Ultrasonic tomography method based on virtual data reconstruction
By using virtual data reconstruction methods, utilizing data acquisition and field-wave separation from a real ultrasonic ring transducer, and combining multi-scale inversion and gradient optimization algorithms, the shortcomings of the full waveform inversion method in imaging speed are solved, and efficient and accurate ultrasonic tomography is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INST OF ACOUSTICS CHINESE ACAD OF SCI
- Filing Date
- 2026-02-28
- Publication Date
- 2026-05-12
AI Technical Summary
Existing full-waveform inversion methods are insufficient in imaging speed and cannot achieve rapid imaging that meets the real-time requirements of clinical practice. In particular, when dealing with ring transducer arrays composed of a large number of array elements, the computational resource requirements are too large, resulting in low imaging efficiency.
Full-matrix data acquisition was performed using a real ultrasonic ring transducer under both loaded and unloaded conditions. Frequency-domain scattered wave data was obtained using field-wave separation and discrete Fourier transform. A virtual ultrasonic ring transducer was constructed for wave field extrapolation. The sound velocity model was optimized by combining multi-scale inversion and gradient optimization algorithms, and finally, ultrasonic tomography was generated.
While reducing computational complexity, it improves imaging efficiency and accuracy, achieving imaging precision close to that of traditional full waveform inversion, thus meeting the needs of real-time clinical imaging.
Smart Images
Figure CN122004941A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of medical ultrasound technology, and in particular to an ultrasound tomography method based on virtual data reconstruction. Background Technology
[0002] Ultrasound computed tomography (USCT) is a non-invasive medical imaging technique that offers a reliable way to obtain structural information and acoustic parameter distribution of patient tissues in clinical practice, thanks to its significant advantages of no ionizing radiation and moderate cost. This technique uses an array of ultrasound transducers to acquire sound wave data from multiple angles, and then utilizes relevant signal processing algorithms to reconstruct internal structures and parameter distributions, demonstrating good diagnostic potential for early tissue pathological changes.
[0003] Currently, ultrasound tomography methods can be mainly divided into two categories: one is imaging algorithms based on the assumption of ray propagation, including delay and sum (DAS) and time-of-flight tomography (TOFT). These two methods, by approximating the sound wave propagation path, have the advantages of fast reconstruction speed and ease of implementation, making them suitable for rapid imaging. The other category is the full waveform inversion (FWI) method based on the wave equation and wave propagation theory. This method, by fitting the propagation waveform, can obtain imaging results with higher resolution and quantitative characteristics, showing significant potential in distinguishing subtle structural changes and providing more accurate acoustic information for early lesion identification. However, due to the complex iterative solution of partial differential equations and the large amount of computational resources involved, full waveform inversion currently has shortcomings in imaging speed and has not yet reached a level of widespread adoption in practical clinical applications, remaining in the stage of continuous optimization and validation.
[0004] To improve imaging efficiency, full waveform inversion can be performed in the frequency domain, a method commonly referred to as Frequency-Domain Full Waveform Inversion (FD-FWI). A multi-scale inversion strategy is employed, starting the inversion process with low-frequency data and gradually introducing higher-frequency components. This reduces computational burden while ensuring solution stability, making image reconstruction possible on personal computers. Although the multi-scale strategy significantly reduces the algorithm's computational resource requirements, the sheer number of independent acoustic wave propagation models requiring numerical solutions is still a significant challenge when dealing with ring transducer arrays composed of large-radius elements. This makes it impossible for this method to achieve the rapid imaging performance required for clinical real-time processing under current hardware conditions. Summary of the Invention
[0005] In a first aspect, this disclosure provides an ultrasound tomography method based on virtual data reconstruction, comprising:
[0006] The real ultrasonic ring transducer performs full-matrix data acquisition at discrete sampling time points under both loaded and unloaded conditions to obtain real total field observation data and real background field observation data. The loaded condition refers to the situation where the target object is placed inside the real ultrasonic ring transducer. The full-matrix data acquisition means that each element of the real ultrasonic ring transducer emits sound wave signals in sequence, and the other elements receive the sound wave signals.
[0007] Field-wave separation is performed using real total field observation data and real background field observation data to obtain real scattered wave observation data, and then frequency domain real scattered wave observation data is obtained through discrete Fourier transform.
[0008] Based on the center frequency of the emitted acoustic signal during full matrix data acquisition, N is selected. f There are N processing frequency points; among them, N f N is a positive integer greater than 1. f The processing frequency points are distributed on both sides of the center frequency.
[0009] A virtual ultrasonic ring transducer is constructed, and wave field extrapolation is performed using the frequency domain real scattered wave observation data of each processing frequency point to achieve virtual data reconstruction and obtain the frequency domain virtual scattered wave observation data of each processing frequency point. The geometric center of the virtual ultrasonic ring transducer and the real ultrasonic ring transducer are consistent. The virtual ring ultrasonic transducer can completely enclose the target body under test and its radius is smaller than that of the real ring transducer.
[0010] An initial sound velocity model is set up, and multi-scale inversion is performed using frequency domain virtual scattered wave observation data and / or frequency domain real scattered wave observation data corresponding to each processing frequency point to iteratively optimize the sound velocity model.
[0011] Ultrasonic tomography is generated using an iteratively optimized sound velocity model.
[0012] For the highest processing frequency, conventional frequency domain full waveform inversion is performed using the corresponding frequency domain real scattered wave observation data.
[0013] Secondly, this disclosure provides a method for virtual data reconstruction through wave field interpolation, including:
[0014] An angular transformation is performed on the frequency domain real scattered wave observation data to obtain the real scattered wave angular spectrum.
[0015] By using the real scattered wave angular spectrum for single-step shifting, a first virtual scattered wave angular spectrum is obtained, which characterizes the acoustic wave signal emitted by the real ultrasonic ring transducer and the acoustic wave signal received by the virtual ultrasonic ring transducer; wherein, the single-step shifting is based on the analytical propagation theory of cylindrical wave angular spectrum.
[0016] Based on the reciprocity of the cylindrical wave angular spectrum analytical propagation theory, a second virtual scattered wave angular spectrum is obtained, which characterizes the acoustic wave signal emitted by the virtual ultrasonic ring transducer and the acoustic wave signal received by the real ultrasonic ring transducer.
[0017] By using the second virtual scattered wave angular spectrum for single-step shifting, a third virtual scattered wave angular spectrum characterizing the acoustic wave signals emitted and received by the virtual ultrasonic ring transducer is obtained.
[0018] The angular spectrum of the third virtual scattered wave is scanned and transformed to obtain the frequency domain virtual scattered wave observation data.
[0019] Thirdly, this disclosure provides a method for optimizing single-step data shifting in wavefield interpolation, including:
[0020] Based on the reversibility of the cylindrical wave angular spectrum analytical propagation theory, the inverse relationship of single-step transfer is established.
[0021] Based on the inverse relationship, the loss function for single-step migration is constructed using the L2 norm.
[0022] Based on the loss function, the data obtained from the single-step transfer is optimized using the first local gradient optimization algorithm.
[0023] Fourthly, this disclosure provides a virtual data inversion method, including:
[0024] The Helmholtz equations for the frequency domain simulated total field observation data and the frequency domain simulated background field observation data are constructed using the second-order constant density acoustic wave equation. The Helmholtz equations are then solved based on the convergent Bern series method to obtain the frequency domain simulated scattering field observation data, thus realizing forward modeling.
[0025] The objective function for virtual data inversion is constructed using the L2 norm of frequency domain virtual scattered wave observation data and frequency domain simulated scattered field observation data.
[0026] Based on the objective function derived from the virtual data, the sound speed model is optimized using the second local gradient optimization algorithm.
[0027] The content described in this section is not intended to identify key or important features of the embodiments of this disclosure, nor does it constitute a limitation on the scope of this disclosure.
[0028] Other features of this disclosure will be described in detail in the following description to aid understanding. Attached Figure Description
[0029] Figure 1 This is a schematic diagram of an ultrasound tomography method based on virtual data reconstruction provided in an embodiment of this disclosure;
[0030] Figure 2 This is a schematic diagram showing the positional relationship between a real ring transducer, a virtual ring transducer, and the target object under test in one embodiment of this disclosure;
[0031] Figure 3 This is a schematic flowchart of a method for virtual data reconstruction by wave field interpolation according to an embodiment of the present disclosure;
[0032] Figure 4 This is a schematic flowchart of a method for optimizing single-step data transfer in wavefield extrapolation according to an embodiment of this disclosure;
[0033] Figure 5 This is a schematic flowchart of a virtual data inversion method provided in one embodiment of the present disclosure;
[0034] Figure 6 This is a schematic diagram of a real annular transducer and its internal pork tissue in one embodiment of this disclosure;
[0035] Figure 7 This is a comparison chart of virtual data and real data at a processing frequency of 0.25MHz in one embodiment of this disclosure;
[0036] Figure 8 This is a comparison chart of the real tissue sound velocity model, the traditional FWI sound velocity model, and the multi-scale inversion sound velocity model disclosed in this paper. Detailed Implementation
[0037] The present disclosure will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments provided below are merely exemplary. Furthermore, for the sake of brevity and clarity, common knowledge has been omitted from the description of the embodiments below.
[0038] In this document, terms such as "first," "second," and "third" are used only to distinguish identical or similar descriptive objects and are not intended to limit the specific order or sequence of the described objects, nor are they used to limit the importance of the described objects. At the same time, in order to enable those skilled in the art to clearly understand the technical solutions provided in this disclosure, expressions such as "red," "yellow," and "blue," as well as the specific colors in the accompanying drawings, are also used only to distinguish identical or similar descriptive objects and do not represent the color attributes possessed by the descriptive objects when this solution is actually deployed.
[0039] One embodiment of this disclosure provides an ultrasound tomography method based on virtual data reconstruction, such as... Figure 1 As shown, it specifically includes:
[0040] Step S101: The real ultrasonic ring transducer performs full matrix data acquisition at discrete sampling time points under both load and no-load conditions to obtain real total field observation data and real background field observation data.
[0041] The load condition refers to the situation where the target object is placed in the actual ring transducer, while the no-load condition refers to the situation where there is no target object in the actual transducer.
[0042] A real transducer consists of multiple array elements. Full data acquisition means that each array element emits acoustic signals in sequence, and all other array elements receive the acoustic signals.
[0043] In this embodiment of the disclosure,
[0044] This represents the actual total field observation data;
[0045] This represents the observational data of the real background field.
[0046] in,
[0047] Indicates the coordinates of the signal transmission point;
[0048] Indicates the location coordinates of the signal receiving point;
[0049] This represents the index of the signal sequence in the time domain, in seconds.
[0050] The aforementioned coordinates of the transmitting and receiving points refer to the positions of the sound wave signals as mapped from the physical entities of the transmitting and receiving array elements in the physical world.
[0051] Step S102: Perform field-wave separation using real total field observation data and real background field observation data to obtain real scattered wave observation data, and obtain frequency domain real scattered wave observation data through discrete Fourier transform.
[0052] The field wave separation operation is shown in Equation 1.
[0053] Formula 1:
[0054]
[0055] in,
[0056] This represents actual scattered wave observation data;
[0057] Frequency domain scattered wave sound pressure observation data express, This represents the signal sequence index in the frequency domain, in Hertz.
[0058] The data volume of the three types of sound pressure observation data is 100%. , Represents the number of discrete time samples, The actual number of receiving array elements and The actual number of launch array elements.
[0059] Step S103: Based on the center frequency of the emitted acoustic signal during full matrix data acquisition, select N. f One processing frequency point.
[0060] Where, N f It is a positive integer greater than 1.
[0061] The processing frequency should be selected on both sides of the center frequency in order to comprehensively utilize the effective information of the observation data and improve the accuracy of ultrasound tomography.
[0062] In a preferred embodiment of this disclosure, N f The frequency difference between each processing frequency point is consistent, which can express.
[0063] Step S104: Construct a virtual ultrasonic ring transducer, and use the frequency domain real scattered wave observation data of each processing frequency point to perform wave field extrapolation in order to realize virtual data reconstruction and obtain the frequency domain virtual scattered wave observation data of each processing frequency point.
[0064] In this disclosed technical solution, the radius of a real ultrasonic ring transducer is known. Based on this, the virtual ultrasonic ring transducer should meet the following requirements:
[0065] 1. The virtual ultrasonic ring transducer maintains the same geometric center as the real ultrasonic ring transducer.
[0066] 2. Radius of the virtual ultrasonic ring transducer It should be smaller than Furthermore, its array of elements can completely enclose the target body under test.
[0067] Figure 2 The diagram illustrates the positional relationship between a real ultrasonic ring transducer, a virtual ultrasonic ring transducer, and the target body under test in one embodiment of this disclosure. The outermost layer is the real ultrasonic ring transducer, and the center position is the target body under test.
[0068] Based on the actual frequency-domain scattered wave observation data of the real ring transducer, the frequency-domain virtual scattered wave observation data of the virtual ring transducer is reconstructed, which has a smaller radius and therefore a smaller computational domain. This data is then used for subsequent multi-scale inversion, which can effectively reduce the computational cost of solving the forward wave field during the calculation process.
[0069] Step S105: Set the initial sound velocity model, and use the frequency domain virtual scattered wave observation data and / or frequency domain real scattered wave observation data corresponding to each processing frequency point to perform multi-scale inversion and iteratively optimize the sound velocity model.
[0070] The purpose of setting the initial sound velocity model is to provide a feasible starting point for multi-scale inversion. It can be set based on prior knowledge of the target body or similar target bodies, using methods such as uniform models, hierarchical models, or pre-estimation.
[0071] Multi-scale inversion is a hierarchical optimization framework. In the technical solution disclosed herein, it specifically refers to performing inversion sequentially from low to high processing frequency points, with the sound velocity model obtained from the inversion at the current processing frequency point serving as the starting point for the inversion at the next processing frequency point.
[0072] In one embodiment of this disclosure, virtual data inversion is performed using the corresponding frequency domain virtual scattering wave observation data for all processing frequency points.
[0073] In a preferred embodiment of this disclosure, for the lowest processing frequency ( ) to the second highest processing frequency ( Virtual data inversion is performed using the corresponding frequency domain virtual scattered wave observation data; for the highest processing frequency point ( The corresponding frequency domain real scattered wave observation data is used for conventional FD-FWI. The beneficial effect of this preferred embodiment is that by introducing real data to constrain the possible error or noise amplification of virtual data in the high-frequency band, the accuracy and robustness of ultrasound tomography can be improved.
[0074] Step S106: Generate ultrasonic tomographic images using the iteratively optimized sound velocity model.
[0075] After obtaining the iteratively optimized sound velocity model, ultrasonic tomography can be generated by combining acoustic parameters such as the sound attenuation coefficient distribution of the target body.
[0076] Acoustic parameters such as the sound attenuation coefficient distribution can also be set based on prior knowledge.
[0077] One embodiment of this disclosure provides a method for virtual data reconstruction through wavefield interpolation, such as... Figure 3 As shown, it specifically includes:
[0078] Step S301: Perform angular transformation on the frequency domain real scattered wave observation data to obtain the real scattered wave angular spectrum.
[0079] This step specifically refers to... Transformed into the true scattered wave angular spectrum .
[0080] in,
[0081] This represents the angle of the Cartesian coordinate transformation of the receiving point;
[0082] The angle represents the Cartesian coordinate transformation of the launch point.
[0083] The subscript r1-r1 indicates that both the transmitting and receiving array elements of the acoustic signal are located in a real ultrasonic ring transducer with a radius of r1.
[0084] Step S302: Use the real scattered wave angular spectrum to perform single-step shifting to obtain the first virtual scattered wave angular spectrum characterizing the acoustic wave signal emitted by the real ultrasonic ring transducer and the acoustic wave signal received by the virtual ultrasonic ring transducer.
[0085] This single-step transfer step aims to be based on the analytical propagation theory of cylindrical wave angular spectrum, by... Reconstruct the virtual first virtual scattered wave angular spectrum The subscripts r1-r2 indicate that the transmitting array elements of the acoustic signal are located in a real ultrasonic ring transducer with a radius of r1, and the receiving array elements are all located in a virtual ultrasonic ring transducer with a radius of r2.
[0086] The expression for this single-step transfer is shown in Formula 2.
[0087] Formula 2:
[0088]
[0089] in,
[0090] This indicates the selected angular mode order, which physically represents the spatial frequency of the cylindrical wave angular harmonics.
[0091] k represents the wave number;
[0092] This represents the first-order m-th order Hankel function, whose physical meaning is the radial propagation characteristic of the m-th order angular mode.
[0093] The analytic propagation operator is defined by its physical meaning, which characterizes the amplitude attenuation and phase delay of the m-order angular mode propagating from a real ultrasonic ring transducer with radius r1 to a virtual ultrasonic ring transducer with radius r2.
[0094] Step S303: Based on the reciprocity of the cylindrical wave angular spectrum analytical propagation theory, the second virtual scattered wave angular spectrum characterizing the acoustic wave signal emitted by the virtual ultrasonic ring transducer and the acoustic wave signal received by the real ultrasonic ring transducer is obtained.
[0095] The reciprocity expression of the analytical propagation theory of cylindrical wave angular spectrum is shown in Equation 3.
[0096] Formula 3:
[0097]
[0098] in,
[0099] The angle representing the Cartesian coordinate transformation of the virtual receiving point;
[0100] The angle represents the Cartesian coordinate transformation of the virtual launch point.
[0101] The subscripts r2-r1 indicate that the transmitting array element of the acoustic signal is located in a virtual ultrasonic ring transducer with a radius of r2, and the receiving array element is located in a real ultrasonic ring transducer with a radius of r1.
[0102] Step S304: Use the second virtual scattered wave angular spectrum to perform single-step shifting to obtain the third virtual scattered wave angular spectrum characterizing the acoustic wave signals emitted and received by the virtual ultrasonic ring transducer.
[0103] Corresponding to step S303, the expression for the single-step transfer in this step is shown in Formula 4.
[0104] Formula 4 (corresponding to Formula 2):
[0105]
[0106] Step S305: Perform a scanning transformation on the angular spectrum of the third virtual scattered wave to obtain the frequency domain virtual scattered wave observation data.
[0107] This step aims to obtain the final frequency domain virtual scattering observation data using the third virtual scattering wave angular spectrum. The amount of data is , This indicates the number of discrete processing frequency points. Indicates the number of virtual receiver array elements. This indicates the number of virtual transmitter array elements.
[0108] It should be understood that steps S301-S305 are execution steps for a single processing frequency point. In actual deployment of the technical solution disclosed herein, steps S301-S305 are executed for all processing frequency points.
[0109] Since single-step transfer operations may introduce errors sufficient to affect the accuracy of ultrasound tomography results, this disclosure provides an embodiment of a method for optimizing single-step transfer data in wavefield interpolation, such as... Figure 4 As shown, it specifically includes:
[0110] Step S401: Based on the reversibility of the cylindrical wave angular spectrum analytical propagation theory, establish the reverse relationship of single-step transfer.
[0111] For the single-step transfer in step S302, the reverse relationship expression is shown in Formula 5.
[0112] Formula 5:
[0113]
[0114] This can be understood as performing inverse virtual data reconstruction on the first virtual scattered wave angular spectrum to obtain an estimate of the real scattered wave angular spectrum.
[0115] Step S402: Based on the inverse relationship, construct the loss function for single-step migration using the L2 norm.
[0116] For the single-step transfer in step S302, the loss function is shown in Formula 6.
[0117] Formula 6:
[0118]
[0119] This can be understood as quantifying the difference between the actual scattered wave sound pressure angle spectrum and the estimated value.
[0120] Step S403: Based on the loss function, optimize the data obtained from the single-step transfer using the first local gradient optimization algorithm.
[0121] For the single-step transfer in step S302, the first local gradient optimization algorithm is shown in Equation 7.
[0122] Formula 7:
[0123]
[0124] Where i represents the number of iterations, the local gradient optimization algorithm is an iterative optimization process, and its termination condition is that the number of iterations reaches the preset number.
[0125] In one embodiment of this disclosure, a line search method is used to obtain the step size. The steepest descent method is used to determine the descent direction. In the steepest descent method, the descent direction is equal to the negative gradient, and the gradient calculation formula is shown in Formula 8.
[0126] Formula 8:
[0127]
[0128] For the single-step transfer data in step S302, its optimized first virtual scattering angular spectrum can be express.
[0129] It should be understood that, in the actual deployment of the technical solution disclosed herein, steps S401-S403 can be embedded as a whole between steps S302 to S303 and steps S304 to S305 to optimize the first and third virtual scattering wave angular spectrum.
[0130] For the single-step transfer in step S304, the formulas corresponding to Formulas 5 to 8 are given here, and the contents will not be repeated here.
[0131] Formula 9 (corresponding to Formula 5):
[0132]
[0133] Formula 10 (corresponding to Formula 6):
[0134]
[0135] Formula 11 (corresponding to Formula 7):
[0136]
[0137] Formula 12 (corresponding to Formula 8):
[0138]
[0139] One embodiment of this disclosure provides a method for optimizing single-step data transfer in wavefield extrapolation, which can effectively reduce the error of the data obtained by single-step transfer and improve the accuracy of ultrasound tomography.
[0140] One embodiment of this disclosure provides a virtual data inversion method, such as... Figure 5 As shown, it specifically includes:
[0141] Step S501: Construct the Helmholtz equations for the frequency domain simulated total field observation data and the frequency domain simulated background field observation data using the second-order constant density acoustic wave equation, and solve them based on the convergent Bern series method to obtain the frequency domain simulated scattering field observation data, thus realizing forward modeling.
[0142] The Helmholtz equations for the frequency domain simulated total field observation data and the frequency domain simulated background field observation data are shown in Equations 13 and 14.
[0143] Formula 13:
[0144]
[0145] Formula 14:
[0146]
[0147] in,
[0148] This represents the frequency domain simulation of total field observation data;
[0149] This represents the frequency domain simulation background field observation data;
[0150] The source term of the Helmholtz equation in the framework of the Convergent Born Series (CBS) is specifically the frequency domain excitation function of the sound source in this disclosed technical solution.
[0151] In one embodiment of this disclosure, field wave separation and parameter separation are first performed according to Equations 13 and 14 to obtain the differential form of the Lippmann-Schwinger (LS) equation for the frequency domain simulated scattering field data, as shown in Equation 15. This equation is the equivalent form of the non-homogeneous Helmholtz equation for the scattering problem.
[0152] Formula 15:
[0153]
[0154] in,
[0155] This represents the frequency domain simulation data of the scattering field observations;
[0156] k represents the total field wavenumber;
[0157] k0 represents the background field. The wave number.
[0158] The field wave separation and parameter separation in this embodiment are shown in Equations 16 and 17.
[0159] Formula 16:
[0160]
[0161] Formula 17:
[0162]
[0163] Subsequently, the CBS method was used to solve Equation 15 to obtain the frequency domain simulated scattering field observation data.
[0164] In another embodiment of this disclosure, the CBS method is first used to solve Equations 13 and 14 to obtain frequency domain simulated total field observation data and frequency domain simulated background field observation data.
[0165] Subsequently, field-wave separation was performed using frequency-domain simulated total field observation data and frequency-domain simulated background field observation data (as shown in Equation 16) to obtain frequency-domain simulated scattering field observation data.
[0166] Step S502: Construct the objective function for virtual data inversion using the L2 norm of frequency domain virtual scattered wave observation data and frequency domain simulated scattered field observation data.
[0167] The objective function for virtual data inversion is shown in Equation 18.
[0168] Formula 18:
[0169]
[0170] Step S503: Based on the objective function derived from the virtual data, optimize the sound speed model using the second local gradient optimization algorithm.
[0171] The second local gradient optimization algorithm is shown in Equation 19.
[0172] Formula 19:
[0173]
[0174] In one embodiment of this disclosure, for the second local gradient optimization algorithm, the adjoint state method is first used to obtain the current gradient, and the gradient calculation formula is shown in Formula 20.
[0175] Formula 20:
[0176]
[0177] in,
[0178] This represents the propagating wave field in the adjoint state method;
[0179] This represents the residual backpropagating wavefield in the adjoint state method;
[0180] Indicates conjugate;
[0181] This indicates taking the real part.
[0182] Subsequently, the current gradient is optimized using either the conjugate gradient method or the L-BFGS (Limited-memory Broyden-Fletcher-Goldfarb-Shanno) method to obtain a better descent direction. And use the line search method to obtain the step size. .
[0183] It should be understood that steps S501 to S503 are also operations under a single processing frequency point. In the actual deployment of the technical solution disclosed herein, steps S501 to S503 are performed for all processing frequency points.
[0184] In one embodiment of this disclosure, the technical solution of this disclosure is applied to computer simulation of pork tissue inside a real annular transducer in a water immersion environment.
[0185] like Figure 6 As shown, the actual ring transducer is a 512-element ring array with a diameter of 22cm. Data is collected by surrounding the transducer with block-shaped components arranged in an approximately hollow cylindrical shape. The data volume is... The center frequency of the transmitted signal is 0.6MHz, and the frequency points from 0.25MHz to 1.2MHz, with an interval of 0.05MHz, are selected as the processing frequency points.
[0186] The radius of the virtual ring transducer is set to 7cm. Virtual data reconstruction is achieved through wave field extrapolation. The virtual data and real data at the 0.25MHz processing frequency are compared, for example... Figure 7 As shown.
[0187] For the processing frequency points of 0.25MHz-1.15MHz, virtual data inversion is performed using the corresponding frequency domain virtual scattered wave observation data. For the processing frequency point of 1.2MHz, conventional FD-FWI is performed using the corresponding frequency domain real scattered wave observation data.
[0188] A comparison of the actual tissue sound velocity, the sound velocity obtained by conventional FWI, and the sound velocity model after multi-scale inversion in the embodiments of this disclosure. Figure 8 As shown.
[0189] The disclosed technical solution can effectively reduce the computational cost of wavefield solution during the operation by reducing the wavefield computation domain, thereby reducing computational complexity and improving the efficiency of ultrasound tomography, while maintaining near-traditional FWI inversion accuracy.
[0190] Any changes made to the above embodiments by those skilled in the art without departing from the true spirit and scope of this disclosure should be included within the scope of protection covered by the claims. The scope of protection claimed by this invention is limited only by the claims.
Claims
1. A method for ultrasound tomography based on virtual data reconstruction, characterized in that, include: The real ultrasonic ring transducer performs full-matrix data acquisition at discrete sampling time points under both loaded and unloaded conditions to obtain real total field observation data and real background field observation data. The loaded condition refers to the situation where the target object is placed inside the real ultrasonic ring transducer. The full-matrix data acquisition means that each element of the real ultrasonic ring transducer emits sound wave signals in sequence, and the other elements receive the sound wave signals. Field-wave separation is performed using the real total field observation data and the real background field observation data to obtain real scattered wave observation data, and then frequency domain real scattered wave observation data is obtained through discrete Fourier transform. Based on the center frequency of the emitted acoustic signal during the full matrix data acquisition, N is selected. f There are N processing frequency points; among them, N f N is a positive integer greater than 1; f The processing frequency points are distributed on both sides of the center frequency; A virtual ultrasonic ring transducer is constructed, and wavefield extrapolation is performed using the frequency domain real scattered wave observation data of each processing frequency point to achieve virtual data reconstruction and obtain the frequency domain virtual scattered wave observation data of each processing frequency point; wherein, the geometric center of the virtual ultrasonic ring transducer and the real ultrasonic ring transducer are consistent; the virtual ring ultrasonic transducer can completely enclose the target body under test and its radius is smaller than the radius of the real ring transducer. An initial sound velocity model is set, and multi-scale inversion is performed using frequency domain virtual scattered wave observation data and / or frequency domain real scattered wave observation data corresponding to each of the processing frequency points to iteratively optimize the sound velocity model. Ultrasonic tomography is generated using an iteratively optimized sound velocity model.
2. The method according to claim 1, characterized in that, include: For the N f For each processing frequency point, virtual data inversion is performed using its corresponding frequency domain virtual scattered wave observation data.
3. The method according to claim 1, characterized in that, include: For the lowest to the second highest processing frequencies, virtual data inversion is performed using the corresponding frequency domain virtual scattered wave observation data. For the highest processing frequency, conventional frequency domain full waveform inversion is performed using the corresponding frequency domain real scattered wave observation data.
4. The method according to claim 1, 2 or 3, characterized in that, Wavefield extrapolation is performed using the actual frequency-domain scattered wave observation data at each of the aforementioned processing frequency points to achieve virtual data reconstruction, resulting in virtual frequency-domain scattered wave observation data for each of the aforementioned processing frequency points, including: The frequency domain real scattered wave observation data are transformed by angular direction to obtain the real scattered wave angular spectrum; By using the real scattered wave angular spectrum for single-step shifting, a first virtual scattered wave angular spectrum characterizing the acoustic wave signal emitted by the real ultrasonic ring transducer and the acoustic wave signal received by the virtual ultrasonic ring transducer is obtained; wherein, the single-step shifting is based on the analytical propagation theory of cylindrical wave angular spectrum; Based on the reciprocity of the cylindrical wave angular spectrum analytical propagation theory, a second virtual scattering wave angular spectrum characterizing the acoustic wave signal emitted by the virtual ultrasonic ring transducer and the acoustic wave signal received by the real ultrasonic ring transducer is obtained. By using the second virtual scattered wave angular spectrum to perform single-step shifting, a third virtual scattered wave angular spectrum characterizing the acoustic wave signals emitted and received by the virtual ultrasonic ring transducer is obtained. The frequency domain virtual scattering wave observation data are obtained by scanning and transforming the angular spectrum of the third virtual scattering wave.
5. The method according to claim 4, characterized in that, Also includes: Based on the reversibility of the cylindrical wave angular spectrum analytical propagation theory, the reverse relationship of the single-step transfer is established; Based on the inverse relationship, the loss function for the single-step transfer is constructed using the L2 norm; Based on the loss function, the data obtained from the single-step transfer is optimized using the first local gradient optimization algorithm.
6. The method according to claim 5, characterized in that, The first local gradient optimization algorithm uses a line search method to obtain the step size and a steepest descent method to obtain the descent gradient.
7. The method according to claim 2, 3 or 6, characterized in that, Virtual data inversion includes: The Helmholtz equations for the frequency domain simulated total field observation data and the frequency domain simulated background field observation data are constructed using the second-order constant density acoustic wave equation, and solved based on the convergent Bern series method to obtain the frequency domain simulated scattering field observation data, thus realizing forward modeling. The objective function for virtual data inversion is constructed using the L2 norm of the frequency domain virtual scattered wave observation data and the frequency domain simulated scattered field observation data. Based on the objective function derived from the virtual data, the sound speed model is optimized using the second local gradient optimization algorithm.
8. The method according to claim 7, characterized in that, The frequency domain simulated scattering field observation data obtained by solving the convergent Bern series method includes: Using the Helmholtz equations of the frequency domain simulated total field observation data and the frequency domain simulated background field observation data, field-wave separation and parameter separation are performed to obtain the differential form of the Lippmann-Schwinger equations for the frequency domain simulated scattering field observation data. The Lippmann-Schwinger equation in differential form was solved using the convergent Bern series method to obtain the frequency domain simulated scattering field observation data.
9. The method according to claim 7, characterized in that, The method based on convergent Berne series to obtain frequency domain simulated scattering field observation data also includes: The Helmholtz equations for the frequency domain simulated total field observation data and the frequency domain simulated background field observation data are solved using the convergent Bern series method to obtain the frequency domain simulated total field observation data and the frequency domain simulated background field observation data. Field-wave separation is performed using the frequency-domain simulated total field observation data and the frequency-domain simulated background field observation data to obtain the frequency-domain simulated scattering field data.
10. The method according to claim 7, characterized in that, The second local gradient optimization algorithm uses the adjoint state method to obtain the current gradient, uses the conjugate gradient method or the L-BFGS method to optimize the current gradient, and uses the line search method to obtain the step size.