Photoacoustic image reconstruction method, device, equipment and medium

Abstract

This application discloses a photoacoustic image reconstruction method, apparatus, device, and medium, relating to the field of photoacoustic imaging. The method includes acquiring time-varying acoustic pressure data of target biological tissue at multiple preset locations; determining a photoacoustic wave equation; embedding the photoacoustic wave equation into the loss function of a physical information neural network to obtain a total loss function; constructing a data sample set using the spatiotemporal coordinates consisting of preset locations and times as input and the time-varying acoustic pressure data as labels; training the physical information neural network using the data sample set based on the total loss function to obtain a trained physical information neural network; inputting the spatiotemporal coordinates at an initial moment into the trained physical information neural network, outputting the acoustic pressure data at the initial moment, and thereby obtaining an initial acoustic pressure distribution. This application enables accurate reconstruction of photoacoustic images.

Description

Technical Field

[0001] The present application relates to the field of photoacoustic imaging, and in particular to a photoacoustic image reconstruction method, device, equipment and medium. Background Art

[0002] Photoacoustic computed tomography (PACT) is a rapidly developing biomedical imaging technology that has garnered widespread attention in recent years. Photoacoustic image reconstruction plays a crucial role in PACT. Combining the high contrast of optical imaging with the high resolution of ultrasound, PACT, as an emerging noninvasive biomedical imaging modality, holds enormous potential for cutting-edge exploration and clinical application in medical imaging.

[0003] Existing PACT methods mainly include time reversal (TR), filtered back-projection (FBP) and deep learning (DL).

[0004] The time reversal method is based on the time reversal symmetry principle of the wave equation, and realizes image reconstruction by inversely solving the forward problem of sound field propagation. Its physical implementation approach is: the received sound pressure signal is reversed in time domain as the boundary condition, and is reversely propagated to the computational domain in the numerical model, and the sound source distribution is reconstructed through the acoustic energy convergence effect. This method utilizes the reversible characteristics of wave propagation and was originally applied to the study of precise focusing of ultrasound in inhomogeneous media. It has the dual advantages of adaptability to complex acoustic media and compatibility with irregular detection geometry. Specifically, the time reversal method can use known limited field of view data to restore the source point information by reverse propagating the signal, and effectively reconstruct the image.

[0005] Overall, the TR method has become the most commonly used photoacoustic image reconstruction method due to its relatively simple operation, high measurement accuracy, and fast computational speed. However, the reconstruction results of this method are still affected by factors such as the number and bandwidth of detectors, as well as boundary conditions. The reconstructed image is prone to artifacts and is very sensitive to noise.

[0006] The basic idea of ​​the filtered back projection method is to first filter the received photoacoustic data to remove noise and enhance the signal. Then, the back projection algorithm is used to convert the filtered data into a two-dimensional or three-dimensional image of the target object. The FBP algorithm has high computational efficiency, simple implementation, and can quickly generate high-quality images with full projection angle coverage. However, the applicability of this method is limited by physical assumptions. The FBP algorithm assumes that the speed of sound in the medium is uniform, while the acoustic heterogeneity of biological tissues may cause wavefront distortion, resulting in a decrease in imaging quality. In addition, the FBP algorithm has high requirements for data integrity and requires a long scanning time to obtain sufficient projection data, which to some extent limits its application in real-time imaging.

[0007] Deep learning has been widely applied to various image processing and pattern recognition tasks, particularly in extracting temporal and spatial features of wave propagation. Through automated learning and data-driven features, neural networks are able to effectively capture the complex patterns in photoacoustic wave propagation. However, deep learning training typically relies on large, high-quality datasets, often requiring paired training data, which is difficult to obtain in practice.

[0008] Reference images used for network training in DL methods, such as full-bandwidth photoacoustic images, are often difficult to obtain, and the network can only be trained using simulated datasets. This training method based on simulated and phantom data differs from the actual situation of real human tissue, resulting in limited generalization and interpretability of the model when applied to the actual human body. In addition, the black box nature of deep learning models makes them difficult to interpret and analyze, further limiting their potential for clinical translation. Therefore, although deep learning has certain advantages in photoacoustic computed tomography, its limitations have also significantly hindered its widespread application.

[0009] The above PACT method can be summarized into the following questions (1)-(5).

[0010] (1) Traditional PACT methods rely on full-angle detection and densely sampled signals. However, this is difficult to achieve in practice, leading to the ill-posed problem of reconstructing images from incomplete projections.

[0011] (2) Traditional PACT methods, such as those based on time reversal and filtered back projection, are subject to inherent limitations of their algorithms, and reconstructed images are prone to artifacts. For example, the conclusion that the impulse response of the FBP algorithm is a delta function is only valid under ideal conditions, such as unbounded plane, unbounded cylinder, and closed spherical imaging geometries and infinite detector bandwidth. However, such ideal conditions can never be fully met in real experiments. Deviations from these ideal conditions will contaminate the PSF (Point Spread Function), ultimately reducing image quality.

[0012] (3) The traditional PACT method is very sensitive to noise and it is difficult to reconstruct a smooth sound pressure distribution map. Therefore, it is necessary to develop a sound pressure spatial distribution inversion algorithm with strong noise resistance.

[0013] (4) Currently, photoacoustic tomography technology based on deep learning can effectively improve the quality of image reconstruction, but the performance of DL methods is heavily dependent on the quality and quantity of training data. Training data is usually required to be paired, and obtaining a large amount of paired data is difficult in practice.

[0014] (5) Reference images used for network training in DL methods, such as full-bandwidth photoacoustic images, are often difficult to obtain, and the network can only be trained using simulated datasets. Since experimental conditions are usually much more complex than simulations, training the network using only simulated datasets will lead to degraded reconstruction performance and poor robustness. At the same time, training methods that rely entirely on data-driven methods have drawbacks such as poor interpretability and poor generalization.

[0015] In summary, existing PACT methods cannot accurately reconstruct photoacoustic images. Summary of the Invention

[0016] The purpose of this application is to provide a photoacoustic image reconstruction method, device, equipment and medium, which can realize accurate reconstruction of photoacoustic images.

[0017] To achieve the above objectives, this application provides the following solutions.

[0018] In the first aspect, the present application provides a photoacoustic image reconstruction method, comprising: obtaining time-varying sound pressure data of a target biological tissue at multiple preset positions; determining a photoacoustic wave equation; embedding the photoacoustic wave equation into a loss function of a physical information neural network to obtain a total loss function; constructing a data sample set using the space-time coordinates consisting of preset positions and times as input and the time-varying sound pressure data as labels; based on the total loss function, using the data sample set to train the physical information neural network to obtain a trained physical information neural network; inputting the space-time coordinates of the initial moment into the trained physical information neural network, and outputting the sound pressure data at the initial moment, thereby obtaining an initial sound pressure distribution.

[0019] Optionally, determining the photoacoustic wave equation specifically includes: establishing three coupled acoustic equations as 、 and Where, is the spatial position, and Represent the horizontal and vertical coordinates of the spatial position, For in time Particles are located in space The speed at is the particle mass density, is the gradient operator; For spatial location The particle mass density at For spatial location The speed of sound at ; is the coefficient of thermal expansion, For in time Located in space Sound pressure data at is the specific heat capacity at constant pressure, In time Particles are located in space The temperature at Represents the heat energy deposited per unit volume per unit time; Based on the three coupled acoustic equations, the photoacoustic wave equation is determined as: Where, is the Laplace operator.

[0020] Optionally, the initial condition satisfied by the photoacoustic wave equation is: and Where, is located at the spatial position at the initial moment The sound pressure data at The spatial position obtained by the photoacoustic effect produced by pulsed laser excitation The sound pressure data at .

[0021] Optionally, the photoacoustic wave equation is embedded into the loss function of the physical information neural network to obtain the total loss function, specifically including: establishing a data-driven loss function as: Where, is the data-driven loss, is the mean square error, is the sound pressure data of the target biological tissue in the historical time and space area, The sound pressure data output by the physical information neural network; the photoacoustic wave equation is converted into a homogeneous wave equation: According to the homogeneous wave equation, the physical information loss function is established as: Where, is the physical information loss; according to the initial conditions satisfied by the photoacoustic wave equation, the initial condition loss function is established as: Where, is the initial condition loss; according to the data-driven loss function, physical information loss function and initial condition loss function, the total loss function is obtained as follows: Where, is the total loss function, is the weight of physical information loss, is the weight of the data-driven loss, is the weight of the initial condition loss.

[0022] Optionally, the physical information neural network includes: an input layer, an output layer and multiple hidden layers; the input layer includes 3 input parameters; each hidden layer includes 100 neurons; and the output layer includes 1 output parameter.

[0023] Optionally, when using a data sample set to train a physical information neural network: a plurality of samples are extracted from the entire spatiotemporal domain using a Latin hypercube sampling method for training every preset number of rounds.

[0024] Optionally, the physical information neural network is a pre-trained physical information neural network.

[0025] In a second aspect, the present application provides a photoacoustic image reconstruction device, comprising: a data acquisition module, an equation determination module, a loss function establishment module, a sample set construction module, a training module and an application module.

[0026] A data acquisition module is used to obtain the time-varying sound pressure data of the target biological tissue at multiple preset positions. An equation determination module is used to determine the photoacoustic wave equation. A loss function establishment module is used to embed the photoacoustic wave equation into the loss function of the physical information neural network to obtain a total loss function. A sample set construction module is used to construct a data sample set using the space-time coordinates consisting of preset positions and times as inputs and the time-varying sound pressure data as labels. A training module is used to train the physical information neural network using the data sample set based on the total loss function to obtain a trained physical information neural network. An application module is used to input the space-time coordinates of the initial moment into the trained physical information neural network and output the sound pressure data at the initial moment, thereby obtaining the initial sound pressure distribution.

[0027] 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 one of the above-described photoacoustic image reconstruction methods.

[0028] 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 one of the above-described photoacoustic image reconstruction methods.

[0029] According to the specific embodiments provided in this application, this application has the following technical effects.

[0030] The present application provides a method, apparatus, device and medium for photoacoustic image reconstruction, which applies a physical information neural network to photoacoustic computed tomography. The physical information neural network can extract more spatiotemporal features of photoacoustic waves, and avoid estimation errors caused by incomplete projection data to the greatest extent in limited viewing angles and sparse sampling. Compared with traditional deep learning methods, the photoacoustic wave equation is encoded as a loss function, so that the physical information neural network no longer learns data features in a disorderly manner, but instead uses physical laws as a guide to achieve parameter solution. At the same time, the physical information neural network no longer relies solely on the stacking of a large amount of training data, which greatly reduces the demand for training data volume. Since the physical information neural network combines the physical information equation, it has a good filtering effect on the data and has good noise resistance compared to traditional methods, thereby enabling accurate reconstruction of photoacoustic images. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments. 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 any creative work.

[0032] Figure 1 A schematic flow chart of a photoacoustic image reconstruction method provided in one embodiment of the present application.

[0033] Figure 2 A schematic diagram of the training process of a physical information neural network provided in another embodiment of the present application.

[0034] Figure 3 A schematic diagram of the functional modules of a photoacoustic image reconstruction device provided in one embodiment of the present application.

[0035] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0037] In order to make the above-mentioned purposes, 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.

[0038] Physics-Informed Neural Networks (PINNs), an emerging approach that combines the powerful modeling capabilities of deep learning with the constraints of physical equations, are becoming a powerful tool for solving tissue acoustic inversion problems. PINNs encode the governing equations of a physical problem, such as partial differential equations, within a fully connected neural network, incorporating physical laws as part of the neural network training. They retain the powerful learning capabilities of deep learning models while also incorporating the laws of physics, making PINNs more robust and physically interpretable when faced with complex problems. PINNs have gradually become an effective tool for solving both direct and inverse problems, showing strong application prospects, particularly in situations involving complex multi-physics coupling.

[0039] Because PINN can more accurately capture the physical properties of wave processes through the constraints of physical equations, this method is expected to overcome many limitations of existing technologies. Specifically, when dealing with highly heterogeneous tissues, PINN can more accurately describe the complexities of photoacoustic wave propagation, thereby providing a new solution for accurately estimating the initial acoustic pressure distribution in photoacoustic imaging. With in-depth research in this field, the PINN method has the potential to significantly enhance the clinical application value of photoacoustic computed tomography technology and promote its widespread application in precision medicine.

[0040] In view of the above, in an exemplary embodiment, Figure 1 As shown, a photoacoustic image reconstruction method is provided, including the following steps 101 to 106.

[0041] Step 101: Acquire time-varying acoustic pressure data of a target biological tissue at a plurality of preset positions.

[0042] Step 102: Determine the photoacoustic wave equation.

[0043] Step 103: Embed the photoacoustic wave equation into the loss function of the physical information neural network to obtain the total loss function.

[0044] Step 104: Using the space-time coordinates consisting of the preset position and time as input and the sound pressure data varying with time as labels, a data sample set is constructed.

[0045] Step 105: Based on the total loss function, the data sample set is used to train the physical information neural network to obtain a trained physical information neural network.

[0046] Step 106: Input the space-time coordinates of the initial moment into the trained physical information neural network, output the sound pressure data of the initial moment, and thus obtain the initial sound pressure distribution.

[0047] By implementing steps 101 to 106 above, applying PINN to photoacoustic computed tomography and embedding the photoacoustic wave equation into the PINN, the problem of poor image quality of sparse sampling reconstruction in the existing PACT method is solved, and the PINN of the present application has stronger noise resistance.

[0048] In another exemplary embodiment of the present application, in the above step 101, a pulsed laser is used to emit laser pulses to the target biological tissue, the target biological tissue expands due to heat and generates an ultrasonic signal, and an ultrasonic transducer is used to collect the ultrasonic signal. The data collected by the ultrasonic transducer is the spatial position of the ultrasonic transducer. The time-varying sound pressure signal collected at the location can be used to form a time-space coordinate using the position and time of the ultrasonic transducer.

[0049] In another exemplary embodiment of the present application, the generation and propagation of photoacoustic waves can be mathematically modeled using the photoacoustic wave equation. Then, the above step 102 can be replaced by the following steps 201 to 202.

[0050] Step 201: Establish three coupled acoustic equations:

[0051] Linearized equations of motion: ;

[0052] Linearized continuity equation: ;

[0053] Thermoelasticity equation: ;

[0054] Where, is the spatial position, and Represent the horizontal and vertical coordinates of the spatial position, For in time Particles are located in space The speed at is the particle mass density, is the gradient operator, For spatial location The particle mass density at For spatial location The speed of sound at ; is the coefficient of thermal expansion, For in time Located in space Sound pressure data at is the specific heat capacity at constant pressure, In time Particles are located in space The temperature at Represents the heat energy deposited per unit volume per unit time.

[0055] Step 202: Based on the three coupled acoustic equations, the photoacoustic wave equation is determined as:

[0056] ;

[0057] Where, is the Laplace operator.

[0058] For example, the initial conditions satisfied by the photoacoustic wave equation are:

[0059] ;

[0060] ;

[0061] Where, is located at the spatial position at the initial moment The sound pressure data at The spatial position obtained by the photoacoustic effect produced by pulsed laser excitation The sound pressure data at .

[0062] In another exemplary embodiment of the present application, the PINN network structure designed by the present application is as follows: Figure 2 As shown, PINN contains three input parameters ( 、 and ), 8 hidden layers and one output parameter , each hidden layer includes 100 neurons. The output of PINN is trained to learn to fit the input data and solve the photoacoustic wave equation. By fitting the input data (the sound pressure data collected by the ultrasonic transducer), PINN will infer the propagation process of the photoacoustic wave in the entire space-time region (that is, the sound pressure distribution in the entire space-time region). By setting it to 0, the sound pressure distribution at the initial moment can be obtained, which is the ultimate goal of photoacoustic image reconstruction in this application.

[0063] In another exemplary embodiment of the present application, the weights and biases of the network will be updated through the training process of the back-propagation algorithm to achieve the purpose of minimizing the loss function. In order to enable the fully connected feedforward neural network to learn the complex mapping relationship between input parameters and output parameters, Tanh is selected as the activation function. Its continuity and smoothness characteristics help to accurately simulate the behavior of the physical field while ensuring the numerical stability of the model during the training process. Sin and other functions can also be used as activation functions. The total loss function of PINN consists of three parts: data-driven loss function, physical information loss function, and initial condition loss function. Then the above step 103 can be replaced by the following steps 301 to 305.

[0064] Step 301: Establish a data-driven loss function: Where, is the data-driven loss, is the mean square error, is the sound pressure data of the target biological tissue in the historical time and space area, This is the sound pressure data output by the physical information neural network.

[0065] Step 302: Convert the photoacoustic wave equation into a homogeneous wave equation: .

[0066] According to the Duhamel principle, the inhomogeneous photoacoustic wave equation is equivalent to the homogeneous wave equation, which can be expressed as follows with the addition of initial conditions:

[0067] ;

[0068] is the time parameter, which represents the laser pulse heating time; It represents the sound pressure obtained by the photoacoustic effect produced by pulsed laser excitation, that is, the initial sound pressure generated after laser heating.

[0069] Sound pressure obtained by photoacoustic effect generated by pulsed laser excitation It can be expressed by the following formula:

[0070] ;

[0071] in, represents the initial distribution of light absorption density. represents the Gruneisen constant.

[0072] Since the time parameter of pulse laser excitation It is generally at the nanosecond level, so it can be ignored for solving the photoacoustic wave equation; the photoacoustic wave equation to be solved is converted to the following formula:

[0073] .

[0074] Embed this equation into the physical information neural network and solve the objective function: .

[0075] Step 303: Based on the homogeneous wave equation, establish the physical information loss function as follows: Where, It is the loss of physical information.

[0076] Step 304: Based on the initial conditions satisfied by the photoacoustic wave equation, an initial condition loss function is established: Where, is the initial condition loss.

[0077] Step 305: According to the data-driven loss function, the physical information loss function and the initial condition loss function, the total loss function is obtained as follows: Where, is the total loss function, is the weight of physical information loss, is the weight of the data-driven loss, is the weight of the initial condition loss. 、 and Used to balance the impact of the three loss functions on network training to ensure that they converge effectively during training.

[0078] In another exemplary embodiment of the present application, to improve computational efficiency and ensure the representativeness of training data, when training a physical information neural network using a data sample set, Latin Hypercube Sampling (LHS) is used to extract multiple samples from the entire spatiotemporal domain for training every preset number of rounds. By ensuring uniform sampling in each dimension, LHS helps maximize sample coverage and reduce sampling bias. This sampling training method not only improves data utilization efficiency but also accelerates the network training process.

[0079] For example, PINN is set to extract 20,000 points from the entire spatiotemporal domain for training every 1,000 epochs. Figure 2 , the ultrasonic transducer collects the ultrasonic signal to obtain the sound pressure data, and the position of the ultrasonic transducer and the acquisition time are combined into the space-time coordinates , and serves as the first batch input of the physical information neural network, and the physical information neural network outputs , Input data drives the loss The data points in the second batch of input of the physical information neural network are collected from the entire spatiotemporal domain ( ) 20,000 points are sampled for training. The Latin hypercube sampling method is used in the sampling process. The second batch of physical information neural network output , input physical information loss function. The initial condition loss function and physical information loss function use It is not the same, it is not the sound pressure at the same location. It can be seen that the input of the data sample set is the spatiotemporal location of the collected data, while the input of the physical information loss function is the spatiotemporal coordinates collected in the entire spatiotemporal domain, so that the sound pressure at any location can be obtained.

[0080] In order to terminate the training when the total loss function reaches convergence, the total number of iterations is defined Equal to 5×10 4 , learning rate Equal to 10 -4 , = =0.1, =1. When the data noise is large, = = =1, which has a better filtering effect on the data. PINN uses Adam optimizer to update parameters to minimize the total loss function. or When , the calculation stops. is the initial total loss ratio. The space-time coordinates at the initial moment are input into a trained physical information neural network, and the network output is the sound pressure data at the initial moment. The method of this application cleverly exploits the similarities between the backpropagation characteristics of neural networks and the chain rule, effectively utilizing the properties of automatic differentiation to accurately solve the partial differential equation (the photoacoustic wave equation).

[0081] In another exemplary embodiment of the present application, when PINN is trained, the weights and biases of the network are randomly initialized, which may be a time-consuming process. In order to better apply the method of the present application to clinical practice and speed up network training, transfer learning is used to accelerate calculations. PINN is pre-trained on existing similar data. The pre-training process is the same as the formal training process, so that the network is forced to meet the solution process of the photoacoustic wave equation. The data sample set of the present application is then input into the pre-trained model for training, so that the gradient descent direction can be found as quickly as possible after the data sample set is input into the pre-trained model. After testing, it was found that the pre-trained PINN can reduce the training time by more than 80% without reducing the imaging quality, greatly speeding up its inversion solution speed, and can meet clinical usage needs to the greatest extent.

[0082] Compared with traditional photoacoustic imaging methods, PINN can extract more spatiotemporal features of photoacoustic waves, and in limited viewing angles and sparse sampling, it avoids the estimation errors caused by incomplete projection data to the greatest extent. In addition, because PINN combines physical information equations, it has a good filtering effect on the data, so it has good noise resistance compared with traditional methods. Compared with traditional deep learning methods, this application encodes the physical information equations as a loss function, so that the network does not learn data features in a disorderly manner, but uses physical laws as a guide to achieve parameter solution. This method of embedding physical information makes the neural network no longer simply rely on the stacking of a large amount of training data, greatly reducing the demand for training data volume, and even achieving accurate solution of single-case data. At the same time, this application can view the solution process of each internal parameter, greatly improving the interpretability of deep learning. Therefore, the method of this application has application advantages far exceeding the existing technology.

[0083] In another exemplary embodiment of the present application, the PINN network structure of the present application can also use a B-spline function. The parameters of the B-spline function can be updated through the training process. The activation function corresponding to each neuron will continue to change with the training process, and the solution of partial differential equations can also be realized.

[0084] In the above step 106, the initial time ( ) Any spacetime coordinate in the spacetime space is input into the trained physical information neural network, which outputs the sound pressure data for any spacetime coordinate in the spacetime space at the initial moment. The sound pressure data for all spacetime coordinates in the spacetime space constitute the initial sound pressure distribution. Since photoacoustic imaging image reconstruction uses a mathematical algorithm to invert the initial sound pressure distribution from the detected ultrasound signal, once the initial sound pressure distribution is obtained in step 106, photoacoustic image reconstruction is achieved.

[0085] Based on the same inventive concept, embodiments of the present application also provide a photoacoustic image reconstruction device for implementing the aforementioned photoacoustic image reconstruction method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations in one or more of the following photoacoustic image reconstruction device embodiments can be found in the aforementioned limitations on the photoacoustic image reconstruction method and will not be further elaborated here.

[0086] In an exemplary embodiment, Figure 3 As shown, a photoacoustic image reconstruction device is provided, which includes: a data acquisition module, an equation determination module, a loss function establishment module, a sample set construction module, a training module and an application module.

[0087] A data acquisition module is used to obtain the time-varying sound pressure data of the target biological tissue at multiple preset positions. An equation determination module is used to determine the photoacoustic wave equation. A loss function establishment module is used to embed the photoacoustic wave equation into the loss function of the physical information neural network to obtain a total loss function. A sample set construction module is used to construct a data sample set using the space-time coordinates consisting of preset positions and times as inputs and the time-varying sound pressure data as labels. A training module is used to train the physical information neural network using the data sample set based on the total loss function to obtain a trained physical information neural network. An application module is used to input the space-time coordinates of the initial moment into the trained physical information neural network and output the sound pressure data at the initial moment, thereby obtaining the initial sound pressure distribution.

[0088] 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 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. 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 computer program in the non-volatile storage medium. The database of the computer device is used to store initial sound pressure distribution. 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, a photoacoustic image reconstruction method is implemented.

[0089] Those skilled in the art will understand that Figure 4 The structure shown in the figure is merely a block diagram of a portion 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. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0090] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0091] 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.

[0092] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. 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 above-mentioned embodiments. In particular, any reference to memory, database, or other media 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), magnetic 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).

[0093] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0094] The technical features of the above embodiments can 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.

[0095] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may 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. A photoacoustic image reconstruction method, characterized in that: include: Acquiring time-varying acoustic pressure data of target biological tissue at a plurality of preset positions; Determine the photoacoustic wave equation; The photoacoustic wave equation is embedded into the loss function of the physical information neural network to obtain the total loss function; The data sample set is constructed by taking the space-time coordinates consisting of the preset position and time as input and the sound pressure data that changes with time as the label; Based on the total loss function, the physical information neural network is trained using the data sample set to obtain a trained physical information neural network; Input the space-time coordinates of the initial moment into the trained physical information neural network, output the sound pressure data at the initial moment, and thus obtain the initial sound pressure distribution; Determine the photoacoustic wave equation, including: The three coupled acoustic equations are established as 、 and Where, is the spatial position, and Represent the horizontal and vertical coordinates of the spatial position, For in time Particles are located in space The speed at is the particle mass density, is the gradient operator; For spatial location The particle mass density at For spatial location The speed of sound at ; is the coefficient of thermal expansion, For in time Located in space Sound pressure data at is the specific heat capacity at constant pressure, In time Particles are located in space The temperature at Represents the heat energy deposited per unit volume and per unit time; According to the three coupled acoustic equations, the photoacoustic wave equation is determined as: Where, is the Laplace operator; The photoacoustic wave equation is embedded into the loss function of the physical information neural network to obtain the total loss function, which specifically includes: The data-driven loss function is established as: Where, is the data-driven loss, is the mean square error, is the sound pressure data of the target biological tissue in the historical time and space area, The sound pressure data output by the physical information neural network; Convert the photoacoustic wave equation to a homogeneous wave equation: ; According to the homogeneous wave equation, the physical information loss function is established as: Where, For physical information loss; According to the initial conditions satisfied by the photoacoustic wave equation, the initial condition loss function is established as: Where, is the initial condition loss; is located at the spatial position at the initial moment Sound pressure data at According to the data-driven loss function, physical information loss function and initial condition loss function, the total loss function is obtained as follows: Where, is the total loss function, is the weight of physical information loss, is the weight of the data-driven loss, is the weight of the initial condition loss.

2. The photoacoustic image reconstruction method according to claim 1, wherein: The initial conditions satisfied by the photoacoustic wave equation are: ； ； Where, The spatial position obtained by the photoacoustic effect produced by pulsed laser excitation The sound pressure data at .

3. The photoacoustic image reconstruction method according to claim 1, wherein: The physical information neural network includes: an input layer, an output layer and multiple hidden layers; The input layer includes 3 input parameters; Each hidden layer consists of 100 neurons; The output layer includes 1 output parameter.

4. The photoacoustic image reconstruction method according to claim 1, wherein: When using data sample sets to train physical information neural networks: Every preset number of rounds, multiple samples are extracted from the entire spatiotemporal domain using the Latin hypercube sampling method for training.

5. The photoacoustic image reconstruction method according to claim 1, wherein: The physical information neural network is a pre-trained physical information neural network.

6. A photoacoustic image reconstruction device, characterized in that: The photoacoustic image reconstruction device is used to implement the photoacoustic image reconstruction method according to any one of claims 1 to 5, and the photoacoustic image reconstruction device includes: A data acquisition module is used to acquire time-varying acoustic pressure data of target biological tissue at multiple preset positions; an equation determination module for determining the photoacoustic wave equation; A loss function building module is used to embed the photoacoustic wave equation into the loss function of the physical information neural network to obtain the total loss function; A sample set construction module is used to construct a data sample set using the space-time coordinates consisting of a preset position and time as input and the time-varying sound pressure data as a label; A training module, configured to train the physical information neural network using a data sample set based on the total loss function to obtain a trained physical information neural network; The application module is used to input the space-time coordinates of the initial moment into the trained physical information neural network, output the sound pressure data of the initial moment, and thus obtain the initial sound pressure distribution.

7. 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 photoacoustic image reconstruction method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the photoacoustic image reconstruction method according to any one of claims 1 to 5 is implemented.

Images