Quantitative magnetic resonance imaging method and system, electronic equipment and product

Abstract

The invention belongs to the technical field of magnetic resonance imaging, and aims to provide a quantitative magnetic resonance imaging method and system, electronic equipment and a product. After a moisture content distribution diagram of a to-be-detected sample and an amplitude diagram of a multi-channel radio frequency receiving magnetic field are obtained according to a magnetic resonance signal, a T1 relaxation parameter diagram of the to-be-detected sample and a first complex electrical characteristic distribution diagram based on the moisture content can be obtained according to the moisture content distribution diagram; inputting the amplitude diagram of the multi-channel radio frequency receiving magnetic field into a preset second deep learning network to obtain a second complex electrical characteristic distribution diagram of the to-be-tested sample based on a radio frequency receiving field vector, and obtaining the amplitude diagram of the multi-channel radio frequency receiving magnetic field according to the first complex electrical characteristic distribution diagram and the second complex electrical characteristic distribution diagram. And obtaining a final complex electrical characteristic distribution diagram of the to-be-tested sample. According to the invention, the imaging range of quantitative magnetic resonance imaging is expanded, and the accuracy of the quantitative magnetic resonance imaging result can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of magnetic resonance imaging, and particularly relates to a quantitative magnetic resonance imaging method, system, electronic device and product. Background Art

[0002] Currently, in order to obtain magnetic resonance images of various physiological parameters of a human tissue to be measured, it is usually necessary to perform multiple magnetic resonance scans on the human tissue to be measured, and then reconstruct magnetic resonance images of different physiological parameters of the human tissue to be measured according to the magnetic resonance signals obtained from the multiple magnetic resonance scans.

[0003] In order to reduce the time consumption of magnetic resonance scanning and improve the consistency between magnetic resonance images of various physiological parameters, currently, Magnetic Resonance Fingerprinting (MRF) technology has emerged. It is a quantitative magnetic resonance imaging method developed in recent years, aiming to simultaneously image multiple physical parameters of tissues through signal comparison in a multi-parameter space. The core of MRF technology lies in using a pre-set scan sequence library to perform dictionary matching with the signals of tissue responses, so as to simultaneously obtain multiple quantitative parameters.

[0004] However, in the process of using the existing technology, the inventor found that there are at least the following problems in the existing technology: MRF technology mainly focuses on imaging magnetic resonance physical properties such as longitudinal relaxation time T1 and transverse relaxation time T2, and has limited direct reconstruction ability for physiological characteristics such as water content and electrical properties. In addition, MRF technology relies on the method of signal comparison and is very sensitive to noise. In a complex tissue environment, such as the brain with complex and delicate structures or ultra-high field MRI (Magnetic Resonance Imaging) with a more complex electromagnetic environment, it will lead to large parameter estimation errors. Summary of the Invention

[0005] The present invention aims to solve the above technical problems at least to a certain extent, and provides a quantitative magnetic resonance imaging method, system, electronic device and product.

[0006] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a quantitative magnetic resonance imaging method, including: Obtaining magnetic resonance signals of a sample to be measured obtained by magnetic resonance scanning; Input the magnetic resonance signal into a preset first deep learning network to obtain the moisture content distribution map of the sample to be measured and the sensitivity information of the magnetic resonance radio frequency receiving coil during magnetic resonance scanning, and obtain the amplitude map of the multi-channel radio frequency receiving magnetic field according to the sensitivity information of the magnetic resonance radio frequency receiving coil; Obtain the T1 relaxation parameter map of the sample to be measured and the first complex electrical property distribution map based on the moisture content according to the moisture content distribution map; Collect the relative phase map of the radio frequency receiving field of the radio frequency receiving coil from the magnetic resonance signal, combine the relative phase map and the amplitude map into a radio frequency receiving field vector, and then input the radio frequency receiving field vector into a preset second deep learning network to obtain the second complex electrical property distribution map of the sample to be measured based on the radio frequency receiving field vector; Obtain the final complex electrical property distribution map of the sample to be measured according to the first complex electrical property distribution map and the second complex electrical property distribution map.

[0007] In a possible design, when the magnetic field strength during magnetic resonance scanning is 3T, the T1 relaxation parameter at any pixel point in the T1 relaxation parameter map of the sample to be measured is: T1 = (522 * PD_py) / (1 - PD_py * 0.858); where PD_py is the moisture content at the any pixel point in the moisture content distribution map; The dielectric constant at any pixel point in the first complex electrical property distribution map is: per = p1×PD_py^2 + p2×PD_py + p3; where p1, p2, and p3 are preset fitting coefficients, and p1 = -287, p2 = 591, p3 = -220; The conductivity at any pixel point in the first complex electrical property distribution map is: con = c1 + c2×exp(c3×PD_py); where c1, c2, and c3 are preset fitting coefficients, and c1 = 0.286, c2 = 1.526×10^-5, c3 = 11.852.

[0008] In a possible design, when the magnetic field strength during magnetic resonance scanning is 7T, the T1 relaxation parameter at any pixel point in the T1 relaxation parameter map of the sample to be measured is: T1 = (522 * PD_py) / (1 - PD_py * 0.858); where PD_py is the moisture content at the any pixel point in the moisture content distribution map; The dielectric constant at any pixel point in the first complex electrical property distribution map is: per = p1'×PD_py^2 + p2'×PD_py + p3'; In the formula, p1', p2', and p3' are preset fitting coefficients, and p1' = -230.5, p2' = 472.5, p3' = -169.6; The conductivity at any pixel point in the first complex electrical property distribution map is: con = c1' + c2'×exp(c3'×PD_py); In the formula, c1', c2', and c3' are preset fitting coefficients, and c1' = -0.6855, c2' = 0.07955, c3' = 3.651.

[0009] In a possible design, the second deep learning network adopts a physically coupled complex convolutional neural network; the second deep learning network includes a complex input layer, a complex convolutional layer, a complex BN layer, a complex activation layer, a complex pooling layer, and a complex output layer connected in sequence.

[0010] In a possible design, the expression of the loss function adopted in the second deep learning network is as follows: ; In the formula, a is a preset first weight coefficient; b is a preset second weight coefficient; c is a preset third weight coefficient; is the mean square error loss value; is the gradient loss value; is the physical loss value constrained by the Helmholtz equation; Among them, the mean square error loss value is: ; In the formula, N is the total number of pixel points of the reference complex electrical property distribution map corresponding to the sample image input to the second deep learning network; is the i th pixel point of the reference complex electrical property distribution map; is the i th pixel point of the predicted complex electrical property distribution map obtained by the second deep learning network according to the sample image; The gradient loss value is: ; In the formula, is the reference complex electrical property distribution map, is the reference complex electrical property distribution map thi The gradient component of a pixel point in the x direction, is the th i gradient component of the pixel point in the y direction of the reference complex electrical property distribution map; is the predicted complex electrical property distribution map, is the predicted complex electrical property distribution map th i gradient component of the pixel point in the x direction of the predicted complex electrical property distribution map, is the predicted complex electrical property distribution map th i gradient component of the pixel point in the y direction of the predicted complex electrical property distribution map; The Helmholtz equation corresponding to the physical loss value constrained by the Helmholtz equation is: ; In the formula, σ is the predicted conductivity; ω is the preset angular frequency; ε is the predicted relative permittivity; μ is the preset vacuum permittivity; j is the imaginary symbol; is the RF receiving field vector of the multi-channel RF receiving magnetic field; The physical loss value constrained by the Helmholtz equation is: .

[0011] In a possible design, according to the first complex electrical property distribution map and the second complex electrical property distribution map, obtaining the final complex electrical property distribution map of the sample to be measured includes: Obtaining first gradient information of the first complex electrical property distribution map; Obtaining second gradient information of the second complex electrical property distribution map; According to the first gradient information and the second gradient information, obtaining the final complex electrical property distribution map of the sample to be measured.

[0012] In a possible design, in the first gradient information, the gradient value of the first sub-gradient information of any pixel point located in the first complex electrical property distribution map is: grad_A = |gx| + |gy|; In the formula, |gx| is the absolute value of the gradient component of any pixel point in the x direction of the first complex electrical property distribution map; |gy| is the absolute value of the gradient component of any pixel point in the y direction of the first complex electrical property distribution map; In the second gradient information, the gradient value of the second sub-gradient information of any pixel point located in the second complex electrical property distribution map is: grad_B = |px| + |py|; In the formula, |px| is the absolute value of the gradient component of any pixel point of the second complex electrical property distribution map in the x direction; |py| is the absolute value of the gradient component of any pixel point of the second complex electrical property distribution map in the y direction; The final electrical property parameter of any pixel point of the final complex electrical property distribution map of the sample to be measured is: EPs = w × B + (1 - w) × A; In the formula, A is the first electrical property parameter of any pixel point of the first complex electrical property distribution map; B is the second electrical property parameter of any pixel point of the second complex electrical property distribution map; w is a preset weight, and w = min(0.7, max(grad_A / (grad_B + eps), 0.3)), eps is a preset minimum value hyperparameter, eps = 0.01.

[0013] In a second aspect, the present invention provides a quantitative magnetic resonance imaging system for implementing a quantitative magnetic resonance imaging method as described in any one of the above; the quantitative magnetic resonance imaging system includes: A signal acquisition module that acquires magnetic resonance signals obtained by magnetic resonance scanning of a sample to be measured; A first signal processing module, communicatively connected to the signal acquisition module, for inputting the magnetic resonance signals into a preset first deep learning network to obtain a water content distribution map of the sample to be measured and sensitivity information of a magnetic resonance radio frequency receiving coil during magnetic resonance scanning, and obtaining an amplitude map of a multi-channel radio frequency receiving magnetic field according to the sensitivity information of the magnetic resonance radio frequency receiving coil; A second signal processing module, communicatively connected to the first signal processing module, for obtaining a T1 relaxation parameter map of the sample to be measured and a first complex electrical property distribution map based on the water content according to the water content distribution map; A third signal processing module, communicatively connected to the first signal processing module, for collecting a relative phase map of a radio frequency receiving field of a radio frequency receiving coil from the magnetic resonance signals, combining the relative phase map and the amplitude map into a radio frequency receiving field vector, and then inputting the radio frequency receiving field vector into a preset second deep learning network to obtain a second complex electrical property distribution map of the sample to be measured based on the radio frequency receiving field vector; A fourth signal processing module, communicatively connected to the second signal processing module and the third signal processing module respectively, for obtaining a final complex electrical property distribution map of the sample to be measured according to the first complex electrical property distribution map and the second complex electrical property distribution map.

[0014] In a third aspect, the present invention provides an electronic device, including: A memory for storing computer program instructions; and, A processor for executing the computer program instructions to complete the operations of a quantitative magnetic resonance imaging method as described in any one of the above.

[0015] In a fourth aspect, the present invention provides a computer program product including a computer program or instructions, where the computer program or the instructions, when executed by a computer, implement a quantitative magnetic resonance imaging method as described in any one of the above.

[0016] The beneficial effects of the present invention are as follows: The present invention discloses a quantitative magnetic resonance imaging method, system, electronic device, and product, which expand the imaging range of quantitative magnetic resonance imaging and can improve the accuracy of the quantitative magnetic resonance imaging results. Specifically, during the implementation of the present invention, after obtaining the moisture content distribution map of the sample to be measured and the amplitude map of the multi-channel radio frequency receiving magnetic field according to the magnetic resonance signal, the T1 relaxation parameter map of the sample to be measured and the first complex electrical property distribution map based on the moisture content can be obtained according to the moisture content distribution map; then, the amplitude map of the multi-channel radio frequency receiving magnetic field is input into a preset second deep learning network to obtain the second complex electrical property distribution map of the sample to be measured based on the radio frequency receiving field vector, and the final complex electrical property distribution map of the sample to be measured is obtained according to the first complex electrical property distribution map and the second complex electrical property distribution map. During this process, through a single magnetic resonance scan of the sample to be measured and based on the settings of the first deep learning network and the second deep learning network, quantitative magnetic resonance images of various physiological parameters such as the moisture content distribution map, T1 relaxation parameter map, and final complex electrical property distribution map of human tissues can be quantitatively reconstructed, expanding the imaging range of quantitative magnetic resonance imaging; at the same time, since the final complex electrical property distribution map is obtained based on the first complex electrical property distribution map based on the moisture content and the second complex electrical property distribution map based on the radio frequency receiving field vector, combining the advantage of clear boundary recognition of the first complex electrical property distribution map and the advantage of more accurate values calculated within the same tissue of the second complex electrical property distribution map, the reconstruction result is clearer, the accuracy is improved compared with traditional MRF technology or traditional electrical property reconstruction technology, and it also has a certain effect of resisting noise interference.

[0017] Other beneficial effects of the present invention will be further described in the specific implementation manner. Description of the Drawings

[0018] Figure 1 is a flowchart of the quantitative magnetic resonance imaging method in Embodiment 1; Figure 2 is a block diagram of the modules of the quantitative magnetic resonance imaging system in Embodiment 2; Figure 3 It is a block diagram of a module of an electronic device in Embodiment 3. Specific implementation manner

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the accompanying drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention.

[0020] Embodiment 1: This embodiment discloses a quantitative magnetic resonance imaging method, which can be, but is not limited to, executed by a computer device or a virtual machine with certain computing resources, such as an electronic device such as a personal computer, a smart phone, a personal digital assistant, or a wearable device, or executed by a virtual machine.

[0021] As Figure 1 shown, a quantitative magnetic resonance imaging method can include, but is not limited to, the following steps: S1. Obtain the magnetic resonance signal obtained by magnetic resonance scanning of the sample to be measured. S2. Input the magnetic resonance signal into a preset first deep learning network to obtain the moisture content distribution map of the sample to be measured and the sensitivity information of the magnetic resonance radio frequency receiving coil during magnetic resonance scanning, and obtain the amplitude map of the multi-channel radio frequency receiving magnetic field according to the sensitivity information of the magnetic resonance radio frequency receiving coil. This amplitude map is also the amplitude information of the multi-channel radio frequency receiving magnetic field, which can be abbreviated with an absolute value symbol as: ; Specifically, in step S2 of this embodiment, the first deep learning network uses a Nested-unet network constrained by the coupled Helmholtz equation. Its specific settings and the specific process of obtaining the moisture content distribution map of the sample to be measured and the sensitivity information of the magnetic resonance radio frequency receiving coil during magnetic resonance scanning from the magnetic resonance signal can all be implemented by the "Method, System, Electronic Device and Medium for Measuring the Moisture Content of Human Tissues" disclosed in the Chinese patent with the publication number CN117653026A, and will not be elaborated here.

[0022] It should be noted that each receiving coil in the multi-channel radio frequency receiving system has a different sensitivity distribution, which directly reflects the induction ability of the receiving coil to the radio frequency receiving magnetic field (B1-field). During the implementation of this embodiment, the sensitivity information obtained from multiple receiving channels can be used, combined with common B1-field mapping methods such as the Double Angle Method (DAM), the Bloch-Siegert frequency shift method, or the Echo Time Method, to reconstruct the amplitude map of the radio frequency receiving magnetic field in the entire imaging area, that is, the amplitude map of the multi-channel radio frequency receiving magnetic field.

[0023] S3. Obtain the T1 relaxation parameter map of the sample to be measured and the first complex electrical property distribution map based on the moisture content from the moisture content distribution map; it should be understood that the real part of the complex electrical property corresponds to the conductivity, and the imaginary part corresponds to the dielectric constant.

[0024] In one implementation, as in step S1, when the magnetic field strength during magnetic resonance scanning is 3T, in step S3, the T1 relaxation parameter at any pixel point in the T1 relaxation parameter map of the sample to be measured is: T1 = (522 * PD_py) / (1 - PD_py * 0.858); Wherein, PD_py is the moisture content at the any pixel point in the moisture content distribution map; In step S3, the dielectric constant at any pixel point in the first complex electrical property distribution map is: per = p1 × PD_py^2 + p2 × PD_py + p3; Wherein, p1, p2, and p3 are preset fitting coefficients, and p1 = -287, p2 = 591, p3 = -220; The conductivity at any pixel point in the first complex electrical property distribution map is: con = c1 + c2 × exp(c3 × PD_py); Wherein, c1, c2, and c3 are preset fitting coefficients, and c1 = 0.286, c2 = 1.526 × 10^-5, c3 = 11.852.

[0025] That is, in the scenario where the magnetic field strength is 3T, the first electrical property distribution map of the sample to be measured based on the moisture content is obtained by using the fitting equations per = p1 × PD_py^2 + p2 × PD_py + p3 and con = c1 + c2 × exp(c3 × PD_py).

[0026] In another embodiment, as in step S1, when the magnetic field strength during magnetic resonance scanning is 7T, in step S3, the T1 relaxation parameter at any pixel point in the T1 relaxation parameter map of the sample to be measured is: T1 = (522 * PD_py) / (1 - PD_py * 0.858); In the formula, PD_py is the moisture content at the any pixel point in the moisture content distribution map; The dielectric constant at any pixel point in the first complex electrical property distribution map is: per = p1' × PD_py^2 + p2' × PD_py + p3'; In the formula, p1', p2' and p3' are preset fitting coefficients, and p1' = -230.5, p2' = 472.5, p3' = -169.6; The conductivity at any pixel point in the first complex electrical property distribution map is: con = c1' + c2' × exp(c3' × PD_py); In the formula, c1', c2' and c3' are preset fitting coefficients, and c1' = -0.6855, c2' = 0.07955, c3' = 3.651.

[0027] That is, in the scenario where the magnetic field strength is 7T, the first complex electrical property distribution map of the sample to be measured based on the moisture content is obtained by using the fitting equations per = p1' × PD_py^2 + p2' × PD_py + p3' and con = c1' + c2' × exp(c3' × PD_py).

[0028] It should be noted that in this embodiment, when the magnetic field strength during magnetic resonance scanning is 3T or 7T, the calculation method of the T1 relaxation parameter at any pixel point in the T1 relaxation parameter map of the sample to be measured is the same, and the calculation expressions of the dielectric constant and conductivity of each pixel point in the first complex electrical property distribution map are the same, but the fitting coefficients are different.

[0029] S4. Acquire the relative phase map of the radio frequency receiving field of the radio frequency receiving coil from the magnetic resonance signal, combine the relative phase map and the amplitude map into a radio frequency receiving field vector, and then input the radio frequency receiving field vector into a preset second deep learning network to obtain the second complex electrical property distribution map of the sample to be measured based on the radio frequency receiving field vector; it should be understood that step S3 and step S4 in this embodiment are parallel steps executed in parallel.

[0030] It should be noted that in this embodiment, the first deep learning network and the second deep learning network are obtained after pre-training. Specifically, they are trained with different simulation data with known true values, and are obtained after being compared with the true values of the simulation data and passing the performance test.

[0031] In step S4 of this embodiment, the second deep learning network adopts a physically coupled complex convolutional neural network, which is obtained by combining a traditional physical model with data-driven deep learning technology. It can adopt a Unet network (residual network) or a ResNet network (convolutional neural network for image segmentation), or other convolutional neural networks, which are not limited here; specifically, the second deep learning network includes a complex input layer, a complex convolutional layer, a complex BN (Batch Normalization) layer, a complex activation layer, a complex pooling layer, and a complex output layer connected in sequence.

[0032] It should be noted that the second deep learning network deeply couples the physical processes of vector operation and magnetic field propagation, enhances the parsing ability of the second deep learning network for complex magnetic field signals, significantly improves the generalization ability and accuracy of the second deep learning network, and solves the interpretability problem of deep learning technology for medical images to a certain extent.

[0033] In step S4, the expression of the loss function adopted in the second deep learning network is as follows: ; In the formula, a is a preset first weight coefficient; b is a preset second weight coefficient; c is a preset third weight coefficient; is the mean square error loss value; is the gradient loss value; is the physical loss value constrained by the Helmholtz equation; It should be noted that during the training of the second deep learning network, a B1-vector simulation sample image with an electrical property distribution label is input into the second deep learning network (this sample image is correspondingly provided with a calibrated complex electrical property distribution map, which is herein referred to as the reference complex electrical property distribution map or the true complex electrical property distribution map). The B1-vector simulation sample image is also the amplitude map of the exemplary multi-channel radio frequency received magnetic field, so that after the training is completed, the second deep learning network is used to predict any scanned amplitude map of the multi-channel radio frequency received magnetic field during the prediction process, and a predicted second complex electrical property distribution map is obtained, which is herein referred to as the predicted complex electrical property distribution map. The loss function is used to measure the error between the predicted complex electrical property distribution map and the actual image, that is, the reference complex electrical property distribution map, so as to adjust the parameters in the second deep learning network, enabling it to be continuously optimized during the training process and reducing the difference between the predicted complex electrical property distribution map and the reference complex electrical property distribution map.

[0034] Among them, the mean square error loss value is: ; In the formula, N is the total number of pixel points of the reference complex electrical property distribution map corresponding to the sample image input into the second deep learning network; is the reference electrical property parameter of the i th pixel point in the reference complex electrical property distribution map; is the predicted electrical property parameter of the i th pixel point in the predicted complex electrical property distribution map obtained by the second deep learning network according to the sample image; It should be noted that the mean square error loss value is used to measure the mean square error between the reference complex electrical property distribution map and the predicted complex electrical property distribution map. The smaller its value, the closer the prediction result of the second deep learning network is to the reference complex electrical property distribution map, and the higher the prediction accuracy.

[0035] The gradient loss value is: ; In the formula, is the reference complex electrical property distribution map, is the th i pixel point of the reference complex electrical property distribution map in the x-direction gradient component, is the th i pixel point of the reference complex electrical property distribution map in the y-direction gradient component; is the predicted complex electrical property distribution map, that is, the predicted complex electrical property distribution image generated by the second deep learning network based on the input sample image. is the predicted complex electrical property distribution map The i -th pixel point of the predicted complex electrical property distribution map in the x-direction gradient component, i is the predicted complex electrical property distribution map The predicted complex electrical property distribution map The predicted complex electrical property distribution map The y-direction gradient component of the -th pixel point; It should be noted that the gradient loss value is used to measure the difference in gradients (such as edge information) between the reference complex electrical property distribution map

[0036] and the predicted complex electrical property distribution map ; In the formula, σ is the predicted conductivity; ω is the preset angular frequency; ε is the predicted relative permittivity; σ , ω and ε are all preset values. μ is the preset vacuum permittivity; j is the imaginary unit symbol; is the radio frequency receiving field vector of the multi-channel radio frequency receiving magnetic field. It should be understood that The conjugate variable of is the radio frequency receiving magnetic field in the Cartesian coordinate system is the Laplace operator; represents the divergence operation of the parameter ; represents the curl operation of ;

[0037] The physical loss value constrained by the Helmholtz equation is: .

[0038] It should be understood that the physical loss value is obtained by taking the absolute value of the Helmholtz equation.

[0039] S5. Obtain the final complex electrical property distribution map of the sample to be measured according to the first complex electrical property distribution map and the second complex electrical property distribution map. It should be noted that in this embodiment, by performing weighted average processing on the first complex electrical property distribution map and the second complex electrical property distribution map, a complex electrical property distribution map that combines the two processing methods and has higher accuracy can be obtained.

[0040] Specifically, in step S5, obtaining the final complex electrical property distribution map of the sample to be measured according to the first complex electrical property distribution map and the second complex electrical property distribution map includes: S501. Obtain the first gradient information of the first complex electrical property distribution map; specifically, the first gradient information includes the first sub-gradient information of each pixel point in the first complex electrical property distribution map, and the first sub-gradient information of any pixel point includes its gradient component in the x direction and its gradient component in the y direction.

[0041] Specifically, in the first gradient information, the gradient value of the first sub-gradient information of any pixel point located in the first complex electrical property distribution map is: grad_A = |gx| + |gy|; In the formula, |gx| is the absolute value of the gradient component of any pixel point in the first complex electrical property distribution map in the x direction, and this gradient component is the change rate of any pixel point in the first complex electrical property distribution map in the x direction; |gy| is the absolute value of the gradient component of any pixel point in the first complex electrical property distribution map in the y direction, and this gradient component is the change rate of any pixel point in the first complex electrical property distribution map in the y direction. It should be understood that |gx| and |gy| are the absolute values of the gradient components in different directions of the same pixel point.

[0042] S502. Obtain the second gradient information of the second complex electrical property distribution map; specifically, the second gradient information includes the second sub-gradient information of each pixel point in the second complex electrical property distribution map, and the second sub-gradient information of any pixel point includes its gradient component in the x direction and its gradient component in the y direction.

[0043] Specifically, in the second gradient information, the gradient value of the second sub-gradient information of any pixel point located in the second complex electrical property distribution map is: grad_B = |px| + |py|; Wherein, |px| is the absolute value of the gradient component of any pixel point of the second complex electrical property distribution map in the x direction, and this gradient component is the change rate of any pixel point of the second complex electrical property distribution map in the x direction; |py| is the absolute value of the gradient component of any pixel point of the second complex electrical property distribution map in the y direction, and this gradient component is the change rate of any pixel point of the second complex electrical property distribution map in the y direction.

[0044] S503. Obtain the final complex electrical property distribution map of the sample to be measured according to the first gradient information and the second gradient information.

[0045] Specifically, the final electrical property parameter of any pixel point of the final complex electrical property distribution map of the sample to be measured is: EPs = w × B + (1 - w) × A; Wherein, A is the first electrical property parameter of any pixel point of the first complex electrical property distribution map; B is the second electrical property parameter of any pixel point of the second complex electrical property distribution map; w is a preset weight, and w = min(0.7, max(grad_A / (grad_B + eps), 0.3)), eps is a preset minimum value hyperparameter, and eps = 0.01. It should be noted that the setting of the minimum value hyperparameter eps here is aimed at preventing the occurrence of a denominator bias term of division by zero to prevent the problem of division by zero error.

[0046] In this embodiment, taking the magnitudes of the gradient information as weights, the values of the first complex electrical property distribution map and the second complex electrical property distribution map are weighted and averaged. The adopted weighted average formula has a boundary recognition function, which can facilitate the recognition of tissue boundaries and relatively uniform regions inside the tissue, and process them separately, so as to obtain the final complex electrical property distribution map of the sample to be measured.

[0047] This embodiment expands the imaging range of quantitative magnetic resonance imaging and can improve the accuracy of the results of quantitative magnetic resonance imaging. Specifically, during the implementation of this embodiment, after obtaining the water content distribution map of the sample to be measured and the amplitude map of the multi-channel radio frequency receiving magnetic field according to the magnetic resonance signal, the T1 relaxation parameter map of the sample to be measured and the first complex electrical property distribution map based on the water content can be obtained according to the water content distribution map; then, the amplitude map of the multi-channel radio frequency receiving magnetic field is input into a preset second deep learning network to obtain the second complex electrical property distribution map of the sample to be measured based on the radio frequency receiving field vector, and the final complex electrical property distribution map of the sample to be measured is obtained according to the first complex electrical property distribution map and the second complex electrical property distribution map. During this process, after performing a magnetic resonance scan on the sample to be measured once, based on the settings of the first deep learning network and the second deep learning network, quantitative magnetic resonance images of various physiological parameters such as the water content distribution map, the T1 relaxation parameter map, and the final complex electrical property distribution map of human tissues can be quantitatively reconstructed, expanding the imaging range of quantitative magnetic resonance imaging; at the same time, since the final complex electrical property distribution map is obtained from the first complex electrical property distribution map based on the water content and the second complex electrical property distribution map based on the radio frequency receiving field vector, combining the advantages of clear boundary recognition of the first complex electrical property distribution map and more accurate values calculated within the same tissue of the second complex electrical property distribution map, the reconstruction result is clearer, and the accuracy is improved compared with traditional MRF technology or traditional electrical property reconstruction technology, and at the same time, it has a certain effect of resisting noise interference.

[0048] It should also be noted that this embodiment can further improve the efficiency of quantitative magnetic resonance imaging. Specifically, this embodiment directly uses common multi-channel receiving coils in clinics, and does not limit the specific number of channels or size. The function of quantitative magnetic resonance imaging can be realized on various magnetic resonance imaging devices (such as different field strengths, 1.5T / 3T / 7T; different numbers of receiving channels, 20 channels / 32 channels). For example, quantitative imaging can be performed on any clinical MRI, providing more imaging modalities, rather than being limited to a certain scientific research MRI machine, greatly improving the feasibility and efficiency of clinical applications. In addition, this embodiment does not require additional calibration steps (such as performing magnetic resonance scans on the calibrated water model and the sample to be measured at the same time, and using the scan results of the water phantom as a reference to calibrate multiple quantitative magnetic resonance images such as the water content distribution map and the final complex electrical property distribution map of the sample to be measured). Only by carefully analyzing the multi-channel received magnetic resonance signals, the multiple degrees of freedom information provided by the multi-channel coil can be fully exploited, further improving the efficiency of quantitative magnetic resonance imaging while ensuring the accuracy of quantitative imaging.

[0049] Embodiment 2: This embodiment discloses a quantitative magnetic resonance imaging system for implementing the quantitative magnetic resonance imaging method in Embodiment 1; as Figure 2 shown, the quantitative magnetic resonance imaging system includes: A signal acquisition module that acquires magnetic resonance signals obtained by magnetic resonance scanning of a sample to be measured; A first signal processing module, communicatively connected to the signal acquisition module, for inputting the magnetic resonance signals into a preset first deep learning network to obtain a moisture content distribution map of the sample to be measured and sensitivity information of a magnetic resonance radio frequency receiving coil during magnetic resonance scanning, and obtaining an amplitude map of a multi-channel radio frequency receiving magnetic field based on the sensitivity information of the magnetic resonance radio frequency receiving coil; A second signal processing module, communicatively connected to the first signal processing module, for obtaining a T1 relaxation parameter map of the sample to be measured and a first complex electrical property distribution map based on the moisture content according to the moisture content distribution map; A third signal processing module, communicatively connected to the first signal processing module, for acquiring a relative phase map of a radio frequency receiving field of a radio frequency receiving coil from the magnetic resonance signals, combining the relative phase map and the amplitude map into a radio frequency receiving field vector, and then inputting the radio frequency receiving field vector into a preset second deep learning network to obtain a second complex electrical property distribution map of the sample to be measured based on the radio frequency receiving field vector; A fourth signal processing module, communicatively connected to the second signal processing module and the third signal processing module respectively, for obtaining a final complex electrical property distribution map of the sample to be measured according to the first complex electrical property distribution map and the second complex electrical property distribution map.

[0050] It should be noted that for the working process, working details and technical effects of the quantitative magnetic resonance imaging system provided in this Embodiment 2, reference can be made to Embodiment 1, which will not be elaborated here.

[0051] Embodiment 3: Based on Embodiment 1 or 2, this embodiment discloses an electronic device, which may be a smart phone, a tablet computer, a notebook computer or a desktop computer, etc. The electronic device may be referred to as a user terminal, a portable terminal, a desktop terminal, etc., as Figure 3 shown, the electronic device includes: A memory for storing computer program instructions; and, A processor for executing the computer program instructions to complete the operations of any one of the quantitative magnetic resonance imaging methods described in Embodiment 1.

[0052] Specifically, the processor 301 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. The processor 301 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor 301 may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake state, also known as the CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 301 may be integrated with a GPU (Graphics Processing Unit), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen.

[0053] The memory 302 may include one or more computer-readable storage media, and the computer-readable storage media may be non-transitory. The memory 302 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices and flash storage devices. In some embodiments, the non-transitory computer-readable storage media in the memory 302 is used to store at least one instruction, and the at least one instruction is used to be executed by the processor 301 to implement the quantitative magnetic resonance imaging method provided in Embodiment 1 of this application.

[0054] In some embodiments, the terminal may also optionally include: a communication interface 303 and at least one peripheral device. The processor 301, the memory 302, and the communication interface 303 may be connected through a bus or signal lines. Each peripheral device may be connected to the communication interface 303 through a bus, signal lines, or a circuit board. Specifically, the peripheral device includes at least one of a radio frequency circuit 304, a display screen 305, and a power supply 306.

[0055] The communication interface 303 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 301 and the memory 302. In some embodiments, the processor 301, the memory 302, and the communication interface 303 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 301, the memory 302, and the communication interface 303 may be implemented on a separate chip or circuit board, and this embodiment does not limit this.

[0056] The radio frequency circuit 304 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 304 communicates with the communication network and other communication devices through electromagnetic signals.

[0057] The display screen 305 is used to display the UI (User Interface). The UI can include any combination of graphics, text, icons, and videos.

[0058] The power supply 306 is used to supply power to each component in the electronic device.

[0059] Embodiment 4: Based on any one of Embodiments 1 to 3, this embodiment discloses a computer program product, including a computer program or instruction, and the computer program or the instruction, when executed by a computer, implements a quantitative magnetic resonance imaging method as described in any one of Embodiments 1. Wherein, the computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0060] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module to implement. In this way, the present invention is not limited to any specific combination of hardware and software.

[0061] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the present invention in each embodiment.

Claims

1. A quantitative magnetic resonance imaging method, characterized in that: include: Acquire a magnetic resonance signal of the sample to be tested obtained by magnetic resonance scanning; Inputting the magnetic resonance signal into a preset first deep learning network, obtaining a moisture content distribution diagram of the sample to be tested and sensitivity information of the magnetic resonance radio frequency receiving coil during magnetic resonance scanning, and obtaining an amplitude diagram of a multi-channel radio frequency receiving magnetic field according to the sensitivity information of the magnetic resonance radio frequency receiving coil; Obtaining a T1 relaxation parameter map of the sample to be tested and a first complex electrical characteristic distribution map based on the moisture content according to the moisture content distribution map; Acquire a relative phase diagram of the radio frequency receiving field of the radio frequency receiving coil from the magnetic resonance signal, combine the relative phase diagram and the amplitude diagram into a radio frequency receiving field vector, and then input the radio frequency receiving field vector into a preset second deep learning network to obtain a second complex electrical characteristic distribution diagram of the sample to be tested based on the radio frequency receiving field vector; A final complex electrical characteristic distribution map of the sample to be tested is obtained according to the first complex electrical characteristic distribution map and the second complex electrical characteristic distribution map.

2. A quantitative magnetic resonance imaging method according to claim 1, characterized in that: When the magnetic field intensity during magnetic resonance scanning is 3T, the T1 relaxation parameter at any pixel point in the T1 relaxation parameter map of the sample to be tested is: T1=(522*PD_py) / (1-PD_py*0.858); Wherein, PD_py is the moisture content at any pixel point in the moisture content distribution map; The dielectric constant at any pixel point in the first complex electrical characteristic distribution graph is: per=p1×PD_py^2+p2×PD_py+p3; Where p1, p2 and p3 are preset fitting coefficients, and p1=-287, p2=591, p3=-220; The conductivity at any pixel point in the first complex electrical characteristic distribution graph is: con = c1 + c2 × exp (c3 × PD_py); Wherein, c1, c2 and c3 are preset fitting coefficients, and c1=0.286, c2=1.526×10^-5, c3=11.

852.

3. A quantitative magnetic resonance imaging method according to claim 1, characterized in that: When the magnetic field intensity during magnetic resonance scanning is 7T, the T1 relaxation parameter at any pixel point in the T1 relaxation parameter map of the sample to be tested is: T1=(522*PD_py) / (1-PD_py*0.858); Wherein, PD_py is the moisture content at any pixel point in the moisture content distribution map; The dielectric constant at any pixel point in the first complex electrical characteristic distribution graph is: per=p1'×PD_py^2+p2'×PD_py+p3'; Wherein, p1', p2' and p3' are preset fitting coefficients, and p1'=-230.5, p2'=472.5, p3'=-169.6; The conductivity at any pixel point in the first complex electrical characteristic distribution graph is: con=c1'+c2'×exp(c3'×PD_py); Wherein, c1', c2' and c3' are preset fitting coefficients, and c1'=-0.6855, c2'=0.07955, c3'=3.

651.

4. A quantitative magnetic resonance imaging method according to claim 1, characterized in that: The second deep learning network adopts a physically coupled complex convolutional neural network; the second deep learning network includes a complex input layer, a complex convolutional layer, a complex BN layer, a complex activation layer, a complex pooling layer and a complex output layer connected in sequence.

5. A quantitative magnetic resonance imaging method according to claim 1, characterized in that: The expression of the loss function used in the second deep learning network is as follows: ； In the formula, a is the preset first weight coefficient; b is the preset second weight coefficient; c is the preset third weight coefficient; is the mean square error loss value; is the gradient loss value; is the physical loss value constrained by the Helmholtz equation; Among them, the mean square error loss value is: ； In the formula, N is the total number of pixels of the reference complex electrical characteristic distribution map corresponding to the sample image input into the second deep learning network; is the reference complex electrical characteristic distribution diagram i Reference electrical characteristic parameters of each pixel; is the first in the predicted complex electrical characteristic distribution map obtained by the second deep learning network according to the sample image i The predicted electrical characteristic parameters of each pixel; The gradient loss value is: ； In the formula, is the reference complex electrical characteristic distribution diagram, The reference complex electrical characteristic distribution diagram No. i The gradient component of a pixel in the x direction, The reference complex electrical characteristic distribution diagram No. i The gradient component of a pixel in the y direction; is the predicted complex electrical characteristic distribution diagram, The predicted complex electrical characteristic distribution map No. i The gradient component of a pixel in the x direction, The predicted complex electrical characteristic distribution map No. i The gradient component of a pixel in the y direction; The Helmholtz equation corresponding to the physical loss value constrained by the Helmholtz equation is: ； In the formula, σ is the predicted conductivity; ω is the preset angular frequency; ε is the predicted dielectric constant; μ is the preset vacuum dielectric constant; j is the imaginary number symbol; is the RF receiving field vector of the multi-channel RF receiving magnetic field; The physical loss value of the Helmholtz equation constraint is: 。 6. A quantitative magnetic resonance imaging method according to claim 1, characterized in that: Obtaining a final complex electrical characteristic distribution map of the sample to be tested according to the first complex electrical characteristic distribution map and the second complex electrical characteristic distribution map, comprising: Acquire first gradient information of the first complex electrical characteristic distribution diagram; Acquire second gradient information of the second complex electrical characteristic distribution diagram; A final complex electrical characteristic distribution diagram of the sample to be tested is obtained according to the first gradient information and the second gradient information.

7. A quantitative magnetic resonance imaging method according to claim 6, characterized in that: In the first gradient information, the gradient value of the first sub-gradient information located at any pixel point of the first complex electrical characteristic distribution graph is: grad_A=|gx|+|gy|; Wherein, |gx| is the absolute value of the gradient component of any pixel point of the first complex electrical characteristic distribution map in the x direction; |gy| is the absolute value of the gradient component of any pixel point of the first complex electrical characteristic distribution map in the y direction; In the second gradient information, the gradient value of the second sub-gradient information located at any pixel point of the second complex electrical characteristic distribution diagram is: grad_B=|px|+|py|; Wherein, |px| is the absolute value of the gradient component of any pixel point of the second complex electrical characteristic distribution map in the x direction; |py| is the absolute value of the gradient component of any pixel point of the second complex electrical characteristic distribution map in the y direction; The final electrical characteristic parameter of any pixel point of the final complex electrical characteristic distribution diagram of the sample to be tested is: EPs = w × B + (1-w) × A; In the formula, A is the first electrical characteristic parameter of any pixel point of the first complex electrical characteristic distribution map; B is the second electrical characteristic parameter of any pixel point of the second complex electrical characteristic distribution map; w is the preset weight, and w=min(0.7,max(grad_A / (grad_B+eps),0.3)), eps is the preset minimum hyperparameter, eps=0.

01.

8. A quantitative magnetic resonance imaging system, characterized in that: Used to implement a quantitative magnetic resonance imaging method as described in any one of claims 1 to 7; the quantitative magnetic resonance imaging system comprises: A signal acquisition module, which acquires a magnetic resonance signal of the sample to be tested obtained by magnetic resonance scanning; a first signal processing module, which is in communication with the signal acquisition module and is used to input the magnetic resonance signal into a preset first deep learning network, obtain a moisture content distribution diagram of the sample to be tested and sensitivity information of the magnetic resonance radio frequency receiving coil during magnetic resonance scanning, and obtain an amplitude diagram of a multi-channel radio frequency receiving magnetic field according to the sensitivity information of the magnetic resonance radio frequency receiving coil; A second signal processing module, which is in communication with the first signal processing module and is used to obtain a T1 relaxation parameter map of the sample to be tested and a first complex electrical characteristic distribution map based on the moisture content according to the moisture content distribution map; A third signal processing module is communicatively connected with the first signal processing module, and is used for acquiring a relative phase diagram of the radio frequency receiving field of the radio frequency receiving coil from the magnetic resonance signal, and combining the relative phase diagram and the amplitude diagram into a radio frequency receiving field vector, and then inputting the radio frequency receiving field vector into a preset second deep learning network to obtain a second complex electrical characteristic distribution diagram of the sample to be tested based on the radio frequency receiving field vector; The fourth signal processing module is respectively connected to the second signal processing module and the third signal processing module for obtaining a final complex electrical characteristic distribution diagram of the sample to be tested according to the first complex electrical characteristic distribution diagram and the second complex electrical characteristic distribution diagram.

9. An electronic device, characterized in that: include: a memory for storing computer program instructions; as well as, A processor, configured to execute the computer program instructions to perform the operation of a quantitative magnetic resonance imaging method as claimed in any one of claims 1 to 7.

10. A computer program product comprising a computer program or instructions, characterized in that The computer program or the instructions implement a quantitative magnetic resonance imaging method as claimed in any one of claims 1 to 7 when executed by a computer.

Images