A method and apparatus for generating a magnetic resonance quantitative physiological parameter map

Abstract

The embodiments of this specification provide a method and device for generating magnetic resonance quantitative physiological parameter maps, including obtaining groups of DCE-MRI images and the corresponding quantitative physiological parameter maps for each group of DCE-MRI images; using each group of DCE-MRI images and the quantitative physiological parameter maps as training inputs to train a deep learning model until the value of its loss function meets the preset requirements. The deep learning model is a generative adversarial network including a generation network and a discrimination network. The generation network is used to generate output images. The discrimination network includes a global discriminator and a local discriminator. Using the deep learning model when the value of the loss function meets the preset requirements as the quantitative physiological parameter map generation model, and processing the to-be-processed DCE-MRI images to obtain quantitative physiological parameter maps. Through the method provided by the embodiments of this specification, the efficiency and accuracy of generating quantitative physiological parameter maps can be improved.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of imaging technology, and in particular, to a method and device for generating magnetic resonance quantitative physiological parameter maps. Background Art

[0002] Quantitative physiological parameters are numerical parameters used to evaluate the physiological functions and metabolic states within tissues. They can assist doctors in diagnosing and monitoring certain diseases, such as tumors, strokes, etc. In medical imaging, quantitative physiological parameters are digital data obtained through imaging technologies, which can reflect the physiological conditions in aspects such as blood flow, blood oxygen, and metabolism within tissues. These parameters are usually calculated from imaging data (such as imaging data from computed tomography (CT), magnetic resonance imaging (MRI), ultrasound imaging (US), and positron emission tomography (PET), etc.). Therefore, different model assumptions and algorithms may lead to different results of the generated parameter maps. The Arterial Input Function (AIF) is one of the important parameters for calculating quantitative physiological parameters. The accuracy and reliability of the AIF often depend on the time resolution and sampling rate of the acquired data. However, due to the difficulty in estimating the AIF for each patient, a population-averaged AIF is usually adopted for PK analysis, and this method has certain limitations and errors. In addition, the generation of traditional quantitative physiological parameters also requires a large amount of computing resources and time.

[0003] In view of this, the embodiments of this specification aim to provide a method for generating magnetic resonance quantitative physiological parameter maps. Summary of the Invention

[0004] Aiming at the above problems of the prior art, the purpose of the embodiments of this specification is to provide a method for generating magnetic resonance quantitative physiological parameter maps to solve the problems of cumbersome operation and low accuracy in generating quantitative physiological parameter maps in the prior art.

[0005] To solve the above technical problems, the specific technical solutions of the embodiments of this specification are as follows:

[0006] In a first aspect, the embodiments of this specification provide a method for generating magnetic resonance quantitative physiological parameter maps, including:

[0007] Obtain groups of DCE-MRI images and the corresponding quantitative physiological parameter maps for each group of DCE-MRI images;

[0008] Using the DCE-MRI images of each group and the quantitative physiological parameter maps as training inputs, train a deep learning model until the loss function value of the deep learning model meets the preset requirements. The deep learning model is a generative adversarial network, which includes a generator network and a discriminator network. The generator network is used to generate output images. The discriminator network includes a global discriminator and a local discriminator. The global discriminator is used to discriminate the similarity between the output image and the quantitative physiological parameter map, and the local discriminator is used to discriminate the similarity between the local region of interest of the output image and the corresponding local region of interest in the quantitative physiological parameter map;

[0009] Using the deep learning model when the loss function value meets the preset requirements as a quantitative physiological parameter map generation model, process the to-be-processed DCE-MRI image to obtain a quantitative physiological parameter map.

[0010] Specifically, the loss function is:

[0011] LOSS = loss_G + loss_D;

[0012] Where LOSS is the loss value of the deep learning model, loss_G is the loss value of the generator network, and loss_D is the loss value of the discriminator network;

[0013] loss_D = loss_D1 * α + loss_D2 * (1 - α);

[0014] Where loss_D1 is the loss value of the global discriminator, loss_D2 is the loss value of the local discriminator, and α is a constant coefficient greater than 0 and less than 1.

[0015] Furthermore, the generator network is a generator network with group normalization added, and the discriminator network is a discriminator network with spectral normalization added.

[0016] Furthermore, obtaining the quantitative physiological parameter maps corresponding to the DCE-MRI images of each group includes:

[0017] Extracting the time-signal curve of any pixel point in the same group of DCE-MRI images;

[0018] Establishing the curve of the change of the contrast agent concentration with time for the corresponding pixel point in this group of DCE-MRI images;

[0019] Fitting the time-signal curve and the change curve to obtain the Ktrans value of the corresponding pixel point in this group of DCE-MRI images;

[0020] Repeating the above process to obtain the Ktrans values of all pixel points in the same group of DCE-MRI images;

[0021] Based on the Ktrans values of all pixels, a Ktrans quantitative physiological parameter map corresponding to this group of DCE-MRI images is obtained.

[0022] Furthermore, before obtaining the quantitative physiological parameter maps corresponding to each group of DCE-MRI images, the method further includes:

[0023] Performing alignment processing and / or registration processing on each group of DCE-MRI images.

[0024] Specifically, performing alignment processing on each group of DCE-MRI images includes:

[0025] Selecting any one of the DCE-MRI images in this group as a reference image, and obtaining the deviation information between the other images in this group of DCE-MRI images and the reference image;

[0026] According to the deviation information, aligning the other images with the reference image.

[0027] Specifically, performing registration processing on each group of DCE-MRI images includes:

[0028] Selecting any one of the DCE-MRI images in this group as a reference image, and extracting the same feature points and / or the same feature regions of the reference image and the other images in this group of DCE-MRI images;

[0029] According to the same feature points and / or the same feature regions, determining a similarity transformation model between the other images and the reference image;

[0030] According to the determined similarity transformation model, registering the other images with the reference image.

[0031] In a second aspect, an embodiment of the present specification further provides a magnetic resonance quantitative physiological parameter map generation device, including:

[0032] An acquisition module, configured to acquire each group of DCE-MRI images and the quantitative physiological parameter maps corresponding to each DCE-MRI image;

[0033] A training module, configured to use the groups of DCE-MRI images and the quantitative physiological parameter maps as training inputs to train a deep learning model until the loss function value of the deep learning model meets a preset requirement. The deep learning model is a generative adversarial network, which includes a generator network and a discriminator network. The generator network is used to generate output images. The discriminator network includes a global discriminator and a local discriminator. The global discriminator is used to discriminate the similarity between the output image and the quantitative physiological parameter map, and the local discriminator is used to discriminate the similarity between the local region of interest of the output image and the corresponding local region of interest in the quantitative physiological parameter map;

[0034] A processing module, configured to use the deep learning model when the loss function value meets the preset requirement as a quantitative physiological parameter map generation model to process the to-be-processed DCE-MRI image to obtain a quantitative physiological parameter map.

[0035] In a third aspect, an embodiment of this specification further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the method provided by the above technical solution is implemented.

[0036] In a fourth aspect, an embodiment of this specification further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method provided by the above technical solution is implemented.

[0037] By adopting the above technical solution, in the model training stage, the method for generating a magnetic resonance quantitative physiological parameter map provided by the embodiment of this specification can learn and identify complex features and patterns from a large amount of training data, without the need to build a model based on a preset hemodynamic model, nor the need to manually select multiple parameters to calculate parameters, thereby avoiding the dependence on assumptions such as vascular continuity and fluid incompressibility, reducing the degree of manual intervention, and being beneficial to improving the accuracy of calculation results.

[0038] In the model usage stage, the requirements for the to-be-processed DCE-MRI image can be reduced. For a single DCE-MRI image or a small number of DCE-MRI images, a relatively accurate Ktrans quantitative physiological parameter map can be generated; and the generation efficiency of the quantitative physiological parameter map can be greatly improved.

[0039] To make the above and other purposes, features, and advantages of the embodiments of this specification more obvious and understandable, the following specifically enumerates preferred embodiments and cooperates with the attached drawings to make a detailed description as follows. Description of the Drawings

[0040] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of this specification. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0041] Figure 1 The schematic diagram of the steps of a method for generating a magnetic resonance quantitative physiological parameter map provided by the embodiments of this specification is shown;

[0042] Figure 2 The comparison diagram of the quantitative physiological parameter map obtained by using the method for generating a magnetic resonance quantitative physiological parameter map provided by the embodiments of this specification and the quantitative physiological parameter map obtained by using the traditional method is shown;

[0043] Figure 3 The schematic diagram of the steps of obtaining the quantitative physiological parameter map corresponding to each group of DCE-MRI images;

[0044] Figure 4 The schematic diagram of the steps of performing alignment processing on each group of DCE-MRI images is shown;

[0045] Figure 5 The schematic diagram of the steps of performing registration processing on each group of DCE-MRI images is shown;

[0046] Figure 6 The schematic diagram of the structure of a device for generating a magnetic resonance quantitative physiological parameter map provided by the embodiments of this specification is shown;

[0047] Figure 7 The schematic diagram of the structure of a computer device provided by the embodiments of this specification is shown.

[0048] Explanation of the reference signs of the drawings:

[0049] 61. Acquisition module;

[0050] 62. Training module;

[0051] 63. Processing module;

[0052] 702. Computer device;

[0053] 704. Processor;

[0054] 706. Memory;

[0055] 708. Driving mechanism;

[0056] 710. Input / output module;

[0057] 712. Input device;

[0058] 714. Output device;

[0059] 716. Rendering device;

[0060] 718. Graphical user interface;

[0061] 720. Network interface;

[0062] 722. Communication link;

[0063] 724. Communication bus. Detailed implementation manners

[0064] Next, the technical solutions in the embodiments of this specification will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the embodiments of this specification.

[0065] It should be noted that the terms "first", "second", etc. in this specification, the claims and the above drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of this specification described here can be implemented in an order other than those illustrated or described here. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or equipment.

[0066] The embodiments of this specification provide a method for generating a magnetic resonance quantitative physiological parameter map, which can reduce manual intervention, reduce errors, and improve processing efficiency. Figure 1 is a schematic diagram of the steps of a method for generating a magnetic resonance quantitative physiological parameter map provided by the embodiments of this specification. This specification provides the method operation steps as described in the embodiments or the flowchart, but based on routine or non-creative labor, it may include more or fewer operation steps. The step sequence listed in the embodiments is only one way among the execution sequences of many steps, and does not represent the only execution sequence. When the actual system or device product executes, it can be executed in the order shown in the embodiments or the drawings or executed in parallel. Specifically, as Figure 1 shown, the method may include:

[0067] S110: Obtain each group of DCE-MRI images and the corresponding quantitative physiological parameter maps for each group of DCE-MRI images.

[0068] Exemplarily, a group of DCE-MRI images in the embodiments of this specification can be image acquisitions of the same tissue (e.g., breast) of the same object at certain time intervals for multiple times (e.g., 45 phases). Before the acquisition, a contrast agent (also known as a contrast medium) needs to be injected, and the aggregation of the contrast agent in the tissue is observed through image acquisition (the inflow and outflow of the contrast agent is a dynamic process, from initially none, to the amount of the contrast agent diffusing (flowing in) into the tissue reaching a peak, and then slowly dissipating (flowing out) from the tissue). By repeating the above operations, each group of DCE-MRI images of different tissues of different objects can be obtained.

[0069] S120: Use each group of DCE-MRI images and the quantitative physiological parameter maps as training inputs to train a deep learning model until the loss function value of the deep learning model meets the preset requirements. The deep learning model is a generative adversarial network, which includes a generator network and a discriminator network. The generator network is used to generate output images, and the discriminator network includes a global discriminator and a local discriminator. The global discriminator is used to discriminate the similarity between the output image and the quantitative physiological parameter map, and the local discriminator is used to discriminate the similarity between the local region of interest of the output image and the corresponding local region of interest in the quantitative physiological parameter map.

[0070] The generator network can be resnet, unet, and unet++.

[0071] In the embodiments of this specification, the generator network is preferably unet.

[0072] The discriminator network is PatchGAN; the batch size is set to 4; the total number of training time periods is set to 200 epochs: the learning rate for the first 100 time periods is set to 0.0002, and the learning rate for the remaining 100 time periods gradually decreases to 0.

[0073] In the embodiments of this specification, the number of local regions of interest (local ROIs) in a group of DCE-MRI images can be set to one or more according to actual needs, and the size of the local regions of interest can also be set according to actual needs. Generally, the tumor region or suspected tumor region in the DCE-MRI image can be selected as the local region of interest. Further, the deep learning model is a generative adversarial network (GAN), which adopts a dual discriminator network, namely a global discriminator and a local discriminator, which can improve the accuracy of the discrimination of the generation effect of the output image, and is conducive to improving the quality of the quantitative physiological parameter map obtained by the subsequent quantitative physiological parameter map generation model trained well.

[0074] S130: Use the deep learning model when the loss function value meets the preset requirements as the quantitative physiological parameter map generation model, and process the to-be-processed DCE-MRI image to obtain the quantitative physiological parameter map.

[0075] As Figure 2 shown, Figure 2 (a) in is the to-be-processed DCE-MRI image; Figure 2 (b) in is the quantitative physiological parameter map obtained by using the magnetic resonance quantitative physiological parameter map generation method provided by the embodiments of this specification; Figure 2 (c) in is the quantitative physiological parameter map obtained by using the traditional method. Comparing Figure 2 (b) and Figure 2 (c), it can be seen that the traditional quantitative physiological parameter map generation method is easily interfered by noise (motion artifacts). The quantitative physiological parameter map generation method provided by the embodiments of this specification adopts a deep learning model, and the deep learning model can learn from a large amount of data, and identify and eliminate useless information (such as noise) during the learning process. Finally, the quantitative physiological parameter map generation model obtained by iterative optimization can remove the interference of noise and improve the generation quality of the quantitative physiological parameter map.

[0076] Moreover, after the training of the quantitative physiological parameter map generation model is completed, to a certain extent, the requirements for the to-be-processed DCE-MRI image can be reduced. For example, the number of DCE-MRI images can be reduced (that is, the corresponding quantitative physiological parameter map is generated based on less than 45-phase images), which reduces the workload.

[0077] The method for generating a magnetic resonance quantitative physiological parameter map provided by the embodiments of this specification obtains a trained quantitative physiological parameter map generation model through a generation network and a dual discriminator network, and uses it to generate a quantitative physiological parameter map, which can reduce the errors caused by manual participation and improve the accuracy of the generated quantitative physiological parameter map; at the same time, there is no need to collect a large number of DCE-MRI images and perform modeling work based on hemodynamic models, etc., greatly reducing the workload and improving the efficiency of generating quantitative physiological parameter maps.

[0078] It should be noted that the method for generating a quantitative physiological parameter map provided by the embodiments of this specification is to generate a corresponding quantitative physiological parameter image for the to-be-processed DCE-MRI image. In some other feasible embodiments, it can also be to process imaging data such as computed tomography, ultrasound imaging, positron emission tomography, etc. to generate a quantitative physiological parameter map.

[0079] Furthermore, in the embodiments of this specification, the loss function is:

[0080] LOSS = loss_G + loss_D;

[0081] Where LOSS is the loss value of the deep learning model, loss_G is the loss value of the generation network, and loss_D is the loss value of the discriminator network. The generator loss and the discriminator loss are independent of each other and restrict each other.

[0082] Specifically,

[0083] loss_D = loss_D1 * α + loss_D2 * (1 - α);

[0084] Where loss_D1 is the loss value of the global discriminator, loss_D2 is the loss value of the local discriminator, and α is a constant coefficient greater than 0 and less than 1.

[0085] In some optional embodiments, the value of α is 0.5, that is, the weights of the global discriminator and the local discriminator are the same.

[0086] It should be noted that when training the deep learning model in the embodiments of this specification, a separate alternating iterative training method is adopted, that is, the generation network is trained first. During this process, the parameters of the discriminative network remain unchanged, so that the output image generated by the generation network is as similar as possible to the quantitative physiological parameter map in the training data, and the local region of interest in the output image is as similar as possible to the corresponding local region of interest in the quantitative physiological parameter map. After multiple iterations of training, the parameters of the generation network are fixed, and the discriminative network is trained to make the discriminative results of the similarity between the output image and the quantitative physiological parameter map, and the discriminative results of the similarity between the local region of interest in the output image and the corresponding local region of interest in the quantitative physiological parameter map more accurate. Finally, the generation network and the discriminative network are alternately trained until the training is completed.

[0087] Furthermore, in the embodiments of this specification, the generation network is a generation network with group normalization added, and the discriminative network is a discriminative network with spectral normalization added.

[0088] The generation network includes a convolutional layer, a normalization layer, and an activation layer. Among them, the normalization layer is a Group Normalization (GN) layer, and the input DCE-MRI image is subjected to group normalization processing in the generation network. During training, group normalization divides each channel into several groups, calculates the mean and standard deviation of the feature maps within each group for each input data, and uses them to standardize the feature maps. This can make the feature maps of each group have the same statistical characteristics, thereby reducing the phenomenon of covariate shift within the group, and then accelerating the training process and improving the performance of the model.

[0089] Similarly, the global discriminator and local discriminator of the discriminative network include a convolutional layer, a normalization layer, and an activation layer. Among them, the normalization layer is a Spectral Normalization (SN) layer. In the embodiments of this specification, the spectral normalization method is as follows:

[0090] For the weight matrix W of each layer of the discriminator, calculate its left singular vector and right singular vector, which are u and v respectively; through singular value decomposition, the spectral norm of the weight matrix is obtained as:

[0091] ||W||=max|u T Wv|;

[0092] where u T is the transpose of u.

[0093] Finally, the weight matrix W is normalized by dividing it by the maximum singular value.

[0094] In addition to the above singular value method, for each weight matrix W, the upper bound sigma of its spectral norm can also be calculated by the power iteration algorithm. The specific steps are as follows:

[0095] Multiply the transpose matrix W of the weight matrix W T repeatedly with a random vector x, and then normalize it to obtain a new vector y until the difference between y and the vector in the previous iteration is less than a threshold. Finally, the value of sigma is set to the inner product of y and x when the power iteration converges.

[0096] Finally, normalize the weight matrix W by dividing it by sigma.

[0097] In the method for generating a magnetic resonance quantitative physiological parameter map according to the embodiment of the present specification, by adding a group normalization generation network, the calculation efficiency of the model is improved, and the scalability and robustness of the model are enhanced; by adding a spectral normalization discriminant network, the singular values of the weight matrix are restricted, avoiding the disappearance or explosion of the weight gradient, improving the stability and generalization ability of the model, and ultimately facilitating the improvement of the quality of the quantitative physiological parameter map generated by the quantitative physiological parameter map generation model.

[0098] As Figure 3 shown, in step S110, obtaining the quantitative physiological parameter map corresponding to each group of DCE-MRI images further includes:

[0099] S310: Extract the time-signal curve of any pixel point in the same group of DCE-MRI images.

[0100] Optionally, before extracting the time-signal curve, preprocessing such as baseline correction and denoising can also be performed on the DCE-MRI images.

[0101] S320: Establish a curve of the change in the contrast agent concentration with time for the corresponding pixel points in this group of DCE-MRI images. Specifically, according to the single-compartment pharmacokinetic model, establish the relationship between the contrast agent concentration in the tissue and time:

[0102]

[0103] where C(t) is the contrast agent concentration of a certain pixel point at time t, Ktrans is the transfer constant in the tissue, AIF(t’) is the AIF curve value at time t’, C(t’) is the contrast agent concentration of a certain pixel point at time t’, and AIF is the arterial input function (Arterial Input Function, AIF). In the embodiment of the present specification, it is preferred to use a large blood vessel in the tissue as the AIF, or use a pre-determined average AIF curve (population-based AIF).

[0104] S330: Fit the time-signal curve and the change curve to obtain the Ktrans value of the corresponding pixel points in this set of DCE-MRI images.

[0105] Specifically, methods such as Nonlinear Least Squares (NLS) or Linear Least Squares (LLS) can be used to fit the time-signal curve and the change curve of each pixel point.

[0106] S340: Repeat the above process to obtain the Ktrans values of all pixel points in the same set of DCE-MRI images.

[0107] S350: Obtain the Ktrans quantitative physiological parameter map corresponding to this set of DCE-MRI images based on the Ktrans values of all pixel points.

[0108] Ktrans represents the rate at which the contrast agent flows from the blood vessels into the tissue interstitium, which is a comprehensive manifestation of perfusion and permeability. It can be understood as the transfer (rate) constant of the contrast agent from the blood vessels into the tissue interstitium, and can be used to show the vascular permeability of different regions in the tissue, providing technical support for subsequent data analysis, tumor diagnosis and treatment, etc. It should be noted that in the embodiments of this specification, the quantitative physiological parameter map used as the training data for the deep learning model is the Ktrans quantitative physiological parameter map. In some other feasible embodiments, the quantitative physiological parameter map used as the training data for the deep learning model can also be the Kep quantitative physiological parameter map (Kep is the rate at which the contrast agent flows out of the tissue interstitium from the tissue interstitium), the ve quantitative physiological parameter map (ve is the proportion of the tissue interstitium), etc. Then, after the model training is completed, the processed quantitative physiological parameter map is adapted to the training data.

[0109] Preferably, in the embodiments of this specification, before obtaining the quantitative physiological parameter map corresponding to each group of DCE-MRI images in step S110, the method further includes:

[0110] Perform alignment processing and / or registration processing on each group of DCE-MRI images.

[0111] As Figure 4 shown, performing alignment processing on each group of DCE-MRI images may include the following steps:

[0112] S410: Select any one of the DCE-MRI images in this group as the reference image to obtain the deviation information between the other images in this group of DCE-MRI images and the reference image.

[0113] S420: Align the other images with the reference image according to the deviation information.

[0114] For example, based on the deviation information, other images can be aligned with the reference image through transformation operations such as translation, rotation, or scaling. In DCE-MRI image acquisition, the acquisition object may move, or the position of the image acquisition site may change due to factors such as breathing, resulting in inaccurate input data or problems where the data cannot be analyzed. In the embodiments of this specification, by performing alignment processing on each group of DCE-MRI images, the accuracy and reliability of the data can be improved.

[0115] As Figure 5 shown, performing registration processing on each group of DCE-MRI images includes:

[0116] S510: Select any one of the DCE-MRI images in this group as the reference image, and extract the same feature points and / or the same feature regions of the reference image and other images in this group of DCE-MRI images.

[0117] Feature points and feature regions can be significant and unique objects in DCE-MRI images. Exemplarily, they can be closed boundary regions, edge regions, line intersection points, etc. After selecting appropriate feature points and / or feature regions, match the feature points and / or feature regions in other images with the corresponding feature points and / or feature regions in the reference image to establish a correlation between other images and the reference image.

[0118] S520: Determine the similarity transformation model between other images and the reference image according to the same feature points and / or the same feature regions.

[0119] That is, determine the range and method of the transformation.

[0120] S530: Register other images with the reference image according to the determined similarity transformation model.

[0121] It should be noted that in the embodiments of this specification, registration processing is performed on two consecutive (or multiple) images in the same group of DCE-MRI images. In some other embodiments, it can also be the registration of two (or multiple) images in DCE-MRI images that belong to the same tissue of the same object but different groups. It can also be the registration between different modality images of the same tissue of the same object. For example, CT images can be registered with DCE-MRI images.

[0122] Registration processing can compare images acquired from different acquisition groups or different devices. Through registration operations, these images can be subjected to unified coordinate system conversion, thereby enabling more accurate comparison and analysis.

[0123] In actual operation, the alignment process and the registration process can be used in combination. In the embodiments of this specification, by performing alignment and registration processing operations on the acquired DCE-MRI images, the positions and orientations between the DCE-MRI images and between the local regions of interest of the DCE-MRI images are correctly aligned. Thus, it is beneficial to improve the accuracy of calculating the Ktrans value and provide a reliable basis for subsequent data analysis and clinical applications.

[0124] In some feasible embodiments, after performing alignment processing and / or registration processing on each group of DCE-MRI images, it may further include:

[0125] Performing an interpolation operation on the aligned DCE-MRI images and / or the registered DCE-MRI images to generate new images with the same pixel spacing and orientation as them, that is, increasing the number of training input data.

[0126] In summary, the method for generating a magnetic resonance quantitative physiological parameter map provided by the embodiments of this specification can learn and identify complex features and patterns from a large amount of training data during the model training stage, without the need to build a model based on a preset hemodynamic model, thereby avoiding the dependence on assumptions such as vascular continuity and fluid incompressibility. And since the deep learning model can automatically learn parameters from the data, there is no need to manually select multiple parameters to calculate perfusion parameters, which reduces the degree of manual intervention and avoids the problem of inaccurate calculation results caused by improper parameter selection.

[0127] When using the trained deep learning model, that is, the quantitative physiological parameter map generation model, to generate a quantitative physiological parameter map, a large amount of mathematical calculations and analyses can be avoided, reducing the computational complexity while greatly improving the data processing efficiency. And the trained model can reduce the requirements for the DCE-MRI images to be processed. For a single DCE-MRI image or a small number of DCE-MRI images, the quantitative physiological parameter map generation model can generate accurate results.

[0128] As Figure 6 shown, the embodiments of this specification also provide a magnetic resonance quantitative physiological parameter map generation device, including:

[0129] An acquisition module 61, configured to acquire each group of DCE-MRI images and the corresponding quantitative physiological parameter maps of the DCE-MRI images;

[0130] A training module 62 is configured to use the DCE-MRI image and the quantitative physiological parameter map as training inputs to train a deep learning model until the loss function value of the deep learning model meets a preset requirement. The deep learning model is a generative adversarial network, which includes a generator network and a discriminator network. The generator network is configured to generate an output image. The discriminator network includes a global discriminator and a local discriminator. The global discriminator is configured to discriminate the similarity between the output image and the quantitative physiological parameter map, and the local discriminator is configured to discriminate the similarity between the local region of interest of the output image and the corresponding local region of interest in the quantitative physiological parameter map.

[0131] A processing module 63 is configured to use the deep learning model when the loss function value meets the preset requirement as a quantitative physiological parameter map generation model to process a to-be-processed DCE-MRI image to obtain a quantitative physiological parameter map.

[0132] The beneficial effects achieved by the device provided in the embodiments of this specification are consistent with those achieved by the above method, and will not be elaborated here.

[0133] As Figure 7 shown, a computer device provided in the embodiments of this specification is presented. The magnetic resonance quantitative physiological parameter map generation device in this specification can be the computer device in this embodiment, which executes the magnetic resonance quantitative physiological parameter map generation method provided in the embodiments of this specification. The computer device 702 may include one or more processors 704, such as one or more central processing units (CPUs), and each processing unit may implement one or more hardware threads. The computer device 702 may also include any memory 706, which is used to store any type of information such as code, settings, data, etc. Non-limiting examples include any type of RAM, any type of ROM, flash memory devices, hard disks, optical discs, etc. More generally, any memory may use any technology to store information. Further, any memory may provide volatile or non-volatile retention of information. Further, any memory may represent a fixed or removable component of the computer device 702. In one case, when the processor 704 executes the associated instructions stored in any memory or combination of memories, the computer device 702 may perform any operation of the associated instructions. The computer device 702 also includes one or more drive mechanisms 708 for interacting with any memory, such as a hard disk drive mechanism, an optical disc drive mechanism, etc.

[0134] The computer device 702 may also include an input / output module 710 (I / O) for receiving various inputs (via the input device 712) and for providing various outputs (via the output device 714). A specific output mechanism may include a presentation device 716 and an associated graphical user interface (GUI) 718. In other embodiments, the input / output module 710 (I / O), the input device 712, and the output device 714 may not be included, and it may only be a computer device in the network. The computer device 702 may also include one or more network interfaces 720 for exchanging data with other devices via one or more communication links 722. One or more communication buses 724 couple the components described above together.

[0135] The communication link 722 may be implemented in any way, for example, through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication link 722 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc. governed by any protocol or combination of protocols.

[0136] Corresponding to Figure 1 、 Figures 3 to 5 In the method of, an embodiment of the present specification also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is run by a processor, it executes the steps of the above method.

[0137] An embodiment of the present specification also provides a computer-readable instruction, wherein when the processor executes the instruction, the program therein causes the processor to execute as Figure 1 、 Figures 3 to 5 shown in the method.

[0138] An embodiment of the present specification also provides a computer program product, including at least one instruction or at least one segment of a program, and the at least one instruction or the at least one segment of the program is loaded and executed by a processor to implement as Figure 1 、 Figures 3 to 5 shown in the method.

[0139] It should be understood that in various embodiments of the present specification, the magnitudes of the serial numbers of the above processes do not mean the order of execution is prior or subsequent. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present specification.

[0140] It should also be understood that in the embodiments of this specification, the term "and / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this specification generally represents an "or" relationship between the associated objects before and after.

[0141] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this specification can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments of this specification.

[0142] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be repeated here.

[0143] In several embodiments provided in the embodiments of this specification, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices, or units, and can also be electrical, mechanical, or other forms of connection.

[0144] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments of this specification.

[0145] In addition, the functional units in the various embodiments of this specification can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0146] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this specification, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this specification. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0147] Specific embodiments are used in this specification to elaborate on the principles and implementation manners of this specification. The description of the above embodiments is only used to help understand the method and its core idea of this specification; at the same time, for those of ordinary skill in the art, according to the idea of this specification, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the embodiments of this specification.

Claims

1. A method for generating a magnetic resonance quantitative physiological parameter map, characterized in that Including: Obtaining DCE-MRI images of each group and quantitative physiological parameter maps corresponding to the DCE-MRI images of each group; Using the DCE-MRI images of each group and the quantitative physiological parameter maps as training inputs to train a deep learning model until the loss function value of the deep learning model meets a preset requirement. The deep learning model is a generative adversarial network, which includes a generator network and a discriminator network. The generator network is used to generate output images. The discriminator network includes a global discriminator and a local discriminator. The global discriminator is used to discriminate the similarity between the output image and the quantitative physiological parameter map, and the local discriminator is used to discriminate the similarity between the local region of interest of the output image and the corresponding local region of interest in the quantitative physiological parameter map; Using the deep learning model when the loss function value meets the preset requirement as a quantitative physiological parameter map generation model to process the to-be-processed DCE-MRI image to obtain a quantitative physiological parameter map; The loss function is: LOSS = loss_G + loss_D; where LOSS is the loss value of the deep learning model, loss_G is the loss value of the generator network, and loss_D is the loss value of the discriminator network; loss_D = loss_D1 * α + loss_D2 * (1 - α); where loss_D1 is the loss value of the global discriminator, loss_D2 is the loss value of the local discriminator, and α is a constant coefficient greater than 0 and less than 1; The generator network is a generator network with group normalization added, and the discriminator network is a discriminator network with spectral normalization added; Obtaining the quantitative physiological parameter map corresponding to the DCE-MRI images of each group includes: Extracting the time-signal curve of any pixel point in the DCE-MRI images of the same group; Establishing a curve of the change of the contrast agent concentration with time for the corresponding pixel points in the DCE-MRI images of this group; Fitting the time-signal curve and the change curve to obtain the Ktrans value of the corresponding pixel points in the DCE-MRI images of this group; Repeating the above process to obtain the Ktrans values of all pixel points in the DCE-MRI images of the same group; Based on the Ktrans values of all pixel points, obtaining the Ktrans quantitative physiological parameter map corresponding to the DCE-MRI images of this group.

2. The method for generating a magnetic resonance quantitative physiological parameter map according to claim 1, wherein Before obtaining the quantitative physiological parameter map corresponding to the DCE-MRI images of each group, the method further includes: Performing alignment processing and / or registration processing on the DCE-MRI images of each group.

3. The method for generating a magnetic resonance quantitative physiological parameter map according to claim 2, wherein Performing alignment processing on the DCE-MRI images of each group includes: Selecting any one of the DCE-MRI images of this group as a reference image to obtain the deviation information between the other images in the DCE-MRI images of this group and the reference image; Aligning the other images with the reference image according to the deviation information.

4. The method for generating a magnetic resonance quantitative physiological parameter map according to claim 2, wherein Performing registration processing on the DCE-MRI images of each group includes: Select any one of the group of DCE-MRI images as the reference image, and extract the same feature points and / or the same feature regions of the reference image and other images in the group of DCE-MRI images; Determine the similarity transformation model between other images and the reference image according to the same feature points and / or the same feature regions; Register other images and the reference image according to the determined similarity transformation model.

5. A magnetic resonance quantitative physiological parameter map generation device, characterized in that, Including: An acquisition module, configured to acquire each group of DCE-MRI images and the corresponding quantitative physiological parameter maps of each DCE-MRI image; wherein, extract the time-signal curve of any pixel point in the same group of DCE-MRI images; establish the change curve of the contrast agent concentration with time of the corresponding pixel point in the group of DCE-MRI images; fit the time-signal curve and the change curve to obtain the Ktrans value of the corresponding pixel point in the group of DCE-MRI images; repeat the above process to obtain the Ktrans values of all pixel points in the same group of DCE-MRI images; obtain the Ktrans quantitative physiological parameter map corresponding to the group of DCE-MRI images based on the Ktrans values of all pixel points; A training module, configured to use the groups of DCE-MRI images and the quantitative physiological parameter maps as training inputs to train a deep learning model until the loss function value of the deep learning model meets the preset requirements. The deep learning model is a generative adversarial network, and the generative adversarial network includes a generator network and a discriminator network. The generator network is used to generate output images, and the discriminator network includes a global discriminator and a local discriminator. The global discriminator is used to discriminate the similarity between the output image and the quantitative physiological parameter map, and the local discriminator is used to discriminate the similarity between the local region of interest of the output image and the corresponding local region of interest in the quantitative physiological parameter map; wherein, the loss function is: LOSS = loss_G + loss_D; wherein, LOSS is the loss value of the deep learning model, loss_G is the loss value of the generator network, and loss_D is the loss value of the discriminator network; loss_D = loss_D1 * α + loss_D2 * (1 - α); wherein, loss_D1 is the loss value of the global discriminator, loss_D2 is the loss value of the local discriminator, and α is a constant coefficient greater than 0 and less than 1; The generator network is a generator network with group normalization added, and the discriminator network is a discriminator network with spectral normalization added; A processing module, configured to use the deep learning model when the loss function value meets the preset requirements as a quantitative physiological parameter map generation model to process the to-be-processed DCE-MRI image to obtain a quantitative physiological parameter map.

6. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the method according to any one of claims 1 to 4 when executing the computer program.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and the computer program implements the method according to any one of claims 1 to 4 when executed by the processor.

Images