Cardiopulmonary magnetic resonance imaging method and system based on low-rank tensor approximation algorithm

Abstract

The invention relates to the technical field of magnetic resonance imaging, in particular to a cardiopulmonary magnetic resonance imaging method and system based on a low-rank tensor approximation algorithm, and the method comprises the steps: S1, synchronously collecting magnetic resonance k-space imaging data, breathing phase data and cardiac phase data of a patient; s2, constructing a three-dimensional data tensor according to the magnetic resonance k-space imaging data, the breathing phase data and the heartbeat phase data; s3, filling an unsampled region corresponding to the magnetic resonance k-space imaging data in the three-dimensional data tensor by adopting low-rank approximate estimation to obtain a filled three-dimensional data tensor; and S4, carrying out image reconstruction by adopting the filled three-dimensional data tensor. The method has the beneficial effects that in the three-dimensional data tensor, the breathing phase data and the heartbeat phase data are combined to jointly carry out low-rank tensor approximate estimation, so that filling of the part of data is realized, a high-quality ULF-MRI image is conveniently obtained, and the scanning time does not need to be prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic resonance imaging, and in particular to a cardiopulmonary magnetic resonance imaging method and system based on a low-rank tensor approximation algorithm. Background Art

[0002] In the field of cardiac and pulmonary magnetic resonance imaging, in order to obtain high-quality imaging results, it is often necessary to counteract the effects of cardiac pulsation and respiratory motion. Traditional high-field magnetic resonance imaging technology tends to use respiratory gating and ECG gating technology to acquire images to reduce motion artifacts.

[0003] Gating mainly refers to synchronizing MR data acquisition with the heart or respiratory cycle, which is called respiratory gating or ECG gating according to the different synchronization objects. Among them, the gating process may be prospective or retrospective. In retrospective gating, MR data is continuously acquired (not in response to specific heart or lung "trigger" events). ECG, pulse or respiratory level is recorded simultaneously with MR data. The MR data can then be reordered, grouped or associated with the stages of the heart and lung cycle. Retrospective gating is commonly used for cardiac cine imaging.

[0004] Taking ECG gating as an example, ECG or peripheral pulse sensors can detect the heart cycle. However, due to the relatively long process between cardiac contraction and the pulse wave reaching the finger, and the difficulty in accurately estimating the delay, peripheral pulse gating cannot be used for cardiac imaging, but is often used for peripheral MRA and CSF movie scanning.

[0005] As for respiratory gating, it is usually applied to some long-term scanning scenarios. In the short-term imaging process of the lower chest and upper abdomen, some methods can be used to freeze the diaphragm movement, such as issuing specific instructions to the patient to make the patient hold his breath for a period of time, and complete the imaging process during the breath-holding period. But it is often not suitable for longer (more than 20-30 seconds) scanning processes. For this, respiratory gating technology is often required to determine the respiratory cycle. Generally speaking, respiratory expansion can be detected by using a chest belt, bellows or pads. For a "chest respirator", the belt should be placed around the lower chest, and for an "abdominal respirator", it should be placed around the mid-abdomen.

[0006] However, in ultra-low field magnetic resonance imaging (ULF-MRI), this gating technique is not suitable for cardiopulmonary imaging in ULF-MRI due to the significantly increased scanning time and reduced signal-to-noise ratio, which reduces the effective data acquisition time. This leads to a series of problems, mainly including: 1) the gated acquisition method is inefficient in ultra-low field magnetic resonance imaging; 2) the low signal-to-noise ratio leads to poor image quality; 3) long data acquisition time causes patient discomfort and low imaging process efficiency. The existence of these problems limits the application of ultra-low field magnetic resonance imaging technology in cardiopulmonary imaging. Summary of the invention

[0007] In view of the above problems existing in the prior art, a cardiopulmonary magnetic resonance imaging method based on a low-rank tensor approximation algorithm is provided; on the other hand, a system for implementing the cardiopulmonary magnetic resonance imaging method is also provided.

[0008] The specific technical solutions are as follows:

[0009] A cardiopulmonary magnetic resonance imaging method based on a low-rank tensor approximation algorithm, comprising:

[0010] Step S1: synchronously acquiring magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data of the patient;

[0011] Step S2: constructing a three-dimensional data tensor according to the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data;

[0012] Step S3: filling an unsampled region in the three-dimensional data tensor corresponding to the magnetic resonance k-space imaging data using low-rank approximate estimation to obtain a filled three-dimensional data tensor;

[0013] Step S4: using the filled three-dimensional data tensor to perform image reconstruction.

[0014] On the other hand, in step S1, the scheme for acquiring the magnetic resonance k-space imaging data includes:

[0015] In three-dimensional k-space, the kz direction is perpendicular to the two-dimensional radial sampling plane;

[0016] When full sampling is performed, a first preset number of radial sampling lines are collected in each kz plane, and each radial sampling line includes a second preset number of sampling points;

[0017] The angle between the radial sampling line in the tth item and the X-axis is:

[0018]

[0019] Wherein, θ is the preset angle, t is the serial number of the radial sampling line, N is the first preset number, [·] represents the rounding operation, Represents a floor operation.

[0020] On the other hand, the step S2 comprises:

[0021] Step S21: determining the tensor phase according to the respiratory phase data and the cardiac phase data;

[0022] Step S22: constructing the three-dimensional data tensor, and sequentially filling the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data into the three-dimensional data tensor according to the tensor phase;

[0023] Step S23: Fill the default values ​​in the three-dimensional data tensor with zeros to form a complete three-dimensional data tensor.

[0024] On the other hand, the step S3 comprises:

[0025] Step S31: decomposing the three-dimensional data tensor to obtain multiple intermediate linear combinations;

[0026] Step S32: performing alternating least squares processing on the intermediate linear combination to generate low-rank approximate data of the three-dimensional data tensor;

[0027] Step S33: filling the three-dimensional data tensor according to the low-rank approximate data;

[0028] Step S34: Return to step S31 to iterate the three-dimensional data tensor until an iteration condition is met and the three-dimensional data tensor generated last time is output as the filled three-dimensional data tensor.

[0029] On the other hand, the step S4 comprises:

[0030] Step S41: extracting the respiratory phase data and the cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data from the filled three-dimensional data tensor as extracted data;

[0031] Step S42: transforming the extracted data and the corresponding magnetic resonance k-space imaging data into k-space, and then performing image reconstruction.

[0032] A cardiopulmonary magnetic resonance imaging system based on a low-rank tensor approximation algorithm, used to implement the above-mentioned cardiopulmonary magnetic resonance imaging method, comprising:

[0033] An acquisition module, wherein the acquisition module synchronously acquires magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data of the patient;

[0034] A data tensor construction module, the data tensor construction module is connected to the acquisition module, and the data tensor construction module constructs a three-dimensional data tensor according to the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data;

[0035] A low-rank approximate estimation module, wherein the low-rank approximate estimation module is connected to the data tensor construction module, and wherein the low-rank approximate estimation module uses low-rank approximate estimation to fill an unsampled region in the three-dimensional data tensor corresponding to the magnetic resonance k-space imaging data to obtain a filled three-dimensional data tensor;

[0036] An image reconstruction module is connected to the low-rank approximation estimation module, and the image reconstruction module uses the filled three-dimensional data tensor to reconstruct the image.

[0037] On the other hand, the acquisition module includes a magnetic resonance acquisition module, and the magnetic resonance acquisition module is used to acquire the magnetic resonance k-space imaging data;

[0038] The magnetic resonance acquisition module comprises:

[0039] A sampling configuration module, wherein the angle generation module configures a first preset number of radial sampling lines for each kz plane in the three-dimensional k space, and each radial sampling line includes a second preset number of sampling points;

[0040] A data integration module is connected to the sampling configuration module, and the data integration module generates the magnetic resonance k-space imaging data according to the sampling data of the sampling points.

[0041] On the other hand, the data tensor construction module includes:

[0042] a phase determination module, wherein the phase determination module determines a tensor phase according to the respiratory phase data and the cardiac phase data;

[0043] A tensor filling module, the tensor filling module is connected to the phase determination module, the tensor filling module constructs the three-dimensional data tensor, and fills the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data into the three-dimensional data tensor in sequence according to the tensor phase;

[0044] A zero-filling module, wherein the zero-filling module is connected to the tensor filling module, and the zero-filling module fills the default values ​​in the three-dimensional data tensor with zeros to form a complete three-dimensional data tensor.

[0045] On the other hand, the low-rank approximate estimation module includes:

[0046] A decomposition module, wherein the decomposition module decomposes the three-dimensional data tensor to obtain a plurality of intermediate linear combinations;

[0047] An estimation module, the estimation module is connected to the decomposition module, and the estimation module performs an alternating least squares process on the intermediate linear combination to generate low-rank approximate data of the three-dimensional data tensor;

[0048] A filling module, the filling module is connected to the estimation module, and the filling module fills the three-dimensional data tensor according to the low-rank approximate data;

[0049] An iteration module, wherein the iteration module is respectively connected to the decomposition module and the filling module, and the iteration module iterates the three-dimensional data tensor until an iteration condition is met and the three-dimensional data tensor generated for the last time is output as the filled three-dimensional data tensor.

[0050] On the other hand, the image reconstruction module comprises:

[0051] A data extraction module, wherein the data extraction module extracts the respiratory phase data and the cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data from the filled three-dimensional data tensor as extracted data;

[0052] A reconstruction module is connected to the data extraction module, and the reconstruction module transforms the extracted data and the corresponding magnetic resonance k-space imaging data into k-space, and then performs image reconstruction.

[0053] The above technical solution has the following advantages or beneficial effects:

[0054] In view of the fact that the gating technology in the prior art is prone to low imaging efficiency and poor images when applied to ultra-low field magnetic resonance imaging, this technical solution introduces a method of estimating and completing the magnetic resonance k-space imaging data based on a low-rank tensor approximation algorithm to achieve better imaging efficiency in this scenario. Specifically, since the signal-to-noise ratio of ULF-MRI is relatively low during the imaging process, and the acquisition time is short after the gating technology is applied, there are fewer effective data components in the magnetic resonance k-space imaging data. For the missing parts, in the three-dimensional data tensor, the respiratory phase data and the cardiac phase data are combined to perform low-rank tensor approximation estimation together, thereby achieving the filling of this part of the data, making it easy to obtain high-quality ULF-MRI images without extending the scanning time. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The embodiments of the present invention will be described more fully with reference to the attached drawings, which are provided for illustration and description only and are not intended to limit the scope of the present invention.

[0056] Figure 1 is an overall schematic diagram of an embodiment of the present invention;

[0057] Figure 2 This is a schematic diagram of step S2 in an embodiment of the present invention;

[0058] Figure 3 A schematic diagram of a three-dimensional data tensor in an embodiment of the present invention;

[0059] Figure 4 This is a schematic diagram of step S3 in an embodiment of the present invention;

[0060] Figure 5 This is a schematic diagram of step S4 in an embodiment of the present invention;

[0061] Figure 6 Schematic diagram of a cardiopulmonary magnetic resonance imaging system in an embodiment of the present invention;

[0062] Figure 7 This is a schematic diagram of a magnetic resonance acquisition module in an embodiment of the present invention;

[0063] Figure 8 A schematic diagram of a data tensor construction module in an embodiment of the present invention;

[0064] Fig. 9 Schematic diagram of a low-rank approximate estimation module in an embodiment of the present invention;

[0065] Fig.10 Schematic diagram of an image reconstruction module in an embodiment of the present invention. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0068] Some of the blocks shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0069] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0070] The present invention comprises:

[0071] A cardiopulmonary magnetic resonance imaging method based on a low-rank tensor approximation algorithm, such as Figure 1 As shown, including:

[0072] Step S1: synchronously acquiring magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data of the patient;

[0073] Step S2: constructing a three-dimensional data tensor according to the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data;

[0074] Step S3: filling an unsampled region corresponding to the magnetic resonance k-space imaging data in the three-dimensional data tensor with a low-rank approximate estimation to obtain a filled three-dimensional data tensor;

[0075] Step S4: Reconstructing the image using the filled three-dimensional data tensor.

[0076] Specifically, in view of the fact that the gating technology in the prior art is prone to lead to low imaging efficiency and poor images when applied to ultra-low field magnetic resonance imaging, the present technical solution introduces a method of estimating and completing the magnetic resonance k-space imaging data based on a low-rank tensor approximation algorithm to achieve better imaging efficiency in this scenario.

[0077] Specifically, during the MRI scanning process, it is necessary to synchronously collect MRI k-space imaging data, respiratory phase data, and cardiac phase data from the patient. The MRI k-space imaging data, respiratory phase data, and cardiac phase data have the same phase in time sequence, and can represent the user's heartbeat and breathing conditions at a certain moment through the respiratory phase data and cardiac phase data, thereby providing a basis for the influence of respiratory and cardiac movements on MRI in the subsequent image reconstruction process.

[0078] Since the MRI k-space imaging data is acquired by ultra-low field MRI equipment, its signal-to-noise ratio is lower than that of the results obtained by traditional high field MRI. In the traditional gating process, the sequence needs to be cut according to the breathing and heartbeat cycles, so the overall information is relatively small and the imaging quality is low.

[0079] To address this problem, in this embodiment, a three-dimensional data tensor is first constructed around the magnetic resonance k-space imaging data, the respiratory phase data, and the cardiac phase data, which has three dimensions corresponding to the magnetic resonance k-space imaging data, the respiratory phase data, and the cardiac phase data, respectively. On this basis, different data among the magnetic resonance k-space imaging data, the respiratory phase data, and the cardiac phase data can be filled into the three-dimensional data tensor according to the position of the phase.

[0080] Since the magnetic resonance k-space imaging data has a small amount of valid data after being processed by the gating technology, the valid values ​​in the three-dimensional data tensor are relatively small. In this regard, the present embodiment adopts a low-rank approximate estimation method to fill the unsampled area of ​​the magnetic resonance k-space imaging data, so that the part is completed. The completed data is used as the filled three-dimensional data tensor for the subsequent image reconstruction process, thereby achieving a better image reconstruction effect.

[0081] In the actual implementation process, the above-mentioned cardiopulmonary magnetic resonance imaging method is mainly configured as a software embodiment in a computer device, such as a processing computer of a magnetic resonance imaging device, a hospital workstation or other equivalent equipment, which is used to receive external data, including magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data, and process them in the above manner to obtain a magnetic resonance image. Among them, the respiratory phase data is mainly collected using an abdominal belt and a breathing pad to obtain the movement of the diaphragm or chest cavity during the patient's breathing process. The cardiac phase data is mainly achieved by electrocardiography or blood oxygen saturation monitoring to obtain the patient's cardiac phase during the magnetic resonance scanning process. During the acquisition process, it is necessary to control the respiratory phase data and the cardiac phase data to align with the time point of the magnetic resonance examination.

[0082] In one embodiment, in step S1, the scheme for acquiring magnetic resonance k-space imaging data includes:

[0083] In the three-dimensional k-space, the kz plane is perpendicular to the two-dimensional radial sampling plane;

[0084] When full sampling is performed, a first preset number of radial sampling lines are collected in each kz plane, and each radial sampling line includes a second preset number of sampling points;

[0085] The angle between the radial sampling line in the tth item and the X-axis is:

[0086]

[0087] Wherein, θ is the preset angle, t is the serial number of the radial sampling line, N is the first preset number, [·] represents the rounding operation, Represents a floor operation.

[0088] Specifically, in order to achieve a better sampling effect, in this embodiment, in the process of collecting magnetic resonance k-space imaging data, a stack-of-stars trajectory based on the golden section angle is used for sampling. Specifically, the sampling process is carried out in the three-dimensional k-space. First, the corresponding radial sampling lines and sampling points are configured for the kz plane. The multiple radial sampling lines are ordered, and their angles relative to the X-axis are generated in turn according to the above formula, so that the radial sampling lines evenly cover the k-space.

[0089] On this basis, the configured radial sampling lines are used for imaging. The imaging direction includes first traversing the kz direction and then traversing each angle in the kz plane. The number of kz planes cannot be an integer multiple of any number in the Fibonacci sequence.

[0090] During the implementation process, the echo sequences used in the above imaging process include fast spin echo (Fast Spin Echo, FSE), echo planar imaging (Echo Planar Imaging, EPI), balanced Steady State Free Precession (balanced Steady State Free Precession, bSSFP), spoiled gradient echo (Spoiled Gradient Recalled Echo, SPGR) and other sequences.

[0091] It should be noted that, when the echo sequence is fast spin echo imaging and echo planar imaging, the code number in the kz direction should also be an integer multiple of the set echo train length (Echo Train Length, ETL).

[0092] In one embodiment, Figure 2 As shown, step S2 includes:

[0093] Step S21: determining the tensor phase according to the respiratory phase data and the cardiac phase data;

[0094] Step S22: constructing a three-dimensional data tensor, and sequentially filling the magnetic resonance k-space imaging data, respiratory phase data, and cardiac phase data into the three-dimensional data tensor according to the tensor phase;

[0095] Step S23: Fill the default values ​​in the three-dimensional data tensor with zeros to form a complete three-dimensional data tensor.

[0096] Specifically, in order to achieve the data processing effect, in this embodiment, after the acquisition of magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data is realized, the corresponding phase information is first extracted from the respiratory phase data and cardiac phase data to determine the tensor phase in the three-dimensional data tensor, which corresponds to the unit step size in the three-dimensional data tensor. Then, a three-dimensional data tensor is constructed with a size of (MN)×N resp ×N card , corresponding to the k-space position, respiratory phase and cardiac phase respectively, and the magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data are filled into the three-dimensional data tensor in sequence according to the tensor phase.

[0097] like Figure 3 As shown, the left side is the k-space of a specific respiratory phase and cardiac phase, wherein the hollow circles represent unsampled positions and the solid circles represent sampled positions, and the right side is the constructed three-dimensional data tensor.

[0098] Accordingly, in the three-dimensional data tensor, after the phase alignment of the respiratory phase data and the cardiac phase data, the magnetic resonance k-space imaging data will have repeated values ​​and default values. The default values ​​are filled with zeros to make the data complete, and the repeated values ​​are filled by averaging sampling, thereby realizing the complete three-dimensional data tensor construction process.

[0099] In one embodiment, Figure 4 As shown, step S3 includes:

[0100] Step S31: decomposing the three-dimensional data tensor to obtain multiple intermediate linear combinations;

[0101] Step S32: performing alternating least squares processing on the intermediate linear combination to generate low-rank approximate data of the three-dimensional data tensor;

[0102] Step S33: Filling the three-dimensional data tensor according to the low-rank approximate data;

[0103] Step S34: Return to step S31 to iterate the three-dimensional data tensor until the iteration condition is met and the three-dimensional data tensor generated last time is output as the filled three-dimensional data tensor.

[0104] Specifically, in order to achieve a better data filling effect, in this embodiment, a low-rank approximate estimation method is used for the three-dimensional data tensor to perform multiple iterative optimizations, so that the magnetic resonance k-space imaging data in the three-dimensional data tensor conforms to the actual situation.

[0105] In each iteration, the three-dimensional data tensor is first decomposed into multiple intermediate linear combinations Where T is a three-dimensional data tensor, Ris the target rank, ar, br, cr are the factor vectors after decomposition.

[0106] Then, these factor vectors can be obtained by using algorithms such as alternating least squares (ALS) to minimize the approximation error, thereby obtaining low-rank approximate data of the three-dimensional data tensor.

[0107] After that, the data consistency operation is performed, that is, the sampled data is refilled into the corresponding position, thus completing this iteration process. The above operation is repeated until the data converges, so as to achieve effective estimation of the uncollected data.

[0108] In one embodiment, Figure 5 As shown, step S4 includes:

[0109] Step S41: extracting respiratory phase data and cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data from the filled three-dimensional data tensor as extracted data;

[0110] Step S42: transforming the extracted data and the corresponding magnetic resonance k-space imaging data into k-space, and then performing image reconstruction.

[0111] Specifically, in order to achieve better image reconstruction effects, in this embodiment, after completing the above-mentioned processing process, respiratory phase data and cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data are extracted from the filled three-dimensional data tensor as extracted data, the extracted data are transformed back to the k-space, and the data are converted to the image domain using a non-uniform fast Fourier transform to obtain corresponding cardiopulmonary magnetic resonance images. Dynamic images of the heart and lungs can also be obtained for further evaluation of cardiopulmonary function.

[0112] A cardiopulmonary magnetic resonance imaging system based on a low-rank tensor approximation algorithm is used to implement the above-mentioned cardiopulmonary magnetic resonance imaging method, such as Figure 6 As shown, including:

[0113] Acquisition module 1, acquisition module 1 synchronously acquires magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data of the patient;

[0114] A data tensor construction module 2, the data tensor construction module 2 is connected to the acquisition module 1, and the data tensor construction module constructs a three-dimensional data tensor according to the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data;

[0115] A low-rank approximate estimation module 3, wherein the low-rank approximate estimation module 3 is connected to the data tensor construction module 2, and the low-rank approximate estimation module 3 uses low-rank approximate estimation to fill the unsampled area corresponding to the magnetic resonance k-space imaging data in the three-dimensional data tensor to obtain a filled three-dimensional data tensor;

[0116] The image reconstruction module 4 is connected to the low-rank approximation estimation module 3, and the image reconstruction module 4 uses the filled three-dimensional data tensor to reconstruct the image.

[0117] Specifically, in view of the fact that the gating technology in the prior art is prone to lead to low imaging efficiency and poor images when applied to ultra-low field magnetic resonance imaging, the present technical solution introduces a method of estimating and completing the magnetic resonance k-space imaging data based on a low-rank tensor approximation algorithm to achieve better imaging efficiency in this scenario.

[0118] Specifically, during the magnetic resonance scanning process, the acquisition module 1 synchronously acquires magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data of the patient.

[0119] Then, the data tensor construction module 2 constructs a three-dimensional data tensor around the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data, which has three dimensions corresponding to the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data, respectively. On this basis, different data among the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data can be filled into the three-dimensional data tensor according to the position of the phase.

[0120] Subsequently, the low-rank approximate estimation module 3 uses a low-rank approximate estimation method to fill the unsampled area of ​​the magnetic resonance k-space imaging data, so that the part is completed. The completed data is used as a filled three-dimensional data tensor for the subsequent image reconstruction module 4 to perform an image reconstruction process, thereby achieving a better image reconstruction effect.

[0121] In one embodiment, the acquisition module 1 includes a magnetic resonance acquisition module 11, and the magnetic resonance acquisition module 11 is used to acquire magnetic resonance k-space imaging data;

[0122] like Figure 7 As shown, the magnetic resonance acquisition module 11 includes:

[0123] The sampling configuration module 111, the angle generation module 111 configures and collects a first preset number of radial sampling lines for each kz plane in the three-dimensional k space, each radial sampling line includes a second preset number of sampling points;

[0124] The data integration module 112 is connected to the sampling configuration module 111 . The data integration module 112 generates magnetic resonance k-space imaging data according to the sampling data of the sampling points.

[0125] Specifically, in order to achieve a better sampling effect, in this embodiment, in the process of collecting magnetic resonance k-space imaging data, a stack-of-stars trajectory based on the golden section angle is used for sampling. Specifically, the sampling process is performed in the three-dimensional k-space. First, the sampling configuration module 111 configures the corresponding radial sampling lines and sampling points on the kz plane. The multiple radial sampling lines are orderly, so that the radial sampling lines evenly cover the k-space.

[0126] On this basis, the data integration module 112 generates magnetic resonance k-space imaging data according to the sampling data of the sampling points, thereby achieving full sampling of the k-space.

[0127] In one embodiment, Figure 8 As shown, the data tensor building module 2 includes:

[0128] A phase determination module 21, the phase determination module 21 determines the tensor phase according to the respiratory phase data and the cardiac phase data;

[0129] A tensor filling module 22, the tensor filling module 22 is connected to the phase determination module 21, the tensor filling module 22 constructs a three-dimensional data tensor, and fills the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data into the three-dimensional data tensor in sequence according to the tensor phase;

[0130] The zero-filling module 23 is connected to the tensor filling module 22, and the zero-filling module 23 fills the default values ​​in the three-dimensional data tensor with zeros to form a complete three-dimensional data tensor.

[0131] Specifically, to achieve the data processing effect, in this embodiment, after the acquisition of the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data is realized, the phase determination module 21 first extracts the corresponding phase information according to the respiratory phase data and the cardiac phase data, thereby determining the tensor phase in the three-dimensional data tensor, and the tensor phase corresponds to the unit step length in the three-dimensional data tensor. Then, the three-dimensional data tensor is constructed, and the tensor filling module 22 fills the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data into the three-dimensional data tensor in sequence according to the tensor phase.

[0132] Accordingly, in the three-dimensional data tensor, after the phase alignment of the respiratory phase data and the cardiac phase data, the magnetic resonance k-space imaging data will have repeated values ​​and default values. The zero-filling module 23 fills the default values ​​with zeros to make the data complete, and fills the repeated values ​​by averaging sampling, thereby realizing a complete three-dimensional data tensor construction process.

[0133] In one embodiment, Fig. 9 As shown, the low-rank approximate estimation module 3 includes:

[0134] A decomposition module 31, the decomposition module 31 decomposes the three-dimensional data tensor to obtain a plurality of intermediate linear combinations;

[0135] An estimation module 32, the estimation module 32 is connected to the decomposition module 31, and the estimation module 32 performs an alternating least squares process on the intermediate linear combination to generate low-rank approximate data of the three-dimensional data tensor;

[0136] A filling module 33, the filling module 33 is connected to the estimation module 32, and the filling module 33 fills the three-dimensional data tensor according to the low-rank approximate data;

[0137] The iteration module 34 is connected to the decomposition module 31 and the filling module 33 respectively. The iteration module 34 iterates the three-dimensional data tensor until the iteration condition is met and the three-dimensional data tensor generated last time is output as the filled three-dimensional data tensor.

[0138] Specifically, in order to achieve a better data filling effect, in this embodiment, a low-rank approximate estimation method is used for the three-dimensional data tensor to perform multiple iterative optimizations, so that the magnetic resonance k-space imaging data in the three-dimensional data tensor conforms to the actual situation.

[0139] In each iteration, the decomposition module 31 first decomposes the three-dimensional data tensor into multiple intermediate linear combinations Where T is a three-dimensional data tensor, R is the target rank, ar, br, cr are the factor vectors after decomposition.

[0140] Then, the estimation module 32 can obtain these factor vectors through algorithms such as alternating least squares (ALS) to minimize the approximation error, thereby obtaining low-rank approximate data of the three-dimensional data tensor.

[0141] Then the filling module 33 performs data consistency operation, that is, refills the sampled data into the corresponding position, thereby completing this iteration process. The iteration module 34 repeats the above operation until the data converges, thereby achieving effective estimation of the uncollected data.

[0142] In one embodiment, Fig.10 As shown, the image reconstruction module 41 includes:

[0143] A data extraction module 41, the data extraction module 41 extracts the respiratory phase data and the cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data from the filled three-dimensional data tensor as extracted data;

[0144] The reconstruction module 42 is connected to the data extraction module 41 . The reconstruction module 42 transforms the extracted data and the corresponding magnetic resonance k-space imaging data into k-space, and then performs image reconstruction.

[0145] Specifically, in order to achieve better image reconstruction effects, in this embodiment, after completing the above-mentioned processing process, the data extraction module 41 also extracts the respiratory phase data and cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data from the filled three-dimensional data tensor as the extracted data, and transforms the extracted data back to the k-space. The reconstruction module 42 uses a non-uniform fast Fourier transform to convert the data to the image domain to obtain the corresponding cardiopulmonary magnetic resonance image, and can also obtain dynamic images of the heart and lungs for further evaluation of cardiopulmonary function.

[0146] The above are only preferred embodiments of the present invention, and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A cardiopulmonary magnetic resonance imaging method based on a low-rank tensor approximation algorithm, characterized in that: include: Step S1: synchronously acquiring magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data of the patient; Step S2: constructing a three-dimensional data tensor according to the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data; Step S3: filling an unsampled region in the three-dimensional data tensor corresponding to the magnetic resonance k-space imaging data using low-rank approximate estimation to obtain a filled three-dimensional data tensor; Step S4: using the filled three-dimensional data tensor to perform image reconstruction.

2. The cardiopulmonary magnetic resonance imaging method according to claim 1, characterized in that: In step S1, the scheme for acquiring the magnetic resonance k-space imaging data includes: In the three-dimensional k-space, the kz plane is perpendicular to the two-dimensional radial sampling plane; When full sampling is performed, a first preset number of radial sampling lines are collected in each kz plane, and each radial sampling line includes a second preset number of sampling points; The angle between the radial sampling line in the tth item and the X-axis is: Wherein, θ is the preset angle, t is the serial number of the radial sampling line, N is the first preset number, [·] represents the rounding operation, Represents a floor operation.

3. The cardiopulmonary magnetic resonance imaging method according to claim 1, characterized in that: The step S2 comprises: Step S21: determining the tensor phase according to the respiratory phase data and the cardiac phase data; Step S22: constructing the three-dimensional data tensor, and sequentially filling the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data into the three-dimensional data tensor according to the tensor phase; Step S23: Fill the default values ​​in the three-dimensional data tensor with zeros to form a complete three-dimensional data tensor.

4. The cardiopulmonary magnetic resonance imaging method according to claim 1, characterized in that: The step S3 comprises: Step S31: decomposing the three-dimensional data tensor to obtain multiple intermediate linear combinations; Step S32: performing alternating least squares processing on the intermediate linear combination to generate low-rank approximate data of the three-dimensional data tensor; Step S33: filling the three-dimensional data tensor according to the low-rank approximate data; Step S34: Return to step S31 to iterate the three-dimensional data tensor until an iteration condition is met and the three-dimensional data tensor generated last time is output as the filled three-dimensional data tensor.

5. The cardiopulmonary magnetic resonance imaging method according to claim 1, characterized in that: The step S4 comprises: Step S41: extracting the respiratory phase data and the cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data from the filled three-dimensional data tensor as extracted data; Step S42: transforming the extracted data and the corresponding magnetic resonance k-space imaging data into k-space, and then performing image reconstruction.

6. A cardiopulmonary magnetic resonance imaging system based on a low-rank tensor approximation algorithm, characterized in that: A method for implementing the cardiopulmonary magnetic resonance imaging method according to any one of claims 1 to 5, comprising: An acquisition module, wherein the acquisition module synchronously acquires magnetic resonance k-space imaging data, respiratory phase data and cardiac phase data of the patient; A data tensor construction module, the data tensor construction module is connected to the acquisition module, and the data tensor construction module constructs a three-dimensional data tensor according to the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data; A low-rank approximate estimation module, wherein the low-rank approximate estimation module is connected to the data tensor construction module, and wherein the low-rank approximate estimation module uses low-rank approximate estimation to fill an unsampled region in the three-dimensional data tensor corresponding to the magnetic resonance k-space imaging data to obtain a filled three-dimensional data tensor; An image reconstruction module is connected to the low-rank approximation estimation module, and the image reconstruction module uses the filled three-dimensional data tensor to reconstruct the image.

7. The cardiopulmonary magnetic resonance imaging system according to claim 6, characterized in that: The acquisition module comprises a magnetic resonance acquisition module, and the magnetic resonance acquisition module is used to acquire the magnetic resonance k-space imaging data; The magnetic resonance acquisition module comprises: A sampling configuration module, wherein the angle generation module configures a first preset number of radial sampling lines for each kz plane in the three-dimensional k space, and each radial sampling line includes a second preset number of sampling points; A data integration module is connected to the sampling configuration module, and the data integration module generates the magnetic resonance k-space imaging data according to the sampling data of the sampling points.

8. The cardiopulmonary magnetic resonance imaging system according to claim 6, characterized in that: The data tensor building module includes: a phase determination module, wherein the phase determination module determines a tensor phase according to the respiratory phase data and the cardiac phase data; A tensor filling module, the tensor filling module is connected to the phase determination module, the tensor filling module constructs the three-dimensional data tensor, and fills the magnetic resonance k-space imaging data, the respiratory phase data and the cardiac phase data into the three-dimensional data tensor in sequence according to the tensor phase; A zero-filling module, wherein the zero-filling module is connected to the tensor filling module, and the zero-filling module fills the default values ​​in the three-dimensional data tensor with zeros to form a complete three-dimensional data tensor.

9. The cardiopulmonary magnetic resonance imaging system according to claim 6, characterized in that: The low-rank approximate estimation module comprises: A decomposition module, wherein the decomposition module decomposes the three-dimensional data tensor to obtain a plurality of intermediate linear combinations; An estimation module, the estimation module is connected to the decomposition module, and the estimation module performs an alternating least squares process on the intermediate linear combination to generate low-rank approximate data of the three-dimensional data tensor; A filling module, the filling module is connected to the estimation module, and the filling module fills the three-dimensional data tensor according to the low-rank approximate data; An iteration module, wherein the iteration module is respectively connected to the decomposition module and the filling module, and the iteration module iterates the three-dimensional data tensor until an iteration condition is met and the three-dimensional data tensor generated for the last time is output as the filled three-dimensional data tensor.

10. The cardiopulmonary magnetic resonance imaging system according to claim 6, characterized in that: The image reconstruction module comprises: A data extraction module, wherein the data extraction module extracts the respiratory phase data and the cardiac phase data having the same phase parameters as the magnetic resonance k-space imaging data from the filled three-dimensional data tensor as extracted data; A reconstruction module is connected to the data extraction module, and the reconstruction module transforms the extracted data and the corresponding magnetic resonance k-space imaging data into k-space, and then performs image reconstruction.

Images