time point is assigned to the data segment of the corresponding type of sampling trajectory, and the data segment of the sampling trajectory of each type not collected at each preset time point is assigned 0.

[0014] Optionally, in the magnetic resonance fingerprinting method, the original data after reconstruction is compressed using a transformation matrix to generate the compressed data; wherein the number of data segments in the compressed data is the same as the number of data segments in the reconstructed original data.

[0015] Optionally, in the magnetic resonance fingerprinting method, the process of generating a quantitative map based on the image time series includes: matching the image time series with a dictionary, and obtaining the quantitative map based on the matching result.

[0016] Optionally, in the magnetic resonance fingerprinting method, before matching the image time series with the dictionary and obtaining the quantitative map based on the matching result, the dictionary is compressed using the transformation matrix to generate the compressed dictionary;

[0017] Furthermore, when matching the image time series with the dictionary, the image time series is matched with the compressed dictionary, and the quantitative map is obtained based on the matching result.

[0018] Optionally, in the magnetic resonance fingerprinting method, the process of generating a quantitative map based on the image time series includes:

[0019] A machine learning model is trained using a dictionary, or the machine learning model is trained using self-supervised learning.

[0020] The image time series is input into a preset machine learning model to generate the quantitative map.

[0021] Based on the same inventive concept, the present invention also provides an imaging device for performing the magnetic resonance fingerprint imaging method described above; wherein, the imaging device includes: an information acquisition unit, a data compression unit, an image reconstruction unit, and a quantitative imaging unit;

[0022] The information acquisition unit is used to collect K-space data at each preset time point as raw data;

[0023] The data compression unit is used to compress the original data using a transformation matrix to generate compressed data;

[0024] The image reconstruction unit is used to reconstruct the image from the compressed data to obtain an image time series.

[0025] The quantitative imaging unit is used to generate a quantitative map based on the image time series.

[0026] Based on the same inventive concept, the present invention also provides a computer storage medium storing executable instructions, which, when executed by a processor, cause the processor to perform the steps in the magnetic resonance fingerprinting method.

[0027] Based on the same inventive concept, the present invention also provides a magnetic resonance imaging device, including the imaging device and / or the computer storage medium.

[0028] In summary, this invention provides a magnetic resonance fingerprint imaging method, imaging device, storage medium, and magnetic resonance device. The magnetic resonance fingerprint imaging method includes: acquiring K-space data at each preset time point as raw data; compressing the raw data using a transformation matrix to generate compressed data; reconstructing the compressed data to obtain an image time series; and generating a quantitative map based on the image time series. It is evident that this invention compresses the data before image reconstruction, so the subsequent image reconstruction calculation process is based on the compressed data, thereby significantly reducing the computational load in the image reconstruction process, effectively improving the speed of image reconstruction, and thus improving the efficiency of magnetic resonance fingerprint imaging. Attached Figure Description

[0029] Those skilled in the art will understand that the accompanying drawings are provided to better understand the invention and do not constitute any limitation on the scope of the invention. Wherein:

[0030] Figure 1 This is a flowchart of the magnetic resonance fingerprinting method in Embodiment 1 of the present invention.

[0031] Figure 2 This is a schematic diagram of the collection of raw data in Embodiment 3 of the present invention.

[0032] Figure 3 This is a schematic diagram of the reconstructed original data in Embodiment 3 of the present invention.

[0033] Figure 4 This is a schematic diagram of compressed data in Embodiment 4 of the present invention.

[0034] Figure 5 This is a flowchart of obtaining a quantitative graph using compressed dictionary matching in Embodiment Six of the present invention.

[0035] Figure 6 This is a flowchart of obtaining a quantitative graph using a machine learning model in Embodiment 7 of the present invention.

[0036] Figure 7 This is a schematic diagram of the imaging device in Embodiment 8 of the present invention.

[0037] In the attached image:

[0038] 100 - Information acquisition unit; 101 - Data compression unit; 102 - Image reconstruction unit; 103 - Quantitative imaging unit; 104 - Data reconstruction unit; 105 - Dictionary generation unit; 106 - Dictionary matching unit. Detailed Implementation

[0039] To make the objectives, advantages, and features of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the drawings are all in a very simplified form and are not drawn to scale, and are only used to facilitate and clearly illustrate the objectives of the embodiments of the present invention. Furthermore, the structures shown in the drawings are often part of the actual structures. In particular, different figures may emphasize different aspects and sometimes use different scales. It should also be understood that, unless specifically stated or indicated, the terms "first," "second," "third," etc., in the specification are only used to distinguish the various components, elements, steps, etc., in the specification, and are not used to indicate the logical or sequential relationships between the various components, elements, steps, etc.

[0040] <Example 1>

[0041] Please see Figure 1 This embodiment provides a magnetic resonance fingerprint imaging method, including:

[0042] Step 1 S10: Collect K-space data at each preset time point as raw data.

[0043] It should be noted that K-space is the dual space of ordinary space under Fourier transform, and K-space includes the spatial range of all scanned voxels. Optionally, the tissue to be detected is scanned using a magnetic resonance imaging (MRI) scanner to obtain acquisition data for each voxel in the tissue, which is then used as the raw data. Furthermore, before acquiring the raw data, acquisition parameters need to be designed, such as the radio frequency flip angle (FA), repetition time (TR), and K-space sampling trajectory. These acquisition parameters change continuously over time so that multiple tissue characteristics in the tissue can be measured in a single acquisition.

[0044] Furthermore, in this embodiment, each preset time point is a repeating time (TR). Optionally, the number of preset time points is greater than or equal to 1000 to ensure imaging quality. Based on this, the raw data includes a time dimension and a data dimension. The time dimension is one-dimensional; the data dimension can be one-dimensional or multi-dimensional.

[0045] Step 2 S20: Compress the original data using a transformation matrix to generate compressed data.

[0046] This embodiment does not specifically limit the transformation matrix A. The transformation matrix A can be a compression matrix used in Singular Value Decomposition (SVD), or it can be any [nT x nt] matrix. For example, a diagonal matrix A1(i, i) = 1; i = 0, 1, 2, ..., nt. That is, after multiplying the diagonal matrix A1 with the reconstructed original data M2, only the data from the first nt time points are retained. It is understood that the data capacity of the compressed data is smaller than that of the original data, so as to reduce the amount of computation and improve imaging efficiency in the subsequent image reconstruction process.

[0047] Step 3 S30: Reconstruct the compressed data to obtain an image time series.

[0048] Image reconstruction is the construction of an image based on the acquisition data of all voxels at corresponding time points, thus generating a detected tissue image at each time point. Furthermore, based on the detected tissue images at all time points, the evolution curves of different quantitative parameters for any voxel can be obtained. These quantitative parameters include, but are not limited to, longitudinal relaxation time (T1) and lateral relaxation time (T2). Since image reconstruction is the most computationally intensive step in the entire magnetic resonance fingerprinting algorithm, directly affecting imaging efficiency, this embodiment uses compressed data for image reconstruction to reduce computational load, shorten computation time, and thus improve imaging efficiency. Further, the image time series includes the reconstructed detected tissue image at each preset time point.

[0049] Step 4S40: Generate a quantitative map based on the image time series.

[0050] Furthermore, after acquiring the image time series, it is necessary to generate the evolution curve of the quantitative parameters of each voxel based on the image time series, and obtain the quantitative map of the detected tissue based on the evolution curve.

[0051] In summary, the magnetic resonance fingerprint imaging method provided in this embodiment compresses data before image reconstruction, so the subsequent image reconstruction calculation process is based on the compressed data, which greatly reduces the amount of computation in the image reconstruction process and effectively improves the speed of magnetic resonance fingerprint imaging.

[0052] <Example 2>

[0053] This embodiment provides a magnetic resonance fingerprinting method, including: based on the method provided in Embodiment 1, the time dimension of the original data is greater than the time dimension of the compressed data. It is understood that the compression of the original data in this embodiment is time-dimension compression. For example, the original data has 1000 time points, while the compressed data has 10 time points. That is, the data corresponding to every ten time points in the original data are compressed into a data segment corresponding to one time point in the compressed data. Optionally, the data corresponding to the ten time points in the original data are located in a data segment corresponding to one time point in the compressed data using a linear combination transformation.

[0054] In summary, the magnetic resonance fingerprinting method provided in this embodiment compresses the acquired raw data in the time dimension, reducing the amount of raw data. Image reconstruction is then performed based on the compressed data, significantly reducing the computational load during image reconstruction and effectively improving image reconstruction speed, thereby enhancing the efficiency of magnetic resonance fingerprinting. For any aspects not detailed in this embodiment, please refer to the descriptions in other embodiments.

[0055] <Example 3>

[0056] This embodiment provides a magnetic resonance fingerprint imaging method, including: based on the method provided in Embodiment 1, before compressing the original data using a transformation matrix to generate compressed data, reconstructing the original data according to the sampling trajectory types in K-space, so that each preset time point in the reconstructed original data corresponds to a data segment of the sampling trajectory of the corresponding type, assigning the K-space data collected at each preset time point to the data segment of the sampling trajectory of the corresponding type, and setting 0 to the data segment of the sampling trajectory of the type not collected at each preset time point.

[0057] It should be noted that there are many types of sampling trajectories in K-space, and the sampling trajectory corresponding to the data collected at each time point is not the same. For example, such as Figure 2As shown, a non-uniform spiral sampling trajectory is used to sparsely sample the K-space data at each time point to obtain the K-space data at that time point. Assuming there are four types of sampling trajectories in the K-space, and these four trajectories cycle sequentially, then the sampling trajectory at the first time point TR1 is the first sampling trajectory S1, the sampling trajectory at the second time point TR2 is the second sampling trajectory S2, the sampling trajectory at the third time point TR3 is the third sampling trajectory S3, the sampling trajectory at the fourth time point TR4 is the fourth sampling trajectory S4, the sampling trajectory at the fifth time point TR5 is the first sampling trajectory S1, and so on. This embodiment does not limit the sampling trajectory to a non-uniform spiral sampling trajectory; it can also be a Cartesian sampling, planar echo imaging, or radial trajectory, etc. Furthermore, this embodiment does not limit the number or types of sampling trajectories; they can be... Figure 2 The example of four types could also be three, five, or six types, etc.

[0058] As mentioned above, the K-space acquisition trajectory changes with time points, so directly compressing the original data without considering the sampling trajectory has no imaging value. Therefore, the original data needs to be reconstructed before compression. Specifically, the original data is reconstructed according to the types of K-space sampling trajectories, so that each preset time point in the reconstructed original data corresponds to a data segment of all types of sampling trajectories. The K-space data acquired at each preset time point is assigned to the data segment of the corresponding type of sampling trajectory, and the data segments of the sampling trajectories of types not acquired at each preset time point are set to 0.

[0059] Please see Figure 3Assuming the number of preset time points is 1000, and the length of the K-space data collected at each preset time point is 1600, then the obtained raw data M1 = [1600 x 1000]. Furthermore, the sampling trajectory of the K-space is a continuously looping first sampling trajectory S1, second sampling trajectory S2, third sampling trajectory S3, and fourth sampling trajectory S4, resulting in 4 types of sampling trajectories in the K-space. Therefore, the data length corresponding to each preset time point is expanded to four times, and divided into a first data segment D1, a second data segment D2, a third data segment D3, and a fourth data segment D4. The first data segment D1 corresponds to the data collected by the first sampling trajectory S1, the second data segment D2 corresponds to the data collected by the second sampling trajectory S2, the third data segment D3 corresponds to the data collected by the third sampling trajectory S3, and the fourth data segment D4 corresponds to the data collected by the fourth sampling trajectory S4. Thus, the raw data can be expanded into a matrix M2 = [6400 x 1000]. That is, while keeping the time dimension of the original data unchanged, the data dimension of the original data is expanded according to the types of the sampling trajectories, so that each column of the data dimension in the reconstructed original data contains all types of sampling trajectories.

[0060] After expanding the data segments of the original data, the K-space data collected at each preset time point is assigned to the data segment of the corresponding type of sampling trajectory, and the data segment of the sampling trajectory of each type not collected at each preset time point is assigned a value of 0. For example, the sampling trajectory within the first preset time point TR1 is the first sampling trajectory S1, and the first K-space data N1 is collected; then the first data segment D1 is assigned a value of N1, and the second data segment D2, the third data segment D3, and the fourth data segment D4 are all assigned a value of 0. Based on this, the data dimension corresponding to the first preset time point TR1 is [N1, 0, 0, 0]. Similarly, the sampling trajectory within the second preset time point TR2 is the second sampling trajectory S2, and the second K-space data N2 is collected; the data dimension corresponding to the second preset time point TR2 is [0, N2, 0, 0]. The sampling trajectory within the third preset time point TR3 is the third sampling trajectory S3, and the third K-space data N3 is collected; the data dimension corresponding to the third preset time point TR3 is [0, 0, N3, 0]. The sampling trajectory within the fourth preset time point TR4 is the fourth sampling trajectory S4, and the fourth K-space data N4 is collected; the data dimension corresponding to the fourth preset time point TR4 is [0, 0, 0, N4]. The sampling trajectory within the fifth preset time point TR5 is the first sampling trajectory S1, and the fifth K-space data N5 is collected; the data dimension corresponding to the fifth preset time point TR5 is [N5, 0, 0, 0]. In summary, the reconstructed original data M2 is:

[0061]

[0062] In summary, the magnetic resonance fingerprinting method provided in this embodiment reconstructs the original data before compressing it. At each preset time point, data segments corresponding to all sampling trajectories are added to clearly define the sampling trajectories during subsequent data processing, avoiding direct compression that could damage the original data and affect subsequent imaging. For details not covered in this embodiment, please refer to the descriptions in other embodiments.

[0063] <Example 4>

[0064] This embodiment provides a magnetic resonance fingerprint imaging method, including: based on the methods provided in Embodiments 1 and 3, compressing and reconstructing the original data using a transformation matrix to generate compressed data; wherein the number of data segments in the compressed data is the same as the number of data segments in the reconstructed original data.

[0065] Please see Figure 4 Assume that the transformation matrix in this embodiment is used to compress the time dimension and is collected cyclically using four different sampling trajectories. Wherein, the original data M1 = [1600x1000], the reconstructed original data M2 = [6400x1000], and the transformation matrix A = [1000x10], then M2*A = [6400x1000]*[1000x10] = [6400x10] = M3.

[0066] Therefore, the transformation matrix A compresses the 1000 time points in the original data to 10 time points, but the data segments in the original data remain unchanged. That is, the data segments in the original data are the same as the data segments in the reconstructed original data M2, both being four, and each data segment includes 1600 data points. It can be understood that each data segment corresponds to one type of sampling trajectory, so the compressed data at each time point also corresponds to data segments of all types of sampling trajectories, preserving all the original data to the greatest extent.

[0067] It should be noted that the time points T1, T2, T3, T4, T5... of the compressed data after the transformation matrix compression are abstract time points and do not correspond to the repetition time of the data acquisition phase. Instead, they are linear combinations of each preset time point before compression. Similarly, the data at each abstract time point is also a linear combination of the original data. For example, the reconstructed original data M2 is:

[0068]

[0069] Assuming the data from the first ten time points is compressed into the first abstract time point T1, then the four data segments at this abstract time point T1 are as follows:

[0070] Q1=a*N1+b*0+c*0+d*0+e*N5+f*0+g*0+h*0+i*N9+j*0

[0071] Q2=a*0+b*N2+c*0+d*0+e*0+f*N6+g*0+h*0+i*0+j*N10

[0072] Q3=a*0+b*0+c*N3+d*0+e*0+f*0+g*N7+h*0+i*0+j*0

[0073] Q4=a*0+b*0+c*0+d*N4+e*0+f*0+g*0+h*N8+i*0+j*0

[0074] Where a, b, c, d, e, f, j, h, and i are all coefficients.

[0075] In summary, the magnetic resonance fingerprinting method provided in this embodiment compresses the reconstructed original data. While reducing the data volume, it compresses the collected data of different sampling trajectories separately to ensure that each compressed abstract time point also corresponds to data segments of all sampling trajectories, thus guaranteeing data integrity and better imaging value. For any details not covered in this embodiment, please refer to the descriptions in other embodiments.

[0076] <Example 5>

[0077] This embodiment provides a magnetic resonance fingerprinting method, including: based on the method provided in Embodiment 1, the process of generating a quantitative map based on the image time series includes: matching the image time series with a dictionary, and obtaining the quantitative map based on the matching result.

[0078] Specifically, firstly, based on the detected tissue images at each time point in the image time series, the evolution curves of the quantitative parameters of each voxel are obtained. Then, the acquired evolution curves are compared with the evolution curves in the dictionary to determine the evolution curve with the highest similarity in the dictionary. This evolution curve best matches the observed signal evolution and is used as the corresponding quantitative parameter evolution curve. Finally, the quantitative map is generated based on all determined quantitative parameter evolution curves. It should be noted that the dictionary is obtained through simulation based on the acquired parameters and different initial states of the magnetization vector, and is a collection of evolution curves. Optionally, the dictionary can be generated using the Bloch equation.

[0079] In summary, the magnetic resonance fingerprinting method provided in this embodiment obtains quantitative spectra using dictionary matching. This is an exhaustive search method that guarantees global maxima within the simulation range, and it is relatively accurate and robust against highly aliased artifacts. For details not covered in this embodiment, please refer to the descriptions in other embodiments.

[0080] <Example 6>

[0081] Please see Figure 5 This embodiment provides a magnetic resonance fingerprinting method, including: based on the methods provided in Embodiments 1 and 5, before matching the image time series with a dictionary and obtaining the quantitative map according to the matching result, compressing the dictionary using the transformation matrix to generate the compressed dictionary; and when matching the image time series with the dictionary, matching the image time series with the compressed dictionary and obtaining the quantitative map according to the matching result.

[0082] It is understood that using the compressed original data for image reconstruction reduces the computational load, significantly shortens the computation time, and improves imaging efficiency. Furthermore, further dictionary compression can shorten dictionary matching time, thereby further reducing imaging time and improving imaging efficiency. For details not covered in this embodiment, please refer to the descriptions in other embodiments.

[0083] <Example 7>

[0084] Please see Figure 6 This embodiment provides a magnetic resonance fingerprint imaging method, including: based on the method provided in Embodiment 1, the process of generating a quantitative map based on the image time series includes: training a machine learning model using a dictionary, or training the machine learning model using self-supervised learning; inputting the image time series into the machine learning model to generate the quantitative map. It is understood that the quantitative map acquisition in this embodiment does not require dictionary matching; it only requires inputting the image time series into a preset machine learning model, which can autonomously determine and generate the corresponding quantitative map, eliminating the need for dictionary generation and matching, further shortening the imaging time and improving imaging efficiency.

[0085] Furthermore, to obtain an accurate quantitative map, the machine learning model needs to be trained on a large amount of data before acquiring the quantitative map to improve its judgment accuracy. Preferably, the machine learning model is trained using self-supervised learning based on online Bloch / EPG simulation, or using the dictionary as a database. Further, the machine learning model includes, but is not limited to, various commonly used deep learning models such as fully connected networks, U-net, V-net, and ResNet, or traditional machine learning models such as Kalman filters and decision trees. For details not covered in this embodiment, please refer to the descriptions in other embodiments.

[0086] <Example 8>

[0087] Please see Figure 7 This embodiment provides an imaging device, including: an information acquisition unit 100, a data compression unit 101, an image reconstruction unit 102, and a quantitative imaging unit 103. The information acquisition unit 100 is used to acquire K-space data at each preset time point as raw data. The data compression unit 101 is used to compress the raw data using a transformation matrix to generate compressed data. The image reconstruction unit 102 is used to reconstruct images from the compressed data to obtain an image time series. The quantitative imaging unit 103 is used to generate a quantitative map based on the image time series. That is, the imaging device executes any one of the magnetic resonance fingerprint imaging methods described in Embodiments 1 to 7 through each unit.

[0088] Furthermore, the data compression unit 101 is used to compress the original data in the time dimension, so that the time dimension of the original data is greater than the time dimension of the compressed data. The imaging device also includes a data reconstruction unit 104, used to reconstruct the original data according to the sampling trajectory types in K-space before compressing the original data using a transformation matrix to generate compressed data, so that each preset time point in the reconstructed original data corresponds to a data segment of the corresponding type of sampling trajectory; and to assign the K-space data collected at each preset time point to the data segment of the corresponding type of sampling trajectory, and to assign 0 to the data segment of the sampling trajectory of types not collected at each preset time point. Based on this, the data compression unit 101 uses a transformation matrix to compress the reconstructed original data to generate the compressed data; wherein the number of data segments in the compressed data is the same as the number of data segments in the reconstructed original data.

[0089] Optionally, the imaging device further includes a dictionary generation unit 105 and a dictionary matching unit 106. The dictionary generation unit 105 generates a dictionary for the image time series. The dictionary matching unit 106 matches the image time series with the dictionary to obtain a matching result. Further, the data compression unit 101 compresses the dictionary using the transformation matrix to generate a compressed dictionary. Therefore, the dictionary matching unit 106 matches the compressed dictionary with the image time series. The quantitative imaging unit 103 obtains the quantitative map based on the matching result. Alternatively, the imaging device does not require the dictionary matching unit 106; the quantitative imaging unit 103 has a machine learning model, and the image time series is input into the machine learning model to generate the quantitative map. Further, the quantitative imaging unit 103 is also used to train the machine learning model using the dictionary generated by the dictionary generation unit 105, or to train the machine learning model using self-supervised learning. It should be noted that during the training of the machine learning model using the dictionary, it is not necessary to pre-compress the dictionary using the data compression unit 101.

[0090] In summary, the imaging device provided in this embodiment can compress the reconstructed original data in the time dimension, thereby reducing the data volume while ensuring data integrity. Furthermore, using compressed data for image reconstruction significantly reduces the computational load during the image reconstruction process, effectively improving image reconstruction speed and thus enhancing the efficiency of magnetic resonance fingerprint imaging. For any aspects not detailed in this embodiment, please refer to the descriptions in other embodiments.

[0091] <Example 9>

[0092] This embodiment provides a computer storage medium. The computer storage medium stores executable instructions, which, when executed by a processor, cause the processor to perform the steps in the magnetic resonance fingerprinting method.

[0093] In summary, the computer storage medium provided in this embodiment compresses data before image reconstruction when executing instructions. The subsequent image reconstruction calculation process is based on the compressed data, thereby significantly reducing the computational load during image reconstruction and effectively improving the speed of magnetic resonance fingerprint imaging. For details not covered in this embodiment, please refer to the descriptions in other embodiments.

[0094] <Example 10>

[0095] This embodiment provides a magnetic resonance imaging (MRI) device, including the imaging device and / or the computer storage medium. In other words, the MRI device may include only the imaging device, or only the computer storage medium, or both the imaging device and the computer storage medium.

[0096] In summary, the magnetic resonance imaging device provided in this embodiment compresses data before image reconstruction during operation. The subsequent image reconstruction calculation process is based on the compressed data, thereby significantly reducing the computational load during image reconstruction and effectively improving the speed of magnetic resonance fingerprint imaging. For any details not covered in this embodiment, please refer to the descriptions in other embodiments.

[0097] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to mutually. In addition, different parts between embodiments can also be combined with each other, and this invention does not limit this.

[0098] Furthermore, it should be understood that although the present invention has been disclosed above with reference to preferred embodiments, these embodiments are not intended to limit the present invention. For any person skilled in the art, many possible variations and modifications can be made to the technical solutions of the present invention based on the disclosed technical content, or equivalent embodiments can be modified accordingly, without departing from the scope of the present invention. Therefore, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention, without departing from the content of the present invention, shall still fall within the scope of protection of the present invention.

Claims

1. A magnetic resonance fingerprint imaging method, characterized in that, include: Collect K-space data at each preset time point as the raw data; The data dimension of the original data is increased according to the sampling trajectory type in K-space to reconstruct the original data, so that each preset time point in the reconstructed original data corresponds to a data segment of the sampling trajectory of the corresponding type; and the K-space data collected at each preset time point is assigned to the data segment of the sampling trajectory of the corresponding type, and the data segment of the sampling trajectory of the type not collected at each preset time point is assigned 0. The original data, after being compressed and reconstructed using a transformation matrix, is used to generate compressed data; The compressed data is used to reconstruct images to obtain image time series. A quantitative map is generated based on the image time series; Wherein, the time dimension of the original data is greater than the time dimension of the compressed data; and, after compressing the time dimension of each piece of original data, the number of data segments in the compressed data is the same as the number of data segments in the reconstructed original data.

2. The magnetic resonance fingerprint imaging method according to claim 1, characterized in that, The process of generating a quantitative map based on the image time series includes: matching the image time series with a dictionary, and obtaining the quantitative map based on the matching result.

3. The magnetic resonance fingerprint imaging method according to claim 2, characterized in that, Before matching the image time series with the dictionary and obtaining the quantitative map based on the matching result, the dictionary is compressed using the transformation matrix to generate the compressed dictionary; Furthermore, when matching the image time series with the dictionary, the image time series is matched with the compressed dictionary, and the quantitative map is obtained based on the matching result.

4. The magnetic resonance fingerprint imaging method according to claim 1, characterized in that, The process of generating a quantitative map based on the image time series includes: A machine learning model is trained using a dictionary, or the machine learning model is trained using self-supervised learning. The image time series is input into the machine learning model to generate the quantitative map.

5. An imaging device, characterized in that, The device is used to perform the magnetic resonance fingerprint imaging method as described in any one of claims 1 to 4; wherein the imaging device includes: an information acquisition unit (100), a data compression unit (101), an image reconstruction unit (102), and a quantitative imaging unit (103). The information acquisition unit (100) is used to acquire K-space data at each preset time point as raw data; The data compression unit (101) is used to compress the original data using a transformation matrix to generate compressed data; The image reconstruction unit (102) is used to reconstruct the image from the compressed data to obtain an image time series; The quantitative imaging unit (103) is used to generate a quantitative map based on the image time series.

6. A computer storage medium, characterized in that, The computer storage medium stores executable instructions, which, when executed by a processor, cause the processor to perform the steps of the magnetic resonance fingerprinting method according to any one of claims 1 to 4.

7. A magnetic resonance imaging device, characterized in that, It includes the imaging device of claim 5, and / or the computer storage medium of claim 6.

Images