A method, system and terminal for generating tissue mechanical characteristics
By preprocessing tissue information images and constructing a physical information neural network model, the robustness and accuracy problems of tissue mechanical feature generation in existing technologies are solved, and efficient calculation and analysis on low-quality images or images with speckles are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN UNIV
- Filing Date
- 2024-05-30
- Publication Date
- 2026-05-12
AI Technical Summary
Existing technologies are not robust enough when processing images of varying quality or those containing speckle noise, resulting in low accuracy and computational efficiency in tissue mechanical features.
The tissue biomechanical feature generation method is adopted. By acquiring tissue information images, preprocessing them, constructing a physical information neural network model, training the target network model, and calculating the tissue information images to be processed to generate tissue biomechanical features.
It improves the robustness and accuracy of tissue mechanical characteristics in cases of poor image quality or high speckle content, enabling more efficient calculation and analysis.
Smart Images
Figure CN118570164B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology, and in particular to a method, system, terminal, and computer-readable storage medium for generating tissue biomechanical features. Background Technology
[0002] Advanced computing technologies, such as machine learning and deep learning, are transforming the field of muscle function assessment. These technologies, through automated analysis of medical images and precise measurement of tissue biomechanical characteristics, are crucial for understanding human condition and function. For example, the application of physical-informed neural networks (PINNs) and ultrasound imaging techniques in tissue elasticity modulus measurement is becoming a cutting-edge research area.
[0003] However, traditional methods for quantifying the elastic modulus of tissue, such as shear wave elastography and grayscale ultrasound, while technologically advanced, still face challenges in terms of high sensitivity to image quality and accuracy and robustness when processing images with speckle noise. Furthermore, these techniques tend to employ simplified geometric assumptions, such as treating tissue as isotropic materials or following simple elastic modulus-strain distribution functions, which may be insufficient to describe the complex deformation of tissue during maximal contraction or complex movements. While deep learning-based medical imaging methods have made progress in some areas, accuracy still needs improvement at different stages of disease. These methods rely on large amounts of data for training and complex model structures, making them difficult to meet the demands of real-time processing. Moreover, these techniques are highly sensitive to image quality and may perform poorly on low-quality or speckled images.
[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention
[0005] The main objective of this invention is to provide a method, system, terminal, and computer-readable storage medium for generating tissue mechanical features. This invention aims to address the problems of low robustness, inability to comprehensively obtain tissue mechanical properties, low accuracy of mechanical features, and low computational efficiency of traditional deep learning methods when processing images of different quality or with high noise levels.
[0006] To achieve the above objectives, the present invention provides a method for generating tissue biomechanical features, the method comprising the following steps:
[0007] A tissue information image obtained by sampling the tissue is acquired; the tissue information image is preprocessed to obtain a preprocessed information image; and the mechanical feature information of the preprocessed information image is acquired.
[0008] A tissue biomechanical feature generation model is constructed, and the model is trained based on the preprocessed information image and the biomechanical feature information to obtain a target network model;
[0009] Obtain a target tissue information image of the tissue to be processed, input the target tissue information image into the target network model for calculation, and obtain the target tissue mechanical characteristics of the tissue to be processed;
[0010] The tissue to be treated is analyzed and evaluated based on the target tissue's mechanical characteristics to obtain the analysis results.
[0011] Optionally, the method for generating tissue mechanical features, wherein acquiring a tissue information image obtained by sampling the tissue, and preprocessing the tissue information image to obtain a preprocessed information image, specifically includes:
[0012] Acquire multiple tissue information obtained by scanning specific locations of multiple tissues using an imaging device;
[0013] Multiple tissue information images are obtained by performing envelope extraction, logarithmic stretching, and region selection on multiple tissue information images;
[0014] Multiple tissue information images are sampled and resampled to obtain multiple adjusted tissue information images;
[0015] Iterative backprojection and denoising processing are performed on multiple adjusted tissue information images to obtain multiple preprocessed information images.
[0016] Optionally, in the tissue mechanical feature generation method, the step of acquiring the mechanical feature information of the preprocessed information image specifically includes:
[0017] Intermediate feature information is obtained by processing the preprocessed information image using automatic image processing technology and high-resolution ultrasound imaging technology;
[0018] The intermediate feature information is verified using time series analysis and frequency domain analysis methods, and the verified intermediate feature information is used as the mechanical feature information of the preprocessed information image.
[0019] Optionally, the tissue biomechanical feature generation method, wherein training the tissue biomechanical feature generation model based on the preprocessed information image and the biomechanical feature information to obtain a target network model specifically includes:
[0020] The preprocessed information image is used as sample data, and the mechanical feature information is used as a label. A dataset is established based on the sample data and the label.
[0021] The dataset is divided into a training set and a test set according to a preset ratio. The training set is used to train the tissue biomechanical feature generation model to obtain a trained tissue biomechanical feature generation model. The test set is used to evaluate the performance of the trained tissue biomechanical feature generation model to obtain the target network model that meets the requirements.
[0022] Optionally, the method for generating tissue biomechanical features, wherein training the tissue biomechanical feature generation model using the training set to obtain the trained tissue biomechanical feature generation model specifically includes:
[0023] The preprocessed information image is input into the tissue mechanical feature generation model, which outputs dimensionless stress and dimensionless parameters.
[0024] The loss function is calculated based on the dimensionless stress, the dimensionless parameters, and the mechanical feature information. The tissue mechanical feature generation model is then adjusted based on the loss function until the loss function meets the preset requirements, thus obtaining the trained tissue mechanical feature generation model.
[0025] The dimensionless stress includes: transverse normal stress S xx Longitudinal normal stress S yy Shear stress S xy The dimensionless parameters include: dimensionless Lamé first parameter Λ and dimensionless Lamé second parameter M.
[0026] Optionally, in the method for generating tissue mechanical features, the step of calculating the loss function based on dimensionless stress, the dimensionless parameters, and the mechanical feature information specifically includes:
[0027] According to the transverse normal stress S xx The longitudinal normal stress S yy The shear stress S xy The dimensionless Lamé first parameter Λ, the dimensionless Lamé second parameter M, and the loss function (Cost Function) for calculating the mechanical characteristic information are as follows:
[0028]
[0029] Where, N Ω This indicates the number of coordinate points in the main region that participated in network training. This indicates the number of coordinate points in the upper and lower boundary regions that participate in network training. This represents the number of coordinate points in the left and right boundary regions participating in network training, where i represents either x or y, j represents either x or y, and x and y represent the horizontal and vertical coordinates, respectively. δ ijThe symbol S represents Kronecker. ij ε represents dimensionless stress. kk S represents the trace of all strain tensors. ix,x Represents the stress tensor S ix The partial derivative of S with respect to the x-coordinate iy,y Represents the stress tensor S iy The partial derivative with respect to the y-coordinate, This represents the known stress components on the left and right sides. This represents the known stress distribution at the top and bottom, ε. ij This represents the mechanical characteristic information corresponding to dimensionless stress.
[0030] Optionally, in the method for generating tissue mechanical features, the target tissue mechanical features include: elastic modulus, Poisson's ratio, transverse normal stress, longitudinal normal stress, and shear stress;
[0031] The step of inputting the target tissue information image into the target network model for calculation to obtain the target tissue mechanical characteristics of the tissue to be processed specifically includes:
[0032] The target tissue information image is input into the target network model to obtain the target dimensionless stress and the target dimensionless form parameters;
[0033] The maximum normal stress at the top boundary is obtained based on the average anatomical data. The first Lamé parameter and the second Lamé parameter are calculated based on the maximum normal stress and the dimensionless parameters of the target. The elastic modulus and the Poisson's ratio are calculated based on the first Lamé parameter and the second Lamé parameter.
[0034] The transverse normal stress, the longitudinal normal stress, and the shear stress are calculated based on the maximum normal stress and the target dimensionless stress.
[0035] Furthermore, to achieve the above objectives, the present invention also provides a tissue biomechanical feature generation system, wherein the tissue biomechanical feature generation system comprises:
[0036] The data acquisition module is used to acquire tissue information images obtained by sampling the tissue, preprocess the tissue information images to obtain preprocessed information images, and acquire the mechanical feature information of the preprocessed information images;
[0037] The model training module is used to construct a tissue mechanical feature generation model. The tissue mechanical feature generation model is trained based on the preprocessed information image and the mechanical feature information to obtain the target network model.
[0038] The model prediction module is used to acquire the target tissue information image of the tissue to be processed, input the target tissue information image into the target network model for calculation, and obtain the target tissue mechanical characteristics of the tissue to be processed.
[0039] The analysis and evaluation module is used to analyze and evaluate the tissue to be treated based on the target tissue's mechanical characteristics, and obtain the analysis results of the tissue to be treated.
[0040] In addition, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a tissue biomechanical feature generation program stored in the memory and executable on the processor, wherein when the tissue biomechanical feature generation program is executed by the processor, it implements the steps of the tissue biomechanical feature generation method as described above.
[0041] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a tissue biomechanical feature generation program, which, when executed by a processor, implements the steps of the tissue biomechanical feature generation method as described above.
[0042] In this invention, tissue information images are acquired, preprocessed to obtain preprocessed information images, and their mechanical feature information is obtained. A tissue mechanical feature generation model is constructed, and trained based on the preprocessed information images and mechanical feature information to obtain a target network model. A target tissue information image of the tissue to be processed is acquired, input into the target network model for calculation, and the target tissue mechanical features of the tissue to be processed are obtained. The analysis results of the tissue to be processed are obtained based on the target tissue mechanical features. This invention integrates physical principles and deep learning for the inference and calculation of tissue mechanical feature information, exhibiting higher robustness and accuracy even when processing images with poor quality and numerous blotches, thus improving the accuracy and reliability of tissue mechanical feature inference. Attached Figure Description
[0043] Figure 1 This is a flowchart of a preferred embodiment of the tissue mechanical feature generation method of the present invention;
[0044] Figure 2 This is a diagram of the architecture of the tissue mechanical feature generation model described in the tissue mechanical feature generation method of the present invention;
[0045] Figure 3 This is a schematic diagram of a preferred embodiment of the tissue mechanical feature generation system of the present invention;
[0046] Figure 4 This is a schematic diagram of the operating environment of a preferred embodiment of the terminal of the present invention. Detailed Implementation
[0047] This application provides a method and related equipment for generating tissue mechanical features. To make the objectives, technical solutions, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.
[0048] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.
[0049] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.
[0050] The tissue biomechanical feature generation method described in the preferred embodiment of the present invention, such as... Figure 1 As shown, the method for generating tissue biomechanical features includes the following steps:
[0051] Step S100: Obtain tissue information images obtained by sampling the tissue, preprocess the tissue information images to obtain preprocessed information images, and obtain the mechanical feature information of the preprocessed information images.
[0052] Specifically, the tissue is sampled to acquire multiple tissue information images (specific locations to be measured for tissue localization) obtained by scanning multiple tissue locations using an imaging device. Tissue imaging or signal information (such as radio frequency data) is recorded in real time to obtain tissue information images. Then, the multiple tissue information images are preprocessed. The preprocessing process includes: performing envelope extraction, logarithmic stretching, and region selection on the multiple tissue information images to obtain multiple tissue information images (which can be ultrasound video or continuous frame image data; in addition to video or continuous frame image analysis, continuous frame images are also used in this method); sampling and resampling the multiple tissue information images to obtain multiple adjusted tissue information images; and iterative backprojection and denoising processing are performed on the multiple adjusted tissue information images to obtain multiple preprocessed information images.
[0053] Understandably, when preprocessing ultrasound videos or consecutive frame images, sampling and resampling steps should be considered first to adjust the resolution, frame rate, or size of the video or consecutive frame images, especially when it is necessary to unify data from different devices to the same format. Iterative Back Projection (IBP) and deep learning methods can be applied to improve the resolution of video or consecutive frame images, which is particularly important for ultrasound images, which typically have low resolution, to enhance the network's learnability. Furthermore, noise reduction techniques can help remove interfering noise from video or consecutive frame images, while contrast enhancement makes image details clearer, especially crucial when displaying soft tissue structures.
[0054] Furthermore, the preprocessed information image is processed using automatic image processing technology and high-resolution ultrasound imaging technology to obtain intermediate feature information; the intermediate feature information is verified using time series analysis and frequency domain analysis methods, and the verified intermediate feature information is used as the mechanical feature information of the preprocessed information image.
[0055] Understandably, the preprocessed information images are processed using automated image processing techniques and high-resolution ultrasound imaging to capture image sequences of tissue under different stress states. Then, by calculating the positional changes of marker points between video or consecutive frame images, the deformation of the tissue under stress can be accurately quantified. This deformation is converted into strain data, which provides crucial information about the tissue's mechanical response and is a key component of the input to Physics-Informed Neural Networks (PINNs). To ensure the accuracy and reliability of the data, various validation measures can be employed, including time-series analysis and frequency domain analysis, to evaluate the consistency and repeatability of the strain data. This validated strain data is then used as the basis for training the PINN to build a model that accurately simulates the mechanical behavior of the tissue.
[0056] Step S200: Construct a tissue mechanical feature generation model. Train the tissue mechanical feature generation model based on the preprocessed information image and the mechanical feature information to obtain the target network model.
[0057] Specifically, a tissue biomechanical feature generation model is constructed, which is a PINN model. The input of this model is processed ultrasound video or continuous frame image data, and the collected tissue biomechanical feature parameter information is used as the reference information for the model input. The output of the model is the corresponding elastic modulus information of the tissue. Taking the PINN network as an example (the network can be modified to build a more complex network, and different basic neural networks can be used for different data conditions, such as convolutional neural networks, generative adversarial networks, recurrent neural networks, etc.), the specific network structure is shown in Figure 2.
[0058] The tissue biomechanical feature generation model consists of two fully connected deep neural networks. The network input (Input Layer) is a preprocessed information image, including the horizontal (x) and vertical (y) information of the ultrasound image. The intermediate hidden layer is a fully connected layer (which can be modified as needed, adding elements such as residual layers and activation functions). The output (Output Layer) is the horizontal normal stress S. xx Longitudinal normal stress S yy Shear stress S xy The dimensionless Lamé first parameter Λ and the dimensionless Lamé second parameter, where the Lamé parameters are a set of physical constants used to describe the elastic properties of materials, and are usually used to describe the elastic behavior of materials.
[0059] Further, training the tissue biomechanical feature generation model using the training set to obtain the trained tissue biomechanical feature generation model specifically includes:
[0060] The preprocessed information image is used as sample data, and the mechanical feature information is used as a label. A dataset is established based on the sample data and the label.
[0061] The dataset is divided into a training set and a test set according to a preset ratio (the division ratio is selected according to the actual situation, for example, the dataset can be divided into a training set and a test set in a 7:3 ratio). The training set is used to train the tissue biomechanical feature generation model to obtain the trained tissue biomechanical feature generation model. The test set is used to evaluate the performance of the trained tissue biomechanical feature generation model and detect the accuracy of the neural network model's prediction results to obtain the target network model that meets the requirements.
[0062] When evaluating the performance of the trained tissue biomechanical feature generation model using the test set, the trained model is applied to the test dataset, which includes preprocessed data and corresponding labels. The model predicts the tissue biomechanical features of each tissue in the test dataset, calculates the overall mean and variance between the tissue biomechanical features and the true tissue biomechanical features in the labels, and evaluates the degree of bias in the model results (data analysis methods such as Pearson correlation coefficient, Spearman correlation coefficient, consistency analysis, or regression analysis can also be used).
[0063] Furthermore, the step of training the tissue biomechanical feature generation model using the training set to obtain the trained tissue biomechanical feature generation model specifically includes:
[0064] The preprocessed information image is input into the tissue mechanical feature generation model, which outputs dimensionless stress and dimensionless parameters. A loss function is calculated based on the dimensionless stress, the dimensionless parameters, and the mechanical feature information. The tissue mechanical feature generation model is adjusted based on the loss function until the loss function meets the preset requirements, thus obtaining the trained tissue mechanical feature generation model.
[0065] The dimensionless stress includes: transverse normal stress S xx Longitudinal normal stress S yy Shear stress S xy The dimensionless parameters include: dimensionless Lamé first parameter Λ and dimensionless Lamé second parameter M (diminishing the dimensions of the parameters simplifies the calculation and reduces errors).
[0066] Loss calculation allows the network-generated data to be constrained within given physical rules, learning relevant organizational information and inferring predicted mechanical characteristics based on strain parameters obtained from grayscale ultrasound and input stress boundary conditions. During training, the convergence of the loss function allows the mechanical characteristics to gradually approximate the input information, meeting measurement requirements. The specific process of calculating the loss function is as follows:
[0067] According to the transverse normal stress S xx The longitudinal normal stress S yy The shear stress S xy The dimensionless Lamé first parameter Λ, the dimensionless Lamé second parameter M, and the loss function (Cost Function) for calculating the mechanical characteristic information are as follows:
[0068]
[0069] Where, N Ω This indicates the number of coordinate points in the main region that participated in network training. This indicates the number of coordinate points in the upper and lower boundary regions that participate in network training. This represents the number of coordinate points in the left and right boundary regions participating in network training, where i represents either x or y, j represents either x or y, and x and y represent the horizontal and vertical coordinates, respectively. δ ij The Kronecker notation is represented by δ when i and j are the same. ij Equal to 1, otherwise δ ij A value of 0 indicates the coupling between the material's elastic constants and the stress-strain relationship; S ij ε represents dimensionless stress. kk S represents the trace of all strain tensors. ix,x Represents the stress tensor S ix The partial derivative of S with respect to the x-coordinate iy,y Represents the stress tensor S iy The partial derivative with respect to the y-coordinate, This represents the known stress components on the left and right sides. This represents the known stress distribution at the top and bottom, ε. ij This represents the mechanical characteristic information corresponding to dimensionless stress.
[0070] Step S300: Obtain the target tissue information image of the tissue to be processed, input the target tissue information image into the target network model for calculation, and obtain the target tissue mechanical characteristics of the tissue to be processed.
[0071] It is understood that the target tissue mechanical characteristics include: elastic modulus, Poisson's ratio, transverse normal stress, longitudinal normal stress, and shear stress. After acquiring the target tissue information image of the tissue to be processed, the target tissue information image is input into the target network model for calculation to obtain the elastic modulus, Poisson's ratio, transverse normal stress, longitudinal normal stress, and shear stress.
[0072] The step of inputting the target tissue information image into the target network model for calculation to obtain the target tissue mechanical characteristics of the tissue to be processed specifically includes:
[0073] The target tissue information image is input into the target network model to obtain the target dimensionless stress and the target dimensionless parameters; the dimensionless stress includes: the target transverse normal stress S' xx Target longitudinal normal stress S' yy Target shear stress S' xy The dimensionless parameters of the target include: the dimensionless Lamé first parameter Λ' and the dimensionless Lamé second parameter M'.
[0074] The maximum normal stress σ0 at the top boundary is obtained based on average anatomical data (empirical value). The first Lamé parameter and the second Lamé parameter are calculated based on the maximum normal stress and the dimensionless parameters of the target. The elastic modulus and Poisson's ratio are then calculated based on the first Lamé parameter and the second Lamé parameter.
[0075] =σ0Λ', μ=σ0M';
[0076]
[0077] Wherein, λ and μ are the first and second Lamé parameters, respectively, E and v are the elastic modulus and Poisson's ratio, respectively, and σ0 is the maximum normal stress at the top boundary.
[0078] The transverse normal stress, the longitudinal normal stress, and the shear stress are calculated based on the maximum normal stress and the target dimensionless stress:
[0079]
[0080] Where, σ xx For the transverse normal stress, σ yy For the longitudinal normal stress, σ xy The shear stress is given.
[0081] Step S400: Analyze and evaluate the tissue to be treated based on the target tissue mechanical characteristics to obtain the analysis results of the tissue to be treated.
[0082] Specifically, the tissue to be treated is analyzed and evaluated based on the obtained elastic modulus, Poisson's ratio, transverse normal stress, longitudinal normal stress, and shear stress, and an analysis result (analysis report) of the tissue to be treated is generated. The user can obtain the physiological condition of the tissue to be treated based on the analysis result.
[0083] It should be noted that different initial stress distribution data or initial stress distribution assumptions are required for different tissues and organs in this invention. That is, different initial conditions are required for different tissues or organs, and different elasticity theories are required for different tissues and organs. For example, linear elasticity laws are used to describe certain solid organs, while nonlinear elasticity laws are used to describe certain organs containing blocky structures.
[0084] This invention obtains the strain distribution of a tissue through grayscale ultrasound video or continuous frame images. Combined with a physical information neural network and constrained by physical elasticity theory, it obtains parameters related to the tissue's mechanical characteristics, such as elastic modulus distribution, stress distribution, and Poisson's ratio distribution. Visually, these parameters can be used to assess the tissue's condition without the need for external excitation sources, such as quasi-static elastography or shear wave elastography.
[0085] As can be seen, this invention acquires tissue information images, preprocesses them to obtain preprocessed information images, and acquires the mechanical feature information of the preprocessed information images. A tissue mechanical feature generation model is constructed, and the model is trained based on the preprocessed information images and mechanical feature information to obtain a target network model. A target tissue information image of the tissue to be processed is acquired, and the target tissue information image is input into the target network model for calculation to obtain the target tissue mechanical features of the tissue to be processed. The analysis results of the tissue to be processed are obtained based on the target tissue mechanical features. This invention integrates physical principles and deep learning for inference and calculation of tissue mechanical feature information, enabling more accurate and real-time calculation of human tissue physiological information. Simultaneously, it has strong generalization ability and can be used on different tissues. Furthermore, training the physical information neural network requires only a small amount of video or continuous frame image data from different periods, with low system and hardware requirements, and can be deployed on different devices while maintaining a relatively fast operating speed. It exhibits higher robustness and accuracy when processing images with poor quality and many speckles, improving the accuracy and reliability of tissue mechanical feature inference.
[0086] Furthermore, such as Figure 3 As shown, based on the above-described method for generating tissue biomechanical features, the present invention also provides a system for generating tissue biomechanical features, wherein the system comprises:
[0087] The data acquisition module 41 is used to acquire tissue information images obtained by sampling the tissue, preprocess the tissue information images to obtain preprocessed information images, and acquire the mechanical feature information of the preprocessed information images;
[0088] Model training module 42 is used to construct a tissue mechanical feature generation model, and to train the tissue mechanical feature generation model based on the preprocessed information image and the mechanical feature information to obtain a target network model;
[0089] Model prediction module 43 is used to acquire the target tissue information image of the tissue to be processed, input the target tissue information image into the target network model for calculation, and obtain the target tissue mechanical characteristics of the tissue to be processed.
[0090] The analysis and evaluation module 44 is used to analyze and evaluate the tissue to be treated based on the target tissue's mechanical characteristics, and obtain the analysis results of the tissue to be treated.
[0091] Furthermore, such as Figure 4 As shown, based on the above-mentioned method and system for generating tissue biomechanical features, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 4 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.
[0092] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a tissue biomechanical feature generation program 40, which can be executed by the processor 10 to implement the tissue biomechanical feature generation method of this application.
[0093] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the tissue biomechanical feature generation method.
[0094] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.
[0095] In one embodiment, when the processor 10 executes the tissue biomechanical feature generation program 40 in the memory 20, the following steps are performed:
[0096] A tissue information image obtained by sampling the tissue is acquired; the tissue information image is preprocessed to obtain a preprocessed information image; and the mechanical feature information of the preprocessed information image is acquired.
[0097] A tissue biomechanical feature generation model is constructed, and the model is trained based on the preprocessed information image and the biomechanical feature information to obtain a target network model;
[0098] Obtain a target tissue information image of the tissue to be processed, input the target tissue information image into the target network model for calculation, and obtain the target tissue mechanical characteristics of the tissue to be processed;
[0099] The tissue to be treated is analyzed and evaluated based on the target tissue's mechanical characteristics to obtain the analysis results.
[0100] Specifically, the step of acquiring a tissue information image obtained by sampling the tissue, and preprocessing the tissue information image to obtain a preprocessed information image, includes:
[0101] Acquire multiple tissue information obtained by scanning specific locations of multiple tissues using an imaging device;
[0102] Multiple tissue information images are obtained by performing envelope extraction, logarithmic stretching, and region selection on multiple tissue information images;
[0103] Multiple tissue information images are sampled and resampled to obtain multiple adjusted tissue information images;
[0104] Iterative backprojection and denoising processing are performed on multiple adjusted tissue information images to obtain multiple preprocessed information images.
[0105] Specifically, obtaining the mechanical feature information of the preprocessed information image includes:
[0106] Intermediate feature information is obtained by processing the preprocessed information image using automatic image processing technology and high-resolution ultrasound imaging technology;
[0107] The intermediate feature information is verified using time series analysis and frequency domain analysis methods, and the verified intermediate feature information is used as the mechanical feature information of the preprocessed information image.
[0108] The step of training the tissue mechanical feature generation model based on the preprocessed information image and the mechanical feature information to obtain the target network model specifically includes:
[0109] The preprocessed information image is used as sample data, and the mechanical feature information is used as a label. A dataset is established based on the sample data and the label.
[0110] The dataset is divided into a training set and a test set according to a preset ratio. The training set is used to train the tissue biomechanical feature generation model to obtain a trained tissue biomechanical feature generation model. The test set is used to evaluate the performance of the trained tissue biomechanical feature generation model to obtain the target network model that meets the requirements.
[0111] The step of training the tissue biomechanical feature generation model using the training set to obtain the trained tissue biomechanical feature generation model specifically includes:
[0112] The preprocessed information image is input into the tissue mechanical feature generation model, which outputs dimensionless stress and dimensionless parameters.
[0113] The loss function is calculated based on the dimensionless stress, the dimensionless parameters, and the mechanical feature information. The tissue mechanical feature generation model is then adjusted based on the loss function until the loss function meets the preset requirements, thus obtaining the trained tissue mechanical feature generation model.
[0114] The dimensionless stress includes: transverse normal stress S xx Longitudinal normal stress S yy Shear stress S xy The dimensionless parameters include: dimensionless Lamé first parameter Λ and dimensionless Lamé second parameter M.
[0115] Specifically, the calculation of the loss function based on the dimensionless stress, the dimensionless parameters, and the mechanical characteristic information includes:
[0116] According to the transverse normal stress S xx The longitudinal normal stress S yy The shear stress S xy The dimensionless Lamé first parameter Λ, the dimensionless Lamé second parameter M, and the loss function (Cost Function) for calculating the mechanical characteristic information are as follows:
[0117]
[0118] Where, N Ω This indicates the number of coordinate points in the main region that participated in network training. This indicates the number of coordinate points in the upper and lower boundary regions that participate in network training. This represents the number of coordinate points in the left and right boundary regions participating in network training, where i represents either x or y, j represents either x or y, and x and y represent the horizontal and vertical coordinates, respectively. δ ij The symbol S represents Kronecker. ij ε represents dimensionless stress. kk S represents the trace of all strain tensors. ix,x Represents the stress tensor S ix The partial derivative of S with respect to the x-coordinate iy,y Represents the stress tensor S iy The partial derivative with respect to the y-coordinate, This represents the known stress components on the left and right sides. This represents the known stress distribution at the top and bottom, ε. ij This represents the mechanical characteristic information corresponding to dimensionless stress.
[0119] The target tissue mechanical characteristics include: elastic modulus, Poisson's ratio, transverse normal stress, longitudinal normal stress, and shear stress;
[0120] The step of inputting the target tissue information image into the target network model for calculation to obtain the target tissue mechanical characteristics of the tissue to be processed specifically includes:
[0121] The target tissue information image is input into the target network model to obtain the target dimensionless stress and the target dimensionless form parameters;
[0122] The maximum normal stress at the top boundary is obtained based on the average anatomical data. The first Lamé parameter and the second Lamé parameter are calculated based on the maximum normal stress and the dimensionless parameters of the target. The elastic modulus and the Poisson's ratio are calculated based on the first Lamé parameter and the second Lamé parameter.
[0123] The transverse normal stress, the longitudinal normal stress, and the shear stress are calculated based on the maximum normal stress and the target dimensionless stress.
[0124] In summary, this invention provides a method and system for generating tissue mechanical features. The method includes: acquiring ultrasound video containing tissue structure information; preprocessing the ultrasound video; performing computational inference on a trained physical neural network of the preprocessed ultrasound video to obtain structural and physiological parameters, obtaining analysis results, and using these results for the analysis and evaluation of tissue physiological state. This invention, based on a physical information neural network, integrates physical principles and deep learning, simultaneously performing computational inference on tissue structure and physiological parameters. This achieves the goal of synchronously and comprehensively estimating tissue structure and physiological parameters using ultrasound examination, reproducing the process of wave propagation in tissue, and providing higher interpretability. The method exhibits higher robustness and accuracy when dealing with image quality differences and numerous speckles, does not rely on large amounts of data, improves the reliability of muscle contraction activity analysis, and enhances the accuracy of tissue parameter generation.
[0125] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.
[0126] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0127] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.
Claims
1. A method for generating tissue mechanical characteristics, characterized in that, The method for generating tissue biomechanical features includes: A tissue information image obtained by sampling the tissue is acquired; the tissue information image is preprocessed to obtain a preprocessed information image; and the mechanical feature information of the preprocessed information image is acquired. A tissue biomechanical feature generation model is constructed, and the model is trained based on the preprocessed information image and the biomechanical feature information to obtain a target network model; Obtain a target tissue information image of the tissue to be processed, input the target tissue information image into the target network model for calculation, and obtain the target tissue mechanical characteristics of the tissue to be processed; The tissue to be treated is analyzed and evaluated based on the target tissue's mechanical characteristics to obtain the analysis results of the tissue to be treated; The step of training the tissue biomechanical feature generation model based on the preprocessed information image and the biomechanical feature information to obtain the target network model specifically includes: The preprocessed information image is used as sample data, and the mechanical feature information is used as a label. A dataset is established based on the sample data and the label. The dataset is divided into training and testing sets according to a preset ratio. The preprocessed information images are input into the tissue mechanical feature generation model, which outputs dimensionless stress and dimensionless parameters to reduce errors. The dimensionless stress includes transverse normal stress. Longitudinal normal stress Shear stress The dimensionless parameters include: the dimensionless Lamé first parameter. and the dimensionless form of Lamé's second parameter ; According to the transverse normal stress The longitudinal normal stress The shear stress The dimensionless form of the first parameter of Lamé. The dimensionless form of the second parameter of Lamé. Calculate the loss function using the mechanical feature information. : ； in, This indicates the number of coordinate points in the main region that participated in network training. This indicates the number of coordinate points in the upper and lower boundary regions that participate in network training. This indicates the number of coordinate points in the left and right boundary regions that participated in network training. express or , express or , , Representing the x-axis and y-axis respectively, The symbol for Kronecker, Indicates dimensionless stress. Represents the trace of all strain tensors. Represents the stress tensor Compared to Partial derivatives of coordinates, Represents the stress tensor Compared to Partial derivatives of coordinates, This represents the known stress components on the left and right sides. This represents the known stress distribution at the top and bottom. This represents the mechanical characteristic information corresponding to dimensionless stress; The tissue mechanical feature generation model is adjusted according to the loss function until the loss function meets the preset requirements to obtain the trained tissue mechanical feature generation model. The performance of the trained tissue mechanical feature generation model is evaluated using the test set to obtain the target network model that meets the requirements. The target tissue mechanical characteristics include: elastic modulus, Poisson's ratio, transverse normal stress, longitudinal normal stress, and shear stress; The step of inputting the target tissue information image into the target network model for calculation to obtain the target tissue mechanical characteristics of the tissue to be processed specifically includes: The target tissue information image is input into the target network model to obtain the target dimensionless stress and the target dimensionless form parameters; The maximum normal stress at the top boundary is obtained based on the average anatomical data. The first Lamé parameter and the second Lamé parameter are calculated based on the maximum normal stress and the dimensionless parameters of the target. The elastic modulus and the Poisson's ratio are calculated based on the first Lamé parameter and the second Lamé parameter. The transverse normal stress, the longitudinal normal stress, and the shear stress are calculated based on the maximum normal stress and the target dimensionless stress.
2. The method for generating tissue biomechanical characteristics according to claim 1, characterized in that, The step of acquiring a tissue information image obtained by sampling the tissue, and preprocessing the tissue information image to obtain a preprocessed information image, specifically includes: Acquire multiple tissue information obtained by scanning specific locations of multiple tissues using an imaging device; Multiple tissue information images are obtained by performing envelope extraction, logarithmic stretching, and region selection on multiple tissue information images; Multiple tissue information images are sampled and resampled to obtain multiple adjusted tissue information images; Iterative backprojection and denoising processing are performed on multiple adjusted tissue information images to obtain multiple preprocessed information images.
3. The method for generating tissue biomechanical characteristics according to claim 1, characterized in that, The acquisition of the mechanical feature information of the preprocessed information image specifically includes: Intermediate feature information is obtained by processing the preprocessed information image using automatic image processing technology and high-resolution ultrasound imaging technology; The intermediate feature information is verified using time series analysis and frequency domain analysis methods, and the verified intermediate feature information is used as the mechanical feature information of the preprocessed information image.
4. A system for generating tissue mechanical characteristics, characterized in that, The tissue biomechanical feature generation system is used to implement the tissue biomechanical feature generation method according to any one of claims 1-3, and the tissue biomechanical feature generation system includes: The data acquisition module is used to acquire tissue information images obtained by sampling the tissue, preprocess the tissue information images to obtain preprocessed information images, and acquire the mechanical feature information of the preprocessed information images; The model training module is used to construct a tissue mechanical feature generation model. The tissue mechanical feature generation model is trained based on the preprocessed information image and the mechanical feature information to obtain the target network model. The model prediction module is used to acquire the target tissue information image of the tissue to be processed, input the target tissue information image into the target network model for calculation, and obtain the target tissue mechanical characteristics of the tissue to be processed. The analysis and evaluation module is used to analyze and evaluate the tissue to be treated based on the target tissue's mechanical characteristics, and obtain the analysis results of the tissue to be treated.
5. A terminal, characterized in that, The terminal includes: a memory, a processor, and a tissue biomechanical feature generation program stored in the memory and executable on the processor, wherein when the tissue biomechanical feature generation program is executed by the processor, it implements the steps of the tissue biomechanical feature generation method as described in any one of claims 1-3.
6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a tissue biomechanical feature generation program, which, when executed by a processor, implements the steps of the tissue biomechanical feature generation method as described in any one of claims 1-3.