Image reconstruction method, magnetic resonance imaging method and computer device
By fitting and restoring the K-space calibration dataset using full-sampled and undersampled datasets in parallel magnetic resonance imaging, the problems of artifacts and quality degradation in image reconstruction are solved, and higher quality image reconstruction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANGHAI UNITED IMAGING HEALTHCARE
- Filing Date
- 2022-06-30
- Publication Date
- 2026-05-29
AI Technical Summary
In parallel magnetic resonance imaging, the reconstructed image quality is poor. Existing techniques suffer from artifacts during image reconstruction, and the image quality is reduced due to the mismatch between the full-sample calibration data and the undersampled data.
By acquiring the K-space calibration dataset of the target region and performing full sampling, a first fitting recovery is performed in the central region of the K-space for each undersampled K-space dataset to construct a data recovery matrix and supplement the unsampled point data; then a second fitting recovery is performed on the non-central region to generate the target K-space dataset.
It improves the quality of reconstructed images, reduces artifacts, ensures high data matching, and generates images that better reflect the actual data obtained from full sampling.
Smart Images
Figure CN117368817B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of magnetic resonance imaging technology, and in particular to an image reconstruction method, a magnetic resonance imaging method, and a computer device. Background Technology
[0002] Parallel magnetic resonance imaging (MRI) technology uses a multi-coil array to simultaneously acquire K-space data, allowing for undersampling of the K-space to reduce the number of phase encoding steps. This significantly shortens the MRI scan time and improves imaging speed while maintaining the image spatial resolution.
[0003] In related technologies, image reconstruction based on K-space data utilizes small-area fully sampled calibration lines in the central region of the acquired K-space data as a reference for restoring unsampled data, thereby synthesizing complete K-space data. To accelerate the acquisition process or due to limitations such as sequence design, a single calibration data set is often shared when reconstructing multiple magnetic resonance images.
[0004] However, the above methods can lead to artifacts in the reconstructed image, resulting in poor image quality. Summary of the Invention
[0005] Therefore, it is necessary to provide an image reconstruction method, a magnetic resonance imaging method, and a computer device that can improve the quality of reconstructed images in parallel magnetic resonance imaging, in order to address the aforementioned technical problems.
[0006] In a first aspect, this application provides an image reconstruction method, the method comprising:
[0007] Obtain the K-space calibration dataset corresponding to the target part. The K-space calibration dataset is fully sampled in the central region of K-space.
[0008] Multiple undersampled K-space datasets corresponding to the target part are obtained, and each undersampled K-space dataset contains data collected from the target part in one excitation.
[0009] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0010] A second fitting is performed on the unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets to recover multiple target K-space datasets;
[0011] Reconstruct multiple target K-space datasets and obtain the corresponding magnetic resonance images of the target regions after multiple excitations.
[0012] In one embodiment, based on the K-space calibration dataset, a first fitting recovery is performed on the unsampled points in the central region of the K-space for each undersampled K-space dataset, including:
[0013] For any undersampled K-space dataset, construct a data recovery matrix based on the K-space calibration dataset and the undersampled K-space dataset;
[0014] Based on the data recovery matrix corresponding to each undersampled K-space dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each sampled K-space dataset.
[0015] In one embodiment, a data recovery matrix is constructed based on the K-space calibration dataset and the undersampled K-space dataset, including:
[0016] Based on the K-space calibration dataset, the first low-rank matrix is constructed using a pre-defined low-rank matrix construction method;
[0017] Based on the undersampled K-space dataset, a second low-rank matrix is constructed using a low-rank matrix construction method;
[0018] Generate the data recovery matrix based on the first low-rank matrix and the second low-rank matrix.
[0019] In one embodiment, the process of constructing a low-rank matrix includes:
[0020] Extract a predetermined number of distinct first data points from the target dataset and obtain the coordinate information of each first data point; the target dataset is a K-space calibration dataset or an undersampled K-space dataset.
[0021] For any given first data point, obtain multiple second data points that are less than a preset length away from the first data point, and obtain the set of data points corresponding to the first data point;
[0022] Obtain the signal values of multiple second data points in each data point set;
[0023] A low-rank matrix is constructed based on the coordinate information of each first data point and the signal values of multiple second data points in the data point set corresponding to each first data point.
[0024] In one embodiment, a second fitting recovery is performed on unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets, including:
[0025] Based on the fitted full-sample data of each intermediate K-space dataset in the central region of K-space, calculate the weight kernel of the unsampled points in the non-central region of K-space for each intermediate K-space dataset.
[0026] For any intermediate K-space dataset, a second fitting recovery is performed on the unsampled points in the non-central region based on the weight kernel of the unsampled points.
[0027] Secondly, this application also provides a magnetic resonance imaging method, the method comprising:
[0028] A partial sampling technique is used to fill the central region of the K-space to obtain the K-space calibration dataset corresponding to the target part;
[0029] The target area is excited multiple times, and the undersampled K-space dataset corresponding to each excitement is collected;
[0030] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0031] A second fitting is performed on the unsampled points in the non-central region of the K-space from multiple intermediate K-space datasets to recover multiple target K-space datasets.
[0032] Reconstruct multiple target K-space datasets and obtain the corresponding magnetic resonance images of the target regions after multiple excitations.
[0033] In one embodiment, the direction of the diffusion gradient applied in each of the multiple excitations is different.
[0034] In one embodiment, the physiological phase of the target site is different for each of the multiple stimulations.
[0035] In one embodiment, a labeling pulse is applied to the target site during each of the multiple excitations, and an undersampled K-space dataset is acquired at different delay times after the labeling pulse is applied.
[0036] Thirdly, this application also provides an image reconstruction apparatus, which includes:
[0037] The calibration data acquisition module is used to acquire the K-space calibration dataset corresponding to the target part. The K-space calibration dataset is fully sampled in the central region of K-space.
[0038] The undersampled data acquisition module is used to acquire multiple undersampled K-space datasets corresponding to the target part. Each undersampled K-space dataset contains data collected from the target part in one excitation.
[0039] The first data recovery module is used to perform a first fitting recovery on the unsampled points in the central region of the K space of each undersampled K space dataset based on the K space calibration dataset, so as to obtain multiple intermediate K space datasets.
[0040] The second data recovery module is used to perform a second fitting recovery on the unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets to obtain multiple target K-space datasets.
[0041] The image reconstruction module is used to reconstruct multiple target K-space datasets and obtain the corresponding magnetic resonance images of the target parts after multiple excitations.
[0042] Fourthly, this application also provides a magnetic resonance imaging device, which includes:
[0043] The data acquisition module is used to fill the central region of the K-space using partial sampling techniques to obtain the K-space calibration dataset corresponding to the target part.
[0044] The scanning module is used to excite the target area multiple times and collect the undersampled K-space dataset corresponding to each excitation;
[0045] The first data recovery module is used to perform a first fitting recovery on the unsampled points in the central region of the K space of each undersampled K space dataset based on the K space calibration dataset, so as to obtain multiple intermediate K space datasets.
[0046] The second data recovery module is used to perform a second fitting recovery on the unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets, thereby obtaining multiple target K-space datasets.
[0047] The imaging module is used to reconstruct multiple target K-space datasets and acquire magnetic resonance images corresponding to multiple excitations of the target area.
[0048] Fifthly, this application also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the method embodiments of the first and second aspects described above.
[0049] In a sixth aspect, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the method embodiments in the first and second aspects described above.
[0050] In a seventh aspect, this application also provides a computer program product comprising a computer program that, when executed by a processor, implements the steps of any of the method embodiments in the first and second aspects described above.
[0051] In the aforementioned image reconstruction method, magnetic resonance imaging method, and computer device, firstly, a K-space calibration dataset corresponding to the target region and multiple undersampled K-space datasets corresponding to the target region are acquired. The K-space calibration dataset is fully sampled in the central region of K-space, and each undersampled K-space dataset contains data collected from the target region during a single excitation. Then, based on the K-space calibration dataset, a first fitting recovery is performed on the unsampled points in the central region of each undersampled K-space dataset to obtain multiple intermediate K-space datasets. Further, a second fitting recovery is performed on the unsampled points in the non-central regions of the multiple intermediate K-space datasets to obtain multiple target K-space datasets. Finally, the multiple target K-space datasets are reconstructed to obtain the magnetic resonance images corresponding to multiple excitations of the target region. That is, this application does not directly use the K-space calibration dataset as a benchmark to fit and recover the unsampled data in the undersampled K-space dataset. Instead, based on the K-space calibration dataset, a first fitting recovery is performed on the unsampled points in the central region of each undersampled K-space dataset, and the fully sampled data in the central region can be obtained through this first fitting recovery. Furthermore, using the fully sampled data of the central region obtained through the first fitting restoration as a benchmark, a second fitting restoration is performed on the unsampled data points in the non-central regions to obtain the target K-space dataset. Thus, using the intermediate K-space dataset obtained after the first fitting restoration of the unsampled points in the central region of K-space as a benchmark ensures that the fully sampled data in the central region of the intermediate K-space dataset and the unsampled points in the non-central regions of the undersampled K-space dataset belong to the same excitation, resulting in a high data matching degree. The target K-space dataset after the second fitting restoration also better matches the actual fully sampled data, leading to better quality magnetic resonance images of the target area generated based on the target K-space dataset. Attached Figure Description
[0052] Figure 1 This is a flowchart illustrating an image reconstruction method in one embodiment;
[0053] Figure 2 This is a schematic diagram of a K-space standard dataset in one embodiment;
[0054] Figure 3 This is a schematic diagram illustrating the fitting and recovery of data in the central region of an undersampled K-space dataset in one embodiment.
[0055] Figure 4 This is a flowchart illustrating the first fitting recovery operation in one embodiment;
[0056] Figure 5 This is a schematic diagram illustrating the construction of a data recovery matrix in one embodiment;
[0057] Figure 6This is a flowchart illustrating the second fitting recovery operation in one embodiment;
[0058] Figure 7 This is a schematic flowchart of a magnetic resonance imaging method in one embodiment;
[0059] Figure 8 This is a structural block diagram of an image reconstruction apparatus in one embodiment;
[0060] Figure 9 This is a structural block diagram of a magnetic resonance imaging device in one embodiment;
[0061] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0063] Magnetic resonance imaging (MRI) is a technique that uses the principle of nuclear magnetic resonance to perform tomographic imaging of the human body. It can provide various images of human soft tissues and has rapidly developed into an important application technology in biomedicine. The imaging process is completely free of radioactive contamination, offers high resolution, and can image at any level. Moreover, unlike existing imaging techniques, MRI involves more factors, resulting in a large amount of image information, giving it significant advantages and application potential in medical diagnosis.
[0064] However, in some clinical applications, in addition to requiring high image spatial resolution, it is also necessary to reduce imaging time to minimize motion artifacts. For example, in real-time imaging of the cardiovascular system and brain function imaging, the effects of factors such as cardiac motion, respiration, and blood flow must be considered. Magnetic resonance imaging (MRI) encodes the spatial information of Fourier images through gradient fields. Acquiring a complete image requires continuous gradient spatial encoding, and imaging speed is highly dependent on the performance of the gradient system of the MRI equipment, such as the intensity and switching rate of the gradient field. To meet the need for rapid imaging, the performance of the gradient field has been greatly enhanced, but this has also created new problems, such as the increasing cost of gradient hardware systems. In addition, excessively high gradient field switching rates can cause neuromuscular electromagnetic stimulation. The improvement of imaging speed dependent on gradient field performance has reached its limit, and clinicians are looking for more effective methods to improve imaging speed.
[0065] To improve imaging speed, parallel magnetic resonance imaging (parallel MRI) uses multiple phased array coils to receive induction signals simultaneously. This reduces the number of gradient encoding steps, thereby significantly shortening the scan time and improving imaging speed.
[0066] Specifically, parallel magnetic resonance imaging uses small-scale full-sample calibration lines acquired in the central region of K-space as a benchmark for recovering unacquired data. It calculates a weighting kernel that can fit the unacquired data and applies this weighting kernel to the undersampled data to be reconstructed in the next data synthesis process to synthesize complete K-space data.
[0067] In this process, accurate full-sample calibration data is crucial for image quality. If the acquired full-sample calibration data does not match the undersampled data to be reconstructed, or if the quality of the full-sample calibration data is poor, errors will occur in the calculated weight kernel. These errors will be introduced into the subsequent data synthesis stage, resulting in image quality degradation such as artifacts or blurring in the reconstructed image.
[0068] In some sequential scanning applications, to accelerate data acquisition or due to limitations such as sequence design, a single full-sample calibration data set is often needed when reconstructing multiple MRI images. However, due to the movement of the scanned object during the scanning process or other factors, a deviation can occur between the aforementioned full-sample calibration data and the undersampled data obtained from the scan, thus affecting the quality of the reconstructed image.
[0069] Based on this, this application provides an image reconstruction method that constructs corresponding target calibration data based on the fully sampled calibration data obtained before scanning, and uses the target calibration data corresponding to each undersampled K-space dataset collected for scanning the target part of the object each time. The unsampled data is then restored using the target calibration data corresponding to each undersampled K-space dataset to obtain complete K-space data, thereby improving the quality of the reconstructed image.
[0070] The image reconstruction method provided in this application can be applied to an image reconstruction device, which can be implemented in software and / or hardware. The device can be integrated into a computer device with medical image processing capabilities, such as the imaging device in a magnetic resonance system, or any computer device outside of a magnetic resonance system.
[0071] The imaging equipment in the magnetic resonance system is used to fill the K-space with magnetic resonance signals and reconstruct the image based on the K-space data to obtain the target magnetic resonance image. The computer equipment outside the magnetic resonance system can be any terminal or server. The terminal can include, but is not limited to, software running on a physical device, such as applications or clients installed on the device, or personal computers, laptops, smartphones, tablets, and portable wearable devices with applications installed. The server can include, but is not limited to, at least one independent server, a distributed server, a cloud server, and a server cluster.
[0072] The technical solutions of the embodiments of this application and how the technical solutions of the embodiments of this application solve the above-mentioned technical problems will be described in detail below with reference to the embodiments and accompanying drawings. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. It should be noted that the image reconstruction method provided in the embodiments of this application can be executed by a magnetic resonance imaging device, a computer device, or the image reconstruction apparatus provided in this application. Obviously, the described embodiments are only some embodiments of the embodiments of this application, and not all embodiments.
[0073] In one embodiment, such as Figure 1 As shown, an image reconstruction method is provided, and the method is illustrated using a computer device as an example. The method includes the following steps:
[0074] Step 110: Obtain the K-space calibration dataset corresponding to the target part. The K-space calibration dataset is fully sampled in the central region of K-space.
[0075] The target area can be any part of the object to be scanned. If the object is a human body, the target area can be the head, chest, abdomen, etc. Of course, the object can also be other living organisms, and this embodiment does not limit this.
[0076] It should be noted that the K-space calibration dataset can be obtained during the pre-scanning process, the localization process, or the sequence scanning process. This embodiment does not limit the timing of acquisition.
[0077] Furthermore, for image reconstruction, reducing scanning time is typically achieved by reducing the number of points used. Therefore, in this embodiment, when acquiring the K-space calibration dataset, partial sampling can be used, performing full sampling only on the central region of the K-space that determines image contrast. For the non-central regions of the K-space, i.e., the peripheral regions, undersampling can be performed, such as... Figure 2 As shown in (a); sampling can also be omitted, such as Figure 2As shown in (b).
[0078] It should be understood that, Figure 2 (b) is a dataset obtained by undersampling in a non-central region of K space using an interlaced method. The undersampling interval can still be multiple rows, and there is no restriction on this.
[0079] Step 120: Obtain multiple undersampled K-space datasets corresponding to the target area. Each undersampled K-space dataset contains data collected from the target area during a single excitation.
[0080] In this context, multiple undersampled K-space datasets are acquired during dynamic or multi-phase scans. For example, in blood oxygen level dependent functional imaging (BOLD fMRI) applications, due to data acquisition time constraints and the requirements of echo-planar imaging (EPI) phase encoding, a K-space calibration dataset is pre-acquired during parallel imaging, followed by the acquisition of multiple undersampled K-space datasets.
[0081] Step 130: Based on the K-space calibration dataset, perform a first fitting recovery on the unsampled points in the central region of the K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets.
[0082] Among them, the intermediate K-space dataset after the first fitting recovery is equivalent to full sampling in the central region of K-space compared to the undersampled K-space dataset.
[0083] That is, by using the full sampled data of the K-space calibration dataset in the central region of K-space, the unsampled points of the undersampled K-space dataset in the central region of K-space are first fitted and recovered to fit the data values of the unsampled points, so as to obtain the full sampled data of the undersampled K-space dataset in the central region of K-space.
[0084] In one possible implementation, the first fitting recovery can leverage the low-rank property by constructing a low-rank matrix and obtaining the null space matrix to determine the intermediate K-space dataset corresponding to each undersampled K-space dataset. This intermediate K-space dataset serves as the benchmark for recovering the unsampled data points in the undersampled K-space dataset.
[0085] It should be noted that the center region location / size is the same in the standard K-space dataset and each undersampled K-space dataset. In other words, based on the location of the center region of the fully sampled K-space dataset in the K-space calibration dataset, the first fitting recovery is performed on the unsampled points at the same location in each sampled K-space dataset.
[0086] As an example, see Figure 3The standard K-space dataset has a central region of size 3*3 in K-space. Therefore, the data range for performing the first fitting recovery in the undersampled K-space dataset is also the unsampled points within the central region of K-space of 3*3.
[0087] Step 140: Perform a second fitting recovery on the unsampled points in the non-central region of the K-space for multiple intermediate K-space datasets to obtain multiple target K-space datasets.
[0088] It should be understood that the phase coding lines filling the central region of K-space primarily determine the image contrast, while the phase coding lines in the surrounding regions primarily determine the anatomical details of the image. The phase coding lines on either side of the zero Fourier line are mirror-symmetric. That is, K-space exhibits mirror-symmetry in both the phase coding direction and the frequency coding direction.
[0089] Based on this, after recovering all the data in the central region of the undersampled K-space dataset through the first fitting, the target K-space dataset under the full sampling condition can be obtained through the second fitting based on the mirror symmetry property of the K-space filled data.
[0090] In one possible implementation, when reconstructing an image based on the K-space domain, the first fitting recovery can be achieved using the Simultaneous Acquisition of Spatial Harmonics (SMASH) reconstruction method, the Generalized Auto-calibrating Partially Parallel Acquisition (GRAPPA) reconstruction method, or the Sensitivity Encoding (SENSE) reconstruction method.
[0091] Specifically, the basic idea of the SMASH reconstruction method is to recover the K-space phase-encoded row data lost due to undersampling by linearly combining the indices of the receiving coils. In contrast, GRAPPA does not fit the data of each array coil to the combined signal, but rather to the ACS row of a single coil, thereby obtaining a series of linear weights to reconstruct the missing K-space data rows of each array coil.
[0092] Step 150: Reconstruct multiple target K-space datasets and obtain the corresponding magnetic resonance images of the target parts after multiple excitations.
[0093] In this step, image reconstruction is performed on each target K-space dataset to obtain the corresponding magnetic resonance image. Specifically, image reconstruction involves performing a Fourier transform on the target K-space dataset to generate a reconstructed image. The reconstructed images from multiple target K-space datasets are the magnetic resonance images obtained after multiple scans of the scanned area.
[0094] In the aforementioned image reconstruction method, based on the K-space calibration dataset, a first fitting restoration is performed on the unsampled points in the central region of each undersampled K-space dataset. This first fitting restoration yields the fully sampled data for the central region. Further, using the fully sampled data obtained from the first fitting restoration as a benchmark, a second fitting restoration is performed on the unsampled data points in the non-central regions to obtain the target K-space dataset. Thus, by using the intermediate K-space dataset obtained after the first fitting restoration of the unsampled points in the central region of K-space as a benchmark, it can be ensured that the fully sampled data in the central region of the intermediate K-space dataset and the unsampled points in the non-central regions of the undersampled K-space dataset belong to the same excitation, resulting in a high data matching degree. The target K-space dataset after the second fitting restoration also better matches the actual fully sampled data, leading to better quality magnetic resonance images of the target area generated based on the target K-space dataset.
[0095] In one embodiment, such as Figure 4 As shown, in step 130 above, based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of the K-space for each undersampled K-space dataset, including the following steps:
[0096] Step 410: For any undersampled K-space dataset, construct a data recovery matrix based on the K-space calibration dataset and the undersampled K-space dataset.
[0097] In one possible implementation, step 410 can be implemented as follows: based on the K-space calibration dataset, a first low-rank matrix is constructed using a preset low-rank matrix construction method; based on the undersampled K-space dataset, a second low-rank matrix is constructed using a low-rank matrix construction method; and a data recovery matrix is generated based on the first low-rank matrix and the second low-rank matrix.
[0098] The process of constructing a low-rank matrix is as follows: extract a predetermined number of different first data points from the target dataset and obtain the coordinate information of each first data point; for any first data point, obtain multiple second data points whose distance from the first data point is less than a predetermined length, and obtain the data point set corresponding to the first data point; obtain the signal values of multiple second data points in each data point set; construct a low-rank matrix based on the coordinate information of each first data point and the signal values of multiple second data points in the data point set corresponding to each first data point.
[0099] That is, the first low-rank matrix and the second low-rank matrix are constructed based on the same low-rank matrix construction method, and the target dataset mentioned above is a K-space calibration dataset or an undersampled K-space dataset.
[0100] As an example, the construction operation of a low-rank matrix is denoted as P. c (·), the construction steps are as follows:
[0101] (1) Randomly select L distinct first data points from the K-space center region of the target dataset, where k represents the encoding index of the selected first data point, 1≤k≤L, n x n y Let x and y represent the x-coordinate and y-coordinate of a first data point, respectively. Then the signal values of these first data points can be expressed as:
[0102] Furthermore, for each first data point selected from the target dataset, select the data point that is the first data point. Other N with a distance within radius R R A set of N second data points. R The coordinate index of the second data point is used express.
[0103] Where m = 1, 2, ..., N R Its corresponding signal value is Usually L≥N R .
[0104] (2) Arrange the following formula (1) to form a low-rank C matrix:
[0105]
[0106] In the formula, k = 1, 2, ..., L, m = 1, 2, ..., N R , This represents the x-coordinate of the m-th nearest neighbor of the k-th point. Let represent the ordinate of the m-th nearest neighbor of the k-th point. Then the size of the C(k, m) matrix is L×N. R .
[0107] See Figure 5 The process of generating the data recovery matrix in step 410 can be as follows: Partial data k from the central region of the K-space calibration dataset. acs The first low-rank matrix P is constructed using the aforementioned low-rank matrix construction method. c (k acs Meanwhile, from the undersampled K-space dataset Q i Take the data k at the corresponding position. i The second low-rank matrix P is constructed using the aforementioned low-rank matrix construction method. c (k i ); then, based on the first low-rank matrix and the second low-rank matrix, a data recovery matrix [P] is generated. c (k acs )Pc (k i )).
[0108] Among them, the undersampled K-space dataset Q i It is any one of multiple undersampled K-space datasets.
[0109] Step 420: Based on the data recovery matrix corresponding to each undersampled K-space dataset, perform the first fitting recovery on the unsampled points in the central region of K-space for each sampled K-space dataset.
[0110] In one possible implementation, step 420 can be performed by: finding the null space matrix N of the data recovery matrix, and N... H N = I, which means that the collected K-space calibration dataset is the full sampled data in the central region of K-space and the Q-th... i The data corresponding to each undersampled K-space dataset are combined to estimate the data of the unsampled points in the central region of the K-space for each undersampled K-space dataset.
[0111] Specifically, the problem of formulating the solution to the requirements is:
[0112]
[0113] Where η represents the proportion of the fully sampled data in the K-space calibration dataset. The larger η is, the more similar the fully sampled data in the central region of the K-space of the fitted and recovered undersampled K-space dataset is to the fully sampled data in the K-space calibration dataset.
[0114] In this embodiment, by using the K-space calibration dataset, a first fitting recovery is performed on the unsampled points in the central region of the K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets. This avoids the mismatch between the K-space calibration dataset and the undersampled K-space dataset. By determining the intermediate K-space dataset corresponding to the full sampling of the central region of each undersampled K-space dataset, potential artifacts or reduced reconstructed image quality can be mitigated.
[0115] Based on the above method embodiments, by using the K-space calibration dataset, the unsampled points in the central region of the K-space for each undersampled K-space dataset are first-fitted and restored to obtain multiple intermediate K-space datasets. The intermediate K-space datasets are obtained by supplementing the unsampled points in the central region of the corresponding undersampled K-space dataset, thus achieving full sampling of the central region of the K-space for the intermediate K-space datasets.
[0116] Furthermore, for each intermediate K-space dataset, using the complete data of its central region as a benchmark, the unsampled points in the non-central region of the K-space can be fitted and restored.
[0117] In one embodiment, such as Figure 6 As shown, step 140 above involves performing a second fitting recovery on the unsampled points in the non-central region of the multiple intermediate K-space datasets, which includes the following steps:
[0118] Step 610: Using the fitted full-sample data of each intermediate K-space dataset in the central region of K-space as a benchmark, calculate the weight kernel of the unsampled points in the non-central region of K-space for each intermediate K-space dataset.
[0119] Specifically, for unsampled points in the non-central region of K-space, the weight of each sampled point in the central region can be calculated based on the full sampled data of the central region when recovering the unsampled points in the non-central region. In other words, it is necessary to calculate the contribution value of each full sampled data point in the central region of K-space to the recovery of the unsampled points in the non-central region.
[0120] As an example, the weight kernel can be calculated using the GRAPPA reconstruction method, which will not be elaborated here.
[0121] Step 620: For any intermediate K-space dataset, perform a second fitting recovery on the unsampled points in the non-center region based on the weight kernel of the unsampled points.
[0122] In other words, by applying the weight kernels calculated from the fully sampled data fitted to the central region of each intermediate K-space dataset to the unsampled points in the non-central region, the complete K-space data can be obtained through GRAPPA reconstruction. This improves the data quality of the target K-space dataset obtained after the second fitting restoration and better suppresses artifacts that may exist in the image.
[0123] Based on the same inventive concept, this application also provides a magnetic resonance imaging method, such as... Figure 7 As shown, the method is illustrated using a computer device as an example. This computer device can specifically be one or more devices within a magnetic resonance imaging system, and includes the following steps:
[0124] Step 710: Use partial sampling techniques to fill the central region of the K-space to obtain the K-space calibration dataset corresponding to the target part.
[0125] In other words, before performing multiple scans on the target area using a set number of excitations, a K-space calibration dataset is acquired during the pre-scanning or positioning process.
[0126] It should be understood that each magnetic resonance signal acquired during the pre-scanning process contains information from the entire layer. Therefore, spatial localization encoding, i.e., frequency encoding and phase encoding, is required for the magnetic resonance signal. The magnetic resonance signal acquired by the receiving coil in the magnetic resonance scanning equipment is actually a radio wave with spatially encoded information. It is an analog signal rather than digital information and needs to be converted into digital information through analog-to-digital conversion (ADC). The digital information is then filled into K-space to obtain a K-space digital dot matrix.
[0127] In this step, obtaining the K-space calibration dataset by undersampling means that the central region of the K-space is fully sampled, while the non-central regions are undersampled.
[0128] Step 720: Excite the target area multiple times and collect the undersampled K-space dataset corresponding to each excitation.
[0129] The number of excitations is determined by a pre-set sequence number; this embodiment does not limit the number of excitations. Each excitation yields an undersampled K-space dataset that can be used to generate a magnetic resonance image.
[0130] Step 730: Based on the K-space calibration dataset, perform a first fitting recovery on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets.
[0131] Step 740: Perform a second fitting recovery on the unsampled points in the non-central region of the K-space for multiple intermediate K-space datasets to obtain multiple target K-space datasets.
[0132] Step 750: Reconstruct multiple target K-space datasets and obtain the corresponding magnetic resonance images of the target parts after multiple excitations.
[0133] Based on the above magnetic resonance imaging method, multiple excitations of the target area can include any of the following settings:
[0134] (1) The direction of the diffusion gradient applied in each excitation is different in multiple excitations.
[0135] This setting can be used by magnetic resonance imaging equipment to perform diffusion tensor imaging (DTI) and diffusion-weighted imaging (DWI) on target areas.
[0136] (2) The physiological phase of the target site is different for each stimulation in multiple stimulations.
[0137] (3) In each of the multiple excitations, a labeling pulse is applied to the target area, and undersampled K-space datasets are collected at different delay times after the labeling pulse is applied.
[0138] This setting can be used by magnetic resonance imaging equipment to perform arterial spin labeling (ASL) on target sites.
[0139] It should be noted that the implementation principle and technical effect of the image reconstruction step in the magnetic resonance imaging method provided in this embodiment are similar to those in the previous image reconstruction method embodiments. For specific limitations and explanations, please refer to the previous method embodiments, which will not be repeated here.
[0140] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0141] Based on the same inventive concept, this application also provides an image reconstruction apparatus for implementing the image reconstruction method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more image reconstruction apparatus embodiments provided below can be found in the limitations of the image reconstruction method described above, and will not be repeated here.
[0142] In one embodiment, such as Figure 8 As shown, an image reconstruction apparatus 800 is provided, comprising: a calibration data acquisition module 810, an undersampling data acquisition module 820, a first data recovery module 830, a second data recovery module 840, and an image reconstruction module 850, wherein:
[0143] The calibration data acquisition module 810 is used to acquire the K-space calibration dataset corresponding to the target part. The K-space calibration dataset is fully sampled in the central region of the K-space.
[0144] The undersampled data acquisition module 820 is used to acquire multiple undersampled K-space datasets corresponding to the target part, and each undersampled K-space dataset is the data collected by the target part in one excitation.
[0145] The first data recovery module 830 is used to perform a first fitting recovery on the unsampled points in the central region of the K space of each undersampled K space dataset based on the K space calibration dataset, so as to obtain multiple intermediate K space datasets.
[0146] The second data recovery module 840 is used to perform a second fitting recovery on the unsampled points in the non-central region of the K space of multiple intermediate K-space datasets to obtain multiple target K-space datasets;
[0147] The image reconstruction module 850 is used to reconstruct multiple target K-space datasets and obtain magnetic resonance images corresponding to multiple excitations of the target parts.
[0148] In one embodiment, the first data recovery module 830 includes:
[0149] The matrix construction unit is used to construct a data recovery matrix for any undersampled K-space dataset, based on the K-space calibration dataset and the undersampled K-space dataset.
[0150] The first recovery unit is used to perform a first fitting recovery of the unsampled points in the central region of the K space of each sampled K space dataset based on the data recovery matrix corresponding to each undersampled K space dataset.
[0151] In one embodiment, the matrix construction unit includes:
[0152] The first construction subunit is used to construct the first low-rank matrix based on the K-space calibration dataset using a preset low-rank matrix construction method;
[0153] The second construction subunit is used to construct a second low-rank matrix based on the undersampled K-space dataset using a low-rank matrix construction method.
[0154] The matrix construction sub-unit is used to generate a data recovery matrix based on the first low-rank matrix and the second low-rank matrix.
[0155] In one embodiment, the process of constructing a low-rank matrix includes:
[0156] Extract a predetermined number of distinct first data points from the target dataset and obtain the coordinate information of each first data point; the target dataset is a K-space calibration dataset or an undersampled K-space dataset.
[0157] For any given first data point, obtain multiple second data points that are less than a preset length away from the first data point, and obtain the set of data points corresponding to the first data point;
[0158] Obtain the signal values of multiple second data points in each data point set;
[0159] A low-rank matrix is constructed based on the coordinate information of each first data point and the signal values of multiple second data points in the data point set corresponding to each first data point.
[0160] In one embodiment, the second data recovery module 840 includes:
[0161] The weight calculation unit is used to calculate the weight kernel of the unsampled points in the non-central region of the K space of each intermediate K space dataset, based on the fully sampled fitted data of each intermediate K space dataset in the central region of the K space;
[0162] The second recovery unit is used to perform a second fitting recovery on the unsampled points in the non-center region for any intermediate K-space dataset, based on the weight kernel of the unsampled points.
[0163] Each module in the aforementioned image reconstruction apparatus 800 can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0164] Based on the same inventive concept, this application also provides a magnetic resonance imaging apparatus for implementing the magnetic resonance imaging method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more magnetic resonance imaging apparatus embodiments provided below can be found in the limitations of the magnetic resonance imaging method described above, and will not be repeated here.
[0165] In one embodiment, such as Figure 9 As shown, a magnetic resonance imaging (MRI) device 900 is provided, comprising: a data acquisition module 910, a scanning module 920, a first data recovery module 930, a second data recovery module 940, and an imaging module 950, wherein:
[0166] The data acquisition module 910 is used to fill the central region of the K-space using partial sampling technology to obtain the K-space calibration dataset corresponding to the target part;
[0167] The scanning module 920 is used to excite the target area multiple times and collect the undersampled K-space dataset corresponding to each excitation.
[0168] The first data recovery module 930 is used to perform a first fitting recovery on the unsampled points in the central region of the K space of each undersampled K space dataset based on the K space calibration dataset, so as to obtain multiple intermediate K space datasets.
[0169] The second data recovery module 940 is used to perform a second fitting recovery on the unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets, thereby obtaining multiple target K-space datasets.
[0170] The imaging module 950 is used to reconstruct multiple target K-space datasets and acquire magnetic resonance images corresponding to multiple excitations of the target area.
[0171] In one embodiment, the direction of the diffusion gradient applied in each of the multiple excitations is different.
[0172] In one embodiment, the physiological phase of the target site is different for each of the multiple stimulations.
[0173] In one embodiment, a labeling pulse is applied to the target site during each of the multiple excitations, and an undersampled K-space dataset is acquired at different delay times after the labeling pulse is applied.
[0174] Each module in the aforementioned magnetic resonance imaging device 900 can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.
[0175] In one embodiment, a computer device is provided, which may be an image reconstruction device, a magnetic resonance imaging device, or other terminal for generating magnetic resonance images. Its internal structure diagram may be as follows: Figure 10 As shown. The computer device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements the image reconstruction method and / or magnetic resonance imaging method provided in this application. The display screen can be a liquid crystal display (LCD) or an e-ink display. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.
[0176] Those skilled in the art will understand that Figure 10The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0177] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0178] Obtain the K-space calibration dataset corresponding to the target part. The K-space calibration dataset is fully sampled in the central region of K-space.
[0179] Multiple undersampled K-space datasets corresponding to the target part are obtained, and each undersampled K-space dataset contains data collected from the target part in one excitation.
[0180] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0181] A second fitting is performed on the unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets to recover multiple target K-space datasets;
[0182] Reconstruct multiple target K-space datasets and obtain corresponding magnetic resonance images of the target regions after multiple excitations.
[0183] In another embodiment, the processor, when executing a computer program, also performs the following steps:
[0184] A partial sampling technique is used to fill the central region of the K-space to obtain the K-space calibration dataset corresponding to the target part;
[0185] The target area is excited multiple times, and the undersampled K-space dataset corresponding to each excitement is collected;
[0186] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0187] A second fitting is performed on the unsampled points in the non-central region of the K-space from multiple intermediate K-space datasets to recover multiple target K-space datasets.
[0188] Reconstruct multiple target K-space datasets and obtain corresponding magnetic resonance images of the target regions after multiple excitations.
[0189] The computer device provided in the above embodiments has a similar implementation principle and technical effect to the above method embodiments, and will not be described again here.
[0190] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:
[0191] Obtain the K-space calibration dataset corresponding to the target part. The K-space calibration dataset is fully sampled in the central region of K-space.
[0192] Multiple undersampled K-space datasets corresponding to the target part are obtained, and each undersampled K-space dataset contains data collected from the target part in one excitation.
[0193] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0194] A second fitting is performed on the unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets to recover multiple target K-space datasets;
[0195] Reconstruct multiple target K-space datasets and obtain corresponding magnetic resonance images of the target regions after multiple excitations.
[0196] In another embodiment, the processor, when executing a computer program, also performs the following steps:
[0197] A partial sampling technique is used to fill the central region of the K-space to obtain the K-space calibration dataset corresponding to the target part;
[0198] The target area is excited multiple times, and the undersampled K-space dataset corresponding to each excitement is collected;
[0199] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0200] A second fitting is performed on the unsampled points in the non-central region of the K-space from multiple intermediate K-space datasets to recover multiple target K-space datasets.
[0201] Reconstruct multiple target K-space datasets and obtain corresponding magnetic resonance images of the target regions after multiple excitations.
[0202] The computer-readable storage medium provided in the above embodiments has a similar implementation principle and technical effect to the above method embodiments, and will not be described again here.
[0203] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:
[0204] Obtain the K-space calibration dataset corresponding to the target part. The K-space calibration dataset is fully sampled in the central region of K-space.
[0205] Multiple undersampled K-space datasets corresponding to the target part are obtained, and each undersampled K-space dataset contains data collected from the target part in one excitation.
[0206] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0207] A second fitting is performed on the unsampled points in the non-central region of the K-space of multiple intermediate K-space datasets to recover multiple target K-space datasets;
[0208] Reconstruct multiple target K-space datasets and obtain corresponding magnetic resonance images of the target regions after multiple excitations.
[0209] In another embodiment, the processor, when executing a computer program, also performs the following steps:
[0210] A partial sampling technique is used to fill the central region of the K-space to obtain the K-space calibration dataset corresponding to the target part;
[0211] The target area is excited multiple times, and the undersampled K-space dataset corresponding to each excitement is collected;
[0212] Based on the K-space calibration dataset, the first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets;
[0213] A second fitting is performed on the unsampled points in the non-central region of the K-space from multiple intermediate K-space datasets to recover multiple target K-space datasets.
[0214] Reconstruct multiple target K-space datasets and obtain corresponding magnetic resonance images of the target regions after multiple excitations.
[0215] The computer program product provided in the above embodiments has a similar implementation principle and technical effect to the above method embodiments, and will not be described again here.
[0216] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.
[0217] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0218] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.
Claims
1. An image reconstruction method, characterized in that, The method includes: Obtain the K-space calibration dataset corresponding to the target part, wherein the K-space calibration dataset is fully sampled in the central region of the K-space; Multiple undersampled K-space datasets corresponding to the target region are obtained, and each undersampled K-space dataset contains data collected from the target region in one excitation. Based on the K-space calibration dataset, a first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets; A second fitting recovery is performed on the unsampled points in the non-central region of the K-space of the multiple intermediate K-space datasets to obtain multiple target K-space datasets; Reconstruct the multiple target K-space datasets and obtain the corresponding magnetic resonance images of the target regions after multiple excitations.
2. The method according to claim 1, characterized in that, The first fitting recovery based on the K-space calibration dataset, for each undersampled K-space dataset, is performed on the unsampled points in the central region of the K-space, including: For any undersampled K-space dataset, a data recovery matrix is constructed based on the K-space calibration dataset and the undersampled K-space dataset; Based on the data recovery matrix corresponding to each of the undersampled K-space datasets, the unsampled points in the central region of the K-space of each sampled K-space dataset are first fitted and recovered.
3. The method according to claim 2, characterized in that, The step of constructing a data recovery matrix based on the K-space calibration dataset and the undersampled K-space dataset includes: Based on the K-space calibration dataset, a first low-rank matrix is constructed using a preset low-rank matrix construction method; Based on the undersampled K-space dataset, a second low-rank matrix is constructed using the low-rank matrix construction method described above; The data recovery matrix is generated based on the first low-rank matrix and the second low-rank matrix.
4. The method according to claim 3, characterized in that, The process of constructing a low-rank matrix includes: Extract a predetermined number of different first data points from the target dataset and obtain the coordinate information of each first data point; the target dataset is the K-space calibration dataset or the undersampled K-space dataset. For any given first data point, obtain multiple second data points whose distance from the first data point is less than a preset length, and obtain the data point set corresponding to the first data point; Obtain the signal values of multiple second data points from each of the data point sets; A low-rank matrix is constructed based on the coordinate information of each first data point and the signal values of multiple second data points in the data point set corresponding to each first data point.
5. The method according to any one of claims 1 to 4, characterized in that, The second fitting recovery of unsampled points in the non-center region of the multiple intermediate K-space datasets includes: Based on the fitted full-sample data of each intermediate K-space dataset in the central region of K-space, calculate the weight kernel of the unsampled points in the non-central region of K-space for each intermediate K-space dataset. For any intermediate K-space dataset, a second fitting recovery is performed on the unsampled points in the non-central region based on the weight kernel of the unsampled points.
6. A magnetic resonance imaging method, characterized in that, The method includes: A partial sampling technique is used to fill the central region of the K-space to obtain the K-space calibration dataset corresponding to the target part; The target area is excited multiple times, and the undersampled K-space dataset corresponding to each excitation is collected; Based on the K-space calibration dataset, a first fitting recovery is performed on the unsampled points in the central region of K-space for each undersampled K-space dataset to obtain multiple intermediate K-space datasets; A second fitting recovery is performed on the unsampled points in the non-central region of the multiple intermediate K-space datasets to obtain multiple target K-space datasets. Reconstruct the multiple target K-space datasets and obtain the corresponding magnetic resonance images of the target regions after multiple excitations.
7. The method according to claim 6, characterized in that, The direction of the diffusion gradient applied in each of the multiple excitations is different.
8. The method according to claim 6, characterized in that, The physiological phase of the target site is different for each of the multiple stimulations.
9. The method according to claim 6, characterized in that, In each of the multiple excitations, a labeling pulse is applied to the target region, and the undersampled K-space dataset is collected at different delay times after the labeling pulse is applied.
10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 9.