Ultrasonic viscoelasticity measurement method, device, equipment and medium

Abstract

The present application discloses an ultrasonic viscoelasticity measurement method, device, equipment and medium, and relates to the field of shear wave elastic imaging. The method includes establishing a non-uniform viscoelastic wave equation; encoding the non-uniform viscoelastic wave equation into a loss function, and constructing a physical information neural network; the physical information neural network includes a spatiotemporal neural network and a spatial neural network; the spatiotemporal neural network is used to predict the stream function; the non-uniform viscoelastic wave equation obtains the vertical vibration velocity of the particle according to the output stream function; the spatial neural network is used to predict the shear modulus and viscosity; the loss function calculates the total loss according to the vertical vibration velocity of the particle, the shear modulus and the viscosity; the vertical vibration velocity of the multi-frame particle of the object to be measured is used as data supervision, and input into the physical information neural network to obtain the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured. The present application can accurately estimate the spatial distribution of the shear modulus and the spatial distribution of the viscosity, and improve the accuracy of elastic prediction.

Description

Technical Field

[0001] The present application relates to the field of shear wave elastic imaging, and in particular to an ultrasonic viscoelasticity measurement method, device, equipment and medium. Background Art

[0002] Currently, Shear Wave Elastography (SWE) has been widely used in clinical practice, mainly for the staging of liver fibrosis, the detection of arterial stiffness, and the differentiation of benign and malignant tumors. However, the traditional SWE method usually uses the shear wave time of flight (ToF) method, relying on the shear wave velocity to estimate the viscoelastic properties of the tissue. This method performs well when dealing with larger tissue structures, but often produces large errors when encountering inclusions less than about 1 cm in size. This is because the estimation of shear wave velocity is essentially a macroscopic measurement, and its resolution is limited by the wavelength and sampling frequency. When there are fine structures or complex geometries in the tissue, the mechanical inversion method that relies solely on shear wave velocity often cannot accurately capture these microscopic features, thereby affecting the accurate estimation of the viscoelastic distribution in complex human soft tissues.

[0003] In recent years, deep learning (DL), as a powerful data-driven tool, has made significant progress in a variety of fields. Deep learning relies on improved artificial neural networks, which are composed of multiple processing layers and can automatically extract and learn representative features of input data. In the field of medical image analysis, DL has been shown to be able to recognize and interpret complex image patterns. The deep learning-based SWE method can extract rich information from the spatial and temporal characteristics of shear waves through the training process, and has the potential to overcome the limitations of traditional methods and provide more accurate estimates of tissue viscoelastic distribution.

[0004] Nevertheless, current deep learning-based SWE methods still face challenges, especially when viscoelastic materials are involved, whose dynamic responses are not only affected by the elastic modulus but also by their viscosity. Therefore, it is particularly important to develop techniques that can accurately estimate the spatial distribution of shear modulus and viscosity. Summary of the invention

[0005] The purpose of this application is to provide an ultrasonic viscoelasticity measurement method, device, equipment and medium, which can accurately estimate the spatial distribution of shear modulus and viscosity, and improve the accuracy of elasticity prediction.

[0006] To achieve the above objectives, this application provides the following solutions.

[0007] In the first aspect, the present application provides an ultrasonic viscoelasticity measurement method, including: establishing a non-uniform viscoelastic wave equation; the non-uniform viscoelastic wave equation characterizes the relationship between the vertical vibration velocity of a particle and the stream function; encoding the non-uniform viscoelastic wave equation into a loss function, and constructing a physical information neural network; the physical information neural network includes a spatiotemporal neural network, a spatial neural network and a loss function; the spatiotemporal neural network is used to predict the stream function; the non-uniform viscoelastic wave equation is used to obtain the vertical vibration velocity of a particle according to the output stream function; the spatial neural network is used to predict the shear modulus and viscosity; the loss function is used to calculate the total loss according to the vertical vibration velocity of the particle, the shear modulus and the viscosity; determining the vertical vibration velocity of multiple frames of particles of the object to be measured according to the continuous multi-frame ultrasonic radio frequency signals of the object to be measured; using the vertical vibration velocity of multiple frames of particles of the object to be measured as data supervision, inputting it into the physical information neural network, and obtaining the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured.

[0008] In the second aspect, the present application provides an ultrasonic viscoelasticity measurement device, including: an equation establishment module, a neural network construction module, a vibration velocity determination module and an application module. The equation establishment module is used to establish a non-uniform viscoelastic wave equation; the non-uniform viscoelastic wave equation characterizes the relationship between the vertical vibration velocity of the particle and the stream function; the neural network construction module is used to encode the non-uniform viscoelastic wave equation into a loss function and construct a physical information neural network; the physical information neural network includes a spatiotemporal neural network, a spatial neural network and a loss function; the spatiotemporal neural network is used to predict the stream function; the non-uniform viscoelastic wave equation is used to obtain the vertical vibration velocity of the particle according to the output stream function; the spatial neural network is used to predict the shear modulus and viscosity; the loss function is used to calculate the total loss according to the vertical vibration velocity of the particle, the shear modulus and the viscosity; the vibration velocity determination module is used to determine the vertical vibration velocity of the multi-frame particle of the object to be measured according to the continuous multi-frame ultrasonic radio frequency signal of the object to be measured; the application module is used to use the vertical vibration velocity of the multi-frame particle of the object to be measured as data supervision, input into the physical information neural network, and obtain the shear modulus spatial distribution and viscosity spatial distribution of the object to be measured.

[0009] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the ultrasonic viscoelasticity measurement methods described above.

[0010] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the ultrasonic viscoelasticity measurement methods described above.

[0011] According to the specific embodiments provided in this application, this application has the following technical effects.

[0012] The present application provides an ultrasonic viscoelasticity measurement method, device, equipment and medium, which applies physical information neural network to the tissue viscoelastic inversion of shear wave elastic imaging. Compared with the traditional shear wave flight time method based on shear wave velocity, the physical information neural network can extract more shear wave spatiotemporal characteristics, and has an inversion accuracy far exceeding that of the traditional shear wave flight time method based on shear wave velocity in complex environments, avoiding the estimation error caused by wave reflection and scattering to the greatest extent; and the physical information neural network can simultaneously realize the spatial distribution of shear modulus and viscosity, thereby improving the elasticity prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0014] Figure 1 Schematic diagram of a flow chart of an ultrasonic viscoelasticity measurement method in one embodiment of the present application.

[0015] Figure 2 A schematic diagram of the structure of a physical information neural network provided in another embodiment of the present application.

[0016] Figure 3 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0019] Since the occurrence and development of diseases are usually accompanied by changes in their mechanical properties, shear wave elastography has broad prospects for clinical application because it can measure the mechanical parameters (elasticity, viscosity, etc.) of soft tissues in vivo.

[0020] Traditional SWE methods usually use the ToF method and rely on the estimation of shear wave velocity. Shear wave velocity is often regarded as an important surrogate for tissue elasticity and can be estimated by analyzing a series of images of shear wave spatiotemporal data. This process is usually implemented in two main steps: first, a cross-correlation analysis is performed on two time-varying signals with known distances, or a linear regression is performed on the peaks in the two-dimensional spatiotemporal image to calculate the propagation time of the shear wave. The ToF method is based on the assumption that shear waves propagate in a fixed direction, so the signal needs to be directional filtered before the cross-correlation analysis. Second, the calculation of viscosity usually relies on the change in the slope of the shear wave velocity at different excitation frequencies, which can be used to characterize the viscoelastic properties of the tissue.

[0021] The ToF method has become the most commonly used viscoelasticity estimation method in clinical practice due to its relatively simple operation and high measurement accuracy. However, this method has significant errors when dealing with non-uniform materials, and this problem is particularly evident in tissues with strong heterogeneity. This error arises from the ToF method's idealized assumptions about the shear wave propagation path and medium uniformity, which is often not true in complex biological tissues. Therefore, the error of the ToF method under specific circumstances has become one of the major challenges in the current clinical application of shear wave elastography technology, limiting its application in accurate viscoelasticity measurement.

[0022] Convolutional Neural Networks (CNN) have been widely used in various image processing and pattern recognition tasks, especially in extracting temporal and spatial features in the wave propagation process. Through automated learning and data-driven features, CNN can effectively capture the complex patterns in shear wave propagation. However, CNN training usually relies on a large number of high-quality data sets, especially in the field of medical imaging, where it is extremely difficult to obtain shear wave propagation data in real human bodies. At present, most studies rely on numerical simulations, phantom experiments, and ex vivo tissue experimental data to generate training data sets for model training.

[0023] However, this training method based on simulation and phantom data is somewhat different from the actual situation of real human tissue, which limits the generalization and interpretability of the model when applied to the actual human body. In addition, the "black box" nature of CNN in deep learning models makes it difficult to interpret and analyze, further limiting its potential in clinical translation. Therefore, although CNN has certain advantages in wave propagation analysis, its limitations have also significantly hindered its widespread application in shear wave elastography.

[0024] It can be seen that the current SWE technology has the following problems and improvement directions.

[0025] 1) Traditional SWE methods rely on the estimation of shear wave velocity, but the propagation of waves in non-uniform soft tissues is prone to reflection and scattering, making the estimation of shear wave velocity more challenging. Therefore, it is necessary to consider more parameters in the inversion of tissue mechanical properties, rather than relying solely on shear wave velocity.

[0026] 2) The ToF method based on cross-correlation is very sensitive to noise and it is difficult to reconstruct a smooth tissue elasticity distribution map. Therefore, it is necessary to develop a tissue mechanical properties inversion algorithm with strong noise resistance.

[0027] 3) Biological tissue is not a simple linear elastic material, but a more complex viscoelastic material. The estimation of tissue elasticity is largely affected by the viscosity distribution. Therefore, an algorithm that can simultaneously realize the spatial distribution of elasticity and viscosity is needed to improve the prediction accuracy of elasticity.

[0028] In view of this, in an exemplary embodiment, Figure 1 As shown, the present application provides an ultrasonic viscoelasticity measurement method, including the following steps 101 to 104.

[0029] Step 101: Establish a non-uniform viscoelastic wave equation; the non-uniform viscoelastic wave equation characterizes the relationship between the vertical vibration velocity of the particle and the stream function.

[0030] Step 102: Encode the non-uniform viscoelastic wave equation into a loss function and construct a physical information neural network; the physical information neural network includes a space-time neural network and a space neural network; the space-time neural network is used to predict the stream function; the non-uniform viscoelastic wave equation is used to obtain the vertical vibration velocity of the particle according to the output stream function; the space neural network is used to predict the shear modulus and viscosity; the loss function is used to calculate the total loss according to the vertical vibration velocity of the particle, the shear modulus and the viscosity.

[0031] Step 103: determining the vertical vibration velocity of the multi-frame particle of the object to be measured according to the continuous multi-frame ultrasonic radio frequency signal of the object to be measured.

[0032] Step 104: using the vertical vibration velocity of the multi-frame particles of the object to be measured as data supervision, inputting it into the physical information neural network, and obtaining the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured.

[0033] By implementing the above steps 101 to 104, the Physics-Informed Neural Networks (PINN) technology is innovatively applied to the tissue viscoelastic inversion of SWE, overcoming the clinical application limitations of many traditional SWE methods. Compared with the traditional ToF method based on shear wave velocity, PINN can extract more spatiotemporal features of shear waves, and has an inversion accuracy far exceeding ToF in complex environments, avoiding the estimation error caused by wave reflection and scattering to the greatest extent. Compared with traditional convolutional neural networks, encoding physical equations as loss functions enables the network to learn data features in a disorderly manner instead of using physical laws as a guide to achieve parameter solution. This method of embedding physical information enables the neural network to no longer rely solely on the stacking of a large amount of training data, greatly reducing the demand for training data, and even achieving accurate solution of single-case data. At the same time, this technology can view the solution process of each internal parameter, greatly improving the interpretability of deep learning. Therefore, this method has application advantages far beyond existing technologies.

[0034] In another exemplary embodiment of the present application, the viscoelastic model is generally divided into the Kelvin-Voigt model (elastic element and viscous element in parallel), the Maxwell model (elastic element and viscous element in series), and other combinations of the two. Considering that the problem of studying viscoelastic solids is, the Kelvin-Voigt model is selected as the constitutive equation, and it is assumed that the material satisfies the conditions of incompressibility and isotropy. Then the process of establishing the non-uniform viscoelastic wave equation in step 101 above can be replaced by the following steps 201 to 203.

[0035] Step 201: According to Newton's second law and the shear stress-shear strain relationship, the expression form of the wave equation of the inhomogeneous medium with respect to the particle displacement is determined as follows: ; In the formula, The default density is uniform and 1000 kg / m 3 . for Direction displacement, For time, is the shear modulus, is the Einstein summation rule parameter, is the spatial direction, is the viscosity, is the shift of the Einstein sum rule, is the Lagrange multiplier associated with the incompressible constraint.

[0036] Newton's second law is: In the formula, For stress.

[0037] The shear stress-shear strain relationship is: In the formula, is the Dirac delta function, in order to introduce the incompressible constraint condition into the equation.

[0038] Step 202: The expression of the non-uniform medium wave equation about the particle displacement is organized into an expression of the particle vibration velocity in two-dimensional space: ; In the formula, is the horizontal vibration velocity of the particle, is the vertical vibration velocity of the particle, is the real space scale of the horizontal axis, is the real space scale of the ordinate, Lagrangian operator that satisfies the incompressibility constraint Taking the partial derivative in the horizontal direction, Lagrangian operator that satisfies the incompressibility constraint Take the partial derivative with respect to the vertical direction.

[0039] It is organized into an expression of the particle vibration velocity in two-dimensional space to facilitate subsequent encoding into the neural network.

[0040] Step 203: Introduce the stream function, simplify the expression of the particle vibration velocity in two-dimensional space, and obtain the non-uniform viscoelastic wave equation as follows: ; In the formula, is the stream function.

[0041] In order to ensure that the incompressible constraint is always satisfied during the solution of the equation, the stream function is introduced , further simplified.

[0042] In another exemplary embodiment of the present application, the elastic inversion technology based on deep learning currently needs to rely on massive data drive, however, the clinical SWE data set is difficult to obtain, and it has defects such as poor interpretability and poor generalization. Therefore, a physical information model that can integrate physical laws into deep learning is needed, so that it can achieve prediction results that conform to basic physical laws under the drive of a small amount of data, avoiding the "black box effect". Physical information neural network is an innovative computing method that uses physical laws as part of neural network training by encoding the control equations of physical problems, such as partial differential equations, in a fully connected neural network. This method not only retains the powerful learning ability of the deep learning model, but also combines the laws of physics, making the model more robust and physically interpretable when facing complex problems. PINN has gradually become an effective tool for solving direct and inverse problems, especially in cases involving complex multi-physical field coupling, showing strong application prospects.

[0043] Physical information neural networks combine the powerful modeling capabilities of deep learning with the constraints of physical equations, and are gradually becoming a powerful tool for solving complex biological tissue mechanics inversion problems. By encoding the inhomogeneous viscoelastic wave equation into the physical information neural network and inputting a set of shear wave particle vibration velocity data, the spatial distribution of viscosity and shear modulus in the tissue can be solved more accurately. This method not only improves the resolution of small-sized inclusions, but also provides new possibilities for the accurate diagnosis of complex soft tissues.

[0044] At present, there is no mature research or application of shear wave elastography viscoelasticity estimation based on physical information neural networks. However, since physical information neural networks can more accurately capture the physical characteristics of the wave process through the constraints of physical equations, this method is expected to overcome many limitations of existing technologies. Specifically, when dealing with highly heterogeneous tissues, physical information neural networks can more accurately describe the complex situations in shear wave propagation, thereby providing new solutions for the accurate estimation of viscoelastic spatial distribution. With the in-depth research of physical information neural networks in this field, it has the potential to significantly enhance the clinical application value of shear wave elastography technology and promote its widespread application in precision medicine.

[0045] This application innovatively applies physical information neural network to tissue viscoelastic inversion of shear wave elastography, such as Figure 2 As shown, the physical information neural network mainly includes a spatiotemporal neural network and a spatial neural network, which respectively solve the four unknown quantities in step 202 and step 203: the stream function , Lagrangian operator satisfying the incompressibility constraint , shear modulus and viscosity .

[0046] Figure 2 The specific structure of the physical information neural network shown is as follows.

[0047] The spatiotemporal neural network contains three input parameters ( , , ), 4 hidden layers, 80 neurons per layer, and two output parameters (stream function and the Lagrangian operator that satisfies the incompressibility constraint ), which is used to output the partial derivatives of the parameters with respect to the three-dimensional space-time coordinate system during the back-propagation process.

[0048] The spatial neural network contains two input parameters ( , ), 4 hidden layers, 80 neurons per layer, and two output parameters (viscosity and shear modulus Spatial distribution), which is used to output the partial derivatives of the parameters with respect to the two-dimensional spatial coordinate system during the back propagation process.

[0049] The loss functions of the physical information neural network include: data error loss function and partial differential equation loss function. The data error loss function is: ; In the formula, is the data error loss, is the mean square error, is the vertical vibration velocity of the particle, is the vertical vibration velocity of the particle used as data supervision. That is, the mean square error loss (Data Loss) is calculated by comparing the actual vertical vibration velocity of the particle collected with the vertical vibration velocity of the particle output by the network.

[0050] The partial differential equation loss (PDE Loss) is composed of the multi-order partial derivatives of the output of each network. Together, they form the code for the inhomogeneous viscoelastic wave equation. The partial differential equation loss function is as follows.

[0051] .

[0052] .

[0053] In the formula, is the first partial differential equation loss, Lagrangian operator that satisfies the incompressibility constraint Taking the partial derivative in the horizontal direction, is the density, Find the second-order derivative of the horizontal vibration velocity of the particle with respect to time, is the shear modulus, Find the second derivative of the horizontal vibration velocity of the particle in the horizontal direction, Find the second derivative of the horizontal vibration velocity of the particle with respect to the vertical direction, is the viscosity, Find the second derivative of the horizontal vibration velocity of the particle with respect to the horizontal direction and then with respect to time Find the partial derivative, Find the second derivative of the horizontal vibration velocity of the particle with respect to the vertical direction and then with respect to time Find the partial derivative, is the partial derivative of the shear modulus with respect to the horizontal direction, Find the partial derivative of the horizontal vibration velocity of the particle with respect to the horizontal direction, is the partial derivative of the shear modulus with respect to the vertical direction, To find the partial derivative of the horizontal vibration velocity of the particle with respect to the vertical direction, To find the partial derivative of the vertical vibration velocity of the particle with respect to the horizontal direction, is the partial derivative of viscosity with respect to the horizontal direction, Find the partial derivative of the horizontal vibration velocity of the particle with respect to the horizontal direction and then with respect to time Find the partial derivative, is the partial derivative of viscosity with respect to the vertical direction, Find the partial derivative of the horizontal vibration velocity of the particle with respect to the vertical direction and then with respect to time Find the partial derivative, Find the partial derivative of the vertical vibration velocity of the particle with respect to the horizontal direction and then with respect to time Find partial derivatives; is the second partial differential equation loss, is the vertical vibration velocity of the particle relative to time Find the second-order derivative, Find the second derivative of the vertical vibration velocity of the particle with respect to the horizontal direction, Find the second derivative of the vertical vibration velocity of the particle in the vertical direction, The second derivative of the vertical vibration velocity of the particle with respect to the horizontal direction is calculated and then the derivative of the vertical vibration velocity with respect to the time Find the partial derivative, The second derivative of the vertical vibration velocity of the particle is calculated with respect to the vertical direction and then with respect to the time Find the partial derivative, is the partial derivative of the vertical vibration velocity of the particle with respect to the vertical direction, Find the partial derivative of the vertical vibration velocity of the particle with respect to the vertical direction and then with respect to time Find the partial derivative.

[0054] The loss function is the sum of the data error loss function and the partial differential equation loss function multiplied by their respective hyperparameters, ensuring that the rates of each loss function remain consistent during the descent process, which is conducive to the overall convergence of the network. The choice of hyperparameters can be appropriately adjusted according to different usage environments. The loss function is: ; In the formula, is the total loss, is the hyperparameter of the partial differential equation loss function, is the hyperparameter of the data error loss function.

[0055] In another exemplary embodiment of the present application, the activation function of the physical information neural network is selected as the Tanh function, but other activation functions such as ReLu can also achieve similar functions. The physical information neural network uses a traditional multi-layer perceptron, that is, only the weights of the neurons are changed during the training process, and the activation function does not change during the training process; and there is currently a strategy of variable activation function, that is, using a B-spline function, in which the parameters can be updated through the training process, so that the activation function corresponding to each neuron will continue to change with the training process, and in theory, it can also solve partial differential equations, but this method has high requirements for video memory.

[0056] In another exemplary embodiment of the present application, in order to better apply the method to clinical practice, the physical information neural network is pre-trained on existing data so that the physical information neural network is forced to satisfy the solution process of the wave equation, so that the new data can find the gradient descent direction as quickly as possible after being input into the network. After testing, it was found that the pre-trained physical information neural network can reduce the training time by more than 80% without reducing the imaging quality, greatly speeding up its inversion solution speed, and can meet the clinical use needs to the greatest extent. Before the above step 104, the method may also include: pre-training the physical information neural network. For example, in order to quickly reconstruct the viscoelastic SWE data of liver patients, the pre-training method can be: pre-input the vertical vibration velocity of multiple frames of particles of a patient's liver into the physical information neural network for solution; and use the model parameters of the physical information neural network after pre-training to convergence as the initialization settings for subsequent solutions.

[0057] In another exemplary embodiment of the present application, the above step 104 can be replaced by the following steps 301 to 309.

[0058] Step 301: Set the time and spatial scale ( , );in, is the real space scale of the horizontal axis, is the real space scale of the vertical axis.

[0059] Step 302: According to time and spatial scale ( , ), through the spatiotemporal neural network, the stream function of this prediction is obtained and the Lagrangian operator that satisfies the incompressibility constraint .

[0060] Step 303: Based on the stream function of this prediction , using the non-uniform viscoelastic wave equation, calculate the horizontal vibration velocity of the particle and the vertical vibration velocity of the particle .

[0061] Step 304: Based on the calculated vertical vibration velocity of the particle The data error loss is calculated by using the data error loss function in the loss function and the vertical vibration velocity of the multi-frame particle of the object to be measured.

[0062] Step 305: According to the spatial scale ( , ), through the spatial neural network, the predicted shear modulus is obtained and viscosity .

[0063] Step 306: Based on the predicted shear modulus , the viscosity predicted this time , the particle horizontal vibration speed and the vertical vibration velocity of the particle , use the partial differential equation loss function in the loss function to calculate the partial differential equation loss.

[0064] Step 307: Combine the partial differential equation loss and the data error loss to obtain the total loss of this time.

[0065] Step 308: If satisfied and , then adjust the network parameters of the physical information neural network and return to execute according to the time and spatial scale ( , ), through the spatiotemporal neural network, the stream function of this prediction is obtained and the Lagrangian operator that satisfies the incompressibility constraint Among them, The total loss for this time is is the initial loss, To set the threshold, is the number of iterations, is the maximum number of iterations.

[0066] Step 309: If satisfied or , then the predicted shear modulus is output and viscosity , as the spatial distribution of shear modulus and viscosity of the object to be measured.

[0067] In another exemplary embodiment of the present application, in order to minimize motion interference and electronic noise, the present application performs 30 times of superposition averaging to improve the signal-to-noise ratio. The ultrasonic radio frequency signal collected by the ultrasonic transducer is a modulated wave, which needs to be demodulated according to the center frequency of the probe. The commonly used method is IQ demodulation. After demodulation, a complex signal is obtained, including amplitude and phase information. Since the amplitude of the shear wave is often in the micron level, traditional structural imaging is not enough to distinguish such weak vibrations. Therefore, it is necessary to track the particle vibration velocity based on the IQ data. Considering the amount of calculation and stability, the particle vibration velocity estimation method of phase difference is selected to calculate the displacement between frames. The center frequency, sound speed and time interval between frames of the ultrasonic signal are known, and the particle vibration velocity of these N frames can be calculated by the phase difference. Then the process of determining the vertical vibration velocity of the multi-frame particle of the object to be measured in step 103 above can be replaced by the following steps 401 to 403.

[0068] Step 401: an ultrasonic transducer collects a plurality of continuous frames of ultrasonic radio frequency signals of an object to be measured, and performs IQ demodulation to obtain a demodulated signal of each frame.

[0069] Step 402: According to the demodulated signal of each frame, a particle vibration velocity estimation method based on phase difference is adopted to obtain the particle vibration displacement of each frame.

[0070] Step 403: Obtain the particle vibration velocity of each frame according to the particle vibration displacement of each frame; the particle vibration velocity includes the particle horizontal vibration velocity and the particle vertical vibration velocity.

[0071] In another exemplary embodiment of the present application, the calculation formula of the method for estimating the particle vibration velocity of phase difference is as follows.

[0072] .

[0073] In the formula, is the particle vibration displacement, is the speed of sound, is the center frequency, is the total number of frames, is the number of rows in the IQ data, For the Frame is the first in IQ data The orthogonal components of the rows, For the Frame is the first in IQ data The orthogonal components of the rows, For the Frame is the first in IQ data The same direction component of the line, For the Frame is the first in IQ data IQ data refers to the demodulated signal of each frame obtained after IQ demodulation.

[0074] This application can accurately solve the spatial distribution of shear modulus and viscosity through single-case particle vibration velocity data. This method cleverly utilizes the similarities between the back-propagation characteristics of neural networks and the chain derivative rule, so this method can effectively utilize the characteristics of automatic differentiation to achieve accurate solutions to partial differential equations. The single case refers to multiple frames of ultrasonic radio frequency signals collected at one time.

[0075] The ultrasonic viscoelasticity measurement method of the present application is a method for solving the spatial distribution of tissue viscoelasticity based on a physical information neural network. The beneficial effects of the method of the present application are as follows.

[0076] 1. This application proposes an innovative method for the first time, encoding the wave control equation in the Kelvin-Voigt model as a loss function, and designing a new network architecture by utilizing the automatic differentiation function of the neural network to achieve accurate and fast viscoelastic spatial inversion. This method significantly improves the interpretability of the model and optimizes the efficiency of the inversion process.

[0077] 2. The method of this application represents a leading PDE (Partial differential equation, an equation containing partial derivatives of unknown functions) solving technology, which does not require a complex and tedious model training process. It can directly make rapid predictions on each case data and accurately calculate the viscoelastic distribution that conforms to the wave equation, showing excellent generalization ability.

[0078] 3. Compared with the traditional TOF method that only relies on the shear wave velocity feature, the method of the present application makes full use of the multiple features of the collected data and effectively avoids the estimation error caused by shear wave scattering or reflection due to the inhomogeneity of the tissue. Therefore, the method of the present application is more applicable in complex human environments and provides more accurate and reliable diagnostic results.

[0079] Based on the same inventive concept, the embodiment of the present application also provides an ultrasonic viscoelasticity measuring device for implementing the ultrasonic viscoelasticity measuring method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more ultrasonic viscoelasticity measuring device embodiments provided below can refer to the limitations of the ultrasonic viscoelasticity measuring method above, and will not be repeated here.

[0080] In an exemplary embodiment, an ultrasonic viscoelasticity measuring device is provided, including: an equation building module, a neural network building module, a vibration velocity determination module, and an application module. The equation building module is used to establish a non-uniform viscoelastic wave equation; the non-uniform viscoelastic wave equation characterizes the relationship between the vertical vibration velocity of a particle and a stream function; the neural network building module is used to encode the non-uniform viscoelastic wave equation into a loss function and construct a physical information neural network; the physical information neural network includes a spatiotemporal neural network, a spatial neural network, and a loss function; the spatiotemporal neural network is used to predict the stream function; the non-uniform viscoelastic wave equation is used to obtain the vertical vibration velocity of a particle according to the output stream function; the spatial neural network is used to predict the shear modulus and viscosity; the loss function is used to calculate the total loss according to the vertical vibration velocity of a particle, the shear modulus, and the viscosity; the vibration velocity determination module is used to determine the vertical vibration velocity of a multi-frame particle of the object to be measured according to the continuous multi-frame ultrasonic radio frequency signal of the object to be measured; the application module is used to use the vertical vibration velocity of a multi-frame particle of the object to be measured as data supervision, input into the physical information neural network, and obtain the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured.

[0081] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an ultrasonic viscoelasticity measurement method is implemented.

[0082] Those skilled in the art will understand that Figure 3The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0083] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0084] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0085] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0086] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0087] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. An ultrasonic viscoelasticity measurement method, characterized in that: include: Establish the inhomogeneous viscoelastic wave equation; The non-uniform viscoelastic wave equation characterizes the relationship between the vertical vibration velocity of the particle and the stream function; The non-uniform viscoelastic wave equation is encoded as a loss function, and a physical information neural network is constructed; the physical information neural network includes a spatiotemporal neural network and a spatial neural network; the spatiotemporal neural network is used to predict the stream function; the non-uniform viscoelastic wave equation is used to obtain the vertical vibration velocity of the particle according to the output stream function; the spatial neural network is used to predict the shear modulus and viscosity; the loss function is used to calculate the total loss according to the vertical vibration velocity of the particle, the shear modulus and the viscosity; Determine the vertical vibration speed of the multi-frame particle of the object to be measured according to the continuous multi-frame ultrasonic radio frequency signal of the object to be measured; The vertical vibration velocity of multiple frames of the object to be measured is used as data supervision and input into the physical information neural network to obtain the spatial distribution of shear modulus and viscosity of the object to be measured; The vertical vibration velocity of the multi-frame particle of the object to be measured is used as data supervision and input into the physical information neural network to obtain the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured, which specifically includes: Set time and spatial scale ( , );in, is the real space scale of the horizontal axis, is the real space scale of the ordinate; According to time and spatial scale ( , ), through the spatiotemporal neural network, the stream function of this prediction is obtained and the Lagrangian operator that satisfies the incompressibility constraint ; According to the stream function of this prediction , using the non-uniform viscoelastic wave equation, calculate the horizontal vibration velocity of the particle and the vertical vibration velocity of the particle ; According to the calculated vertical vibration velocity of the particle and the vertical vibration velocity of the multi-frame particle of the object to be measured, and using the data error loss function in the loss function to calculate the data error loss; According to the spatial scale ( , ), through the spatial neural network, the predicted shear modulus is obtained and viscosity ; According to the predicted shear modulus , the viscosity predicted this time , the particle horizontal vibration speed and the vertical vibration velocity of the particle , using the partial differential equation loss function in the loss function, calculate the partial differential equation loss; Combine the partial differential equation loss and data error loss to get the total loss of this time; If satisfied and , then adjust the network parameters of the physical information neural network and return to execute according to the time and spatial scale ( , ), through the spatiotemporal neural network, the stream function of this prediction is obtained and the Lagrangian operator that satisfies the incompressibility constraint steps; wherein, The total loss for this time is is the initial loss, To set the threshold, is the number of iterations, is the maximum number of iterations; If satisfied or , then the predicted shear modulus is output and viscosity , as the spatial distribution of shear modulus and viscosity of the object to be measured.

2. The ultrasonic viscoelasticity measurement method according to claim 1, characterized in that: The inhomogeneous viscoelastic wave equation is established, including: According to Newton's second law and the shear stress-shear strain relationship, the expression form of the wave equation of inhomogeneous media with respect to particle displacement is determined as follows: ; In the formula, is the density, for Direction displacement, For time, is the shear modulus, is the Einstein summation rule parameter, is the spatial direction, is the viscosity, is the shift of the Einstein sum rule, is the Lagrange multiplier associated with the incompressible constraint; The expression of the wave equation of inhomogeneous media about the displacement of the particle is organized into an expression of the particle vibration velocity in two-dimensional space: ; In the formula, is the horizontal vibration velocity of the particle, is the vertical vibration velocity of the particle, is the real space scale of the horizontal axis, is the real space scale of the ordinate, Lagrangian operator that satisfies the incompressibility constraint Taking the partial derivative in the horizontal direction, Lagrangian operator that satisfies the incompressibility constraint Find the partial derivative in the vertical direction; By introducing the stream function and simplifying the expression of the particle vibration velocity in two-dimensional space, the non-uniform viscoelastic wave equation is obtained: ; In the formula, is the stream function.

3. The ultrasonic viscoelasticity measurement method according to claim 1, characterized in that: The loss function includes: a data error loss function and a partial differential equation loss function; The data error loss function is: ; In the formula, is the data error loss, is the mean square error, is the vertical vibration velocity of the particle, is the vertical vibration velocity of the particle used as data supervision; The partial differential equation loss function is: ； ； In the formula, is the first partial differential equation loss, Lagrangian operator that satisfies the incompressibility constraint Taking the partial derivative in the horizontal direction, is the density, Find the second-order derivative of the horizontal vibration velocity of the particle with respect to time, is the shear modulus, Find the second derivative of the horizontal vibration velocity of the particle in the horizontal direction, Find the second derivative of the horizontal vibration velocity of the particle with respect to the vertical direction, is the viscosity, Find the second derivative of the horizontal vibration velocity of the particle with respect to the horizontal direction and then with respect to time Find the partial derivative, Find the second derivative of the horizontal vibration velocity of the particle with respect to the vertical direction and then with respect to time Find the partial derivative, is the partial derivative of the shear modulus with respect to the horizontal direction, Find the partial derivative of the horizontal vibration velocity of the particle with respect to the horizontal direction, is the partial derivative of the shear modulus with respect to the vertical direction, To find the partial derivative of the horizontal vibration velocity of the particle with respect to the vertical direction, To find the partial derivative of the vertical vibration velocity of the particle with respect to the horizontal direction, is the partial derivative of viscosity with respect to the horizontal direction, Find the partial derivative of the horizontal vibration velocity of the particle with respect to the horizontal direction and then with respect to time Find the partial derivative, is the partial derivative of viscosity with respect to the vertical direction, Find the partial derivative of the horizontal vibration velocity of the particle with respect to the vertical direction and then with respect to time Find the partial derivative, Find the partial derivative of the vertical vibration velocity of the particle with respect to the horizontal direction and then with respect to time Find partial derivatives; is the second partial differential equation loss, Lagrangian operator that satisfies the incompressibility constraint Taking the partial derivative in the vertical direction, is the vertical vibration velocity of the particle relative to time Find the second-order derivative, Find the second derivative of the vertical vibration velocity of the particle with respect to the horizontal direction, Find the second derivative of the vertical vibration velocity of the particle in the vertical direction, The second derivative of the vertical vibration velocity of the particle with respect to the horizontal direction is calculated and then the derivative of the vertical vibration velocity with respect to the time Find the partial derivative, The second derivative of the vertical vibration velocity of the particle is calculated with respect to the vertical direction and then with respect to the time Find the partial derivative, is the partial derivative of the vertical vibration velocity of the particle with respect to the vertical direction, Find the partial derivative of the vertical vibration velocity of the particle with respect to the vertical direction and then with respect to time Find partial derivatives; The loss function is: ; In the formula, is the total loss, is the hyperparameter of the partial differential equation loss function, is the hyperparameter of the data error loss function.

4. The ultrasonic viscoelasticity measurement method according to claim 1, characterized in that: The vertical vibration velocity of the multi-frame particle of the object to be measured is used as data supervision and input into the physical information neural network to obtain the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured, which also includes: The physical information neural network is pre-trained.

5. The ultrasonic viscoelasticity measurement method according to claim 1, characterized in that: Determining the vertical vibration velocity of the multi-frame particle points of the object to be measured according to the continuous multi-frame ultrasonic radio frequency signals of the object to be measured, specifically including: Acquire a continuous multi-frame ultrasonic radio frequency signal of the object to be measured collected by the ultrasonic transducer, and obtain a demodulated signal of each frame through IQ demodulation; According to the demodulated signal of each frame, the particle vibration velocity estimation method of phase difference is adopted to obtain the particle vibration displacement of each frame; According to the particle vibration displacement of each frame, the particle vibration speed of each frame is obtained; the particle vibration speed includes the particle horizontal vibration speed and the particle vertical vibration speed.

6. The ultrasonic viscoelasticity measurement method according to claim 5, characterized in that: The calculation formula of the phase difference particle vibration velocity estimation method is: ； In the formula, is the particle vibration displacement, is the speed of sound, is the center frequency, is the total number of frames, is the number of rows in the IQ data, For the Frame is the first in IQ data The orthogonal components of the rows, For the Frame is the first in IQ data The orthogonal components of the rows, For the Frame is the first in IQ data The same direction component of the line, For the Frame is the first in IQ data The same direction component of the row.

7. An ultrasonic viscoelasticity measuring device, characterized in that: The ultrasonic viscoelasticity measuring device comprises: An equation building module is used to build a non-uniform viscoelastic wave equation; the non-uniform viscoelastic wave equation represents the relationship between the vertical vibration velocity of the particle and the stream function; A neural network construction module is used to encode the non-uniform viscoelastic wave equation into a loss function to construct a physical information neural network; the physical information neural network includes a spatiotemporal neural network, a spatial neural network and a loss function; the spatiotemporal neural network is used to predict the stream function; the non-uniform viscoelastic wave equation is used to obtain the vertical vibration velocity of the particle according to the output stream function; the spatial neural network is used to predict the shear modulus and viscosity; the loss function is used to calculate the total loss according to the vertical vibration velocity of the particle, the shear modulus and the viscosity; A vibration velocity determination module, used to determine the vertical vibration velocity of multiple frames of particles of the object to be measured according to the continuous multiple frames of ultrasonic radio frequency signals of the object to be measured; An application module is used to use the vertical vibration velocity of multiple frames of particles of the object to be measured as data supervision, input it into the physical information neural network, and obtain the spatial distribution of shear modulus and viscosity of the object to be measured; The vertical vibration velocity of the multi-frame particle of the object to be measured is used as data supervision and input into the physical information neural network to obtain the spatial distribution of the shear modulus and the spatial distribution of the viscosity of the object to be measured, which specifically includes: Set time and spatial scale ( , );in, is the real space scale of the horizontal axis, is the real space scale of the ordinate; According to time and spatial scale ( , ), through the spatiotemporal neural network, the stream function of this prediction is obtained and the Lagrangian operator that satisfies the incompressibility constraint ; According to the stream function of this prediction , using the non-uniform viscoelastic wave equation, calculate the horizontal vibration velocity of the particle and the vertical vibration velocity of the particle ; According to the calculated vertical vibration velocity of the particle and the vertical vibration velocity of the multi-frame particle of the object to be measured, and using the data error loss function in the loss function to calculate the data error loss; According to the spatial scale ( , ), through the spatial neural network, the predicted shear modulus is obtained and viscosity ; According to the predicted shear modulus , the viscosity predicted this time , the particle horizontal vibration speed and the vertical vibration velocity of the particle , using the partial differential equation loss function in the loss function, calculate the partial differential equation loss; Combine the partial differential equation loss and data error loss to get the total loss of this time; If satisfied and , then adjust the network parameters of the physical information neural network and return to execute according to the time and spatial scale ( , ), through the spatiotemporal neural network, the stream function of this prediction is obtained and the Lagrangian operator that satisfies the incompressibility constraint steps; wherein, The total loss for this time is is the initial loss, To set the threshold, is the number of iterations, is the maximum number of iterations; If satisfied or , then the predicted shear modulus is output and viscosity , as the spatial distribution of shear modulus and viscosity of the object to be measured.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the ultrasonic viscoelasticity measurement method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the ultrasonic viscoelasticity measurement method according to any one of claims 1 to 6 is implemented.

Images