FPINN-based low-frequency reverberation breast ultrasonic viscoelastic ultrafast imaging method and system

Abstract

The invention belongs to the technical field of medical ultrasonic imaging, and particularly relates to a low-frequency reverberation breast ultrasonic viscoelastic ultrafast imaging method and system based on FPINN. The low-frequency reverberation breast ultrasonic viscoelastic ultrafast imaging method comprises the following steps: acquiring and processing displacement data of a to-be-imaged breast tissue to obtain frequency domain information; obtaining shear wave velocities under different frequencies based on the frequency domain information, and obtaining elasticity, viscosity and flow data of the mammary tissue according to the shear wave velocities under different frequencies; and based on the obtained elasticity, viscosity and flow data of the mammary tissue, performing imaging processing by using the trained FPINN model to obtain an imaging result of the mammary tissue. According to the FPINN model, in the imaging processing process, the input elasticity, viscosity and flow data of the mammary tissue can be quickly converted into imaging results, and the real-time performance and accuracy of imaging can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical ultrasonic imaging, and in particular relates to a low-frequency reverberation breast ultrasonic viscoelastic ultrafast imaging method and system based on FPINN (Fractional Physics-Informed Neural Network). Background Art

[0002] Breast cancer is a disease in which breast cells grow abnormally and uncontrollably to form tumors. It is one of the three most common cancers in the world. To explain further, in women, breast cancer, lung cancer and colorectal cancer account for 51% of all newly diagnosed cancers, of which breast cancer accounts for 32%. Although there is a lot of evidence that early breast cancer is curable in many cases, the incidence and mortality of breast cancer are still increasing, which is most likely due to the insufficient screening of early breast tissue lesions by existing diagnostic and treatment technologies.

[0003] At present, the commonly used breast lesion screening methods in clinical practice mainly include mammography, breast ultrasound, magnetic resonance imaging, breast tissue biopsy, thermal imaging and molecular imaging. Among them, mammography is widely used in large-scale screening due to its low price, but mammography has low sensitivity to dense breast tissue, there is a possibility of missing small tumors, and there is a risk of low-dose radiation. More importantly, mammography only provides two-dimensional images, and there is a possibility of missing hidden lesions due to overlapping lesions. Breast ultrasound and thermal imaging are radiation-free and suitable for pregnant women, lactating women and young women, especially for evaluating masses in dense breast tissue, but their sensitivity and specificity are low, it is difficult to accurately locate or detect deep lesions, and there is also a risk of missed diagnosis. Although magnetic resonance imaging and molecular imaging have high resolution and can clearly display the structure of breast tissue, their high cost limits their application in large-scale screening. In summary, there is an urgent need for a convenient, simple to operate, low-cost, non-invasive and radiation-free breast tissue lesion screening tool in clinical practice, which has important clinical significance for the early detection and treatment of breast cancer.

[0004] In recent years, ultrasonic shear wave elastography (SWE) has developed rapidly. As a means of non-destructive measurement of tissue biomechanical properties, it has been widely used in many fields such as liver, breast, heart, etc. However, the existing ultrasonic shear wave imaging technology based on acoustic radiation force or external vibration excitation by steady-state exciter has certain safety risks. Specifically, acoustic radiation force elastography requires focused ultrasound to act directly on biological tissues, and its safety for diseased tissues has not yet been clarified; the external vibration excitation method using a steady-state exciter may cause transient huge acceleration, thereby generating shock waves, which may damage the microenvironment of diseased tissues; in addition, multiple reflections from organ boundaries, internal tissue heterogeneity, and mode conversion will produce complex wave fields in time and space, resulting in signal interference and superposition, affecting the accuracy of the shear wave propagation path, increasing the difficulty of tissue property evaluation, and thus causing uncertainty and errors in the results.

[0005] The mechanical properties of soft tissue are complex and cannot be fully described by simple integer-order models. In the absence of fractional-order theory, existing technologies may not accurately characterize the elasticity, viscosity and other mechanical behaviors of soft tissues such as breast tissue, thereby affecting the imaging effect. When processing these data to generate imaging results, existing technologies may not be able to do so quickly and effectively, affecting the real-time and accuracy of imaging. Summary of the invention

[0006] The purpose of the present invention is to provide a low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method and system based on FPINN to solve one or more of the above-mentioned technical problems. In the technical solution provided by the present invention, the FPINN model can quickly convert the input breast tissue elasticity, viscosity, and flow data into imaging results during the imaging process. Compared with the existing traditional imaging methods, it does not require complex manual feature extraction and lengthy calculation processes. The present invention utilizes the parallel computing capabilities of neural networks and the learned mapping relationships to efficiently generate imaging results, thereby improving the real-time and accuracy of imaging.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method based on FPINN, comprising the following steps:

[0009] Acquire and process the displacement data of the breast tissue to be imaged to obtain frequency domain information; obtain the shear wave velocities at different frequencies based on the frequency domain information, and obtain the elasticity, viscosity and fluidity data of the breast tissue according to the shear wave velocities at different frequencies;

[0010] Based on the elasticity, viscosity and fluidity data of breast tissue, the trained FPINN model is used for imaging processing to obtain the imaging results of breast tissue;

[0011] In the training process of the FPINN model, the input is the sample data of breast tissue elasticity, viscosity and fluidity, and the output is the breast tissue imaging result obtained according to the viscoelastic properties of different positions inside the breast tissue. The parameters are updated by minimizing the defined loss function, and the training is completed after reaching the preset convergence condition.

[0012] The loss function L is defined as fPINN The expression is:

[0013] L fPINN =L fPDE +β1L data1 +β2L dat ;

[0014] Where, L fPDE is the model-driven loss, L data1 , L data2 They are the first experimental data driven loss and the second experimental data driven loss respectively; β1 and β2 are L data1 , L dat The weight coefficient of

[0015]

[0016] In the formula, is the equivalent Lame coefficient; u is the particle vibration displacement; t is time; f idc is the reverberation excitation source; μ and λ are the Lame coefficients; α is the fractional order; ω is the vibration angular frequency; k is the wave number; x is the spatial position vector; i is the imaginary unit; A j is the excitation amplitude of the jth excitation source;

[0017]

[0018] Where u1 is the displacement field predicted by the neural network in the FPINN model; u1 * displacement fields estimated for ultrafast ultrasound plane wave imaging;

[0019]

[0020] In the formula, c s is the shear wave velocity derived from the Lame coefficient predicted by the neural network in the FPINN model; c s * is the shear wave velocity obtained by the reverberation method.

[0021] A further improvement of the method of the present invention is that

[0022] In the step of obtaining the shear wave velocity at different frequencies based on the frequency domain information, firstly, an approximate wave value is extracted from the frequency domain information using finite differences, and then the shear wave velocity is calculated based on the extracted approximate wave value;

[0023] The calculation expression of the approximate wave value is:

[0024]

[0025] In the formula, is the approximate wave number; C is equal to 10 / (Δx 2 Bvv(0)) is a constant; Δx is the value of autocorrelation in a certain period of time; Bvv(0) and Bvv(Δx) are the autocorrelation coefficients at Δt=0 and Δt=x;

[0026] The calculation expression of shear wave velocity is:

[0027]

[0028] In the formula, C s is the shear wave velocity; ω1 is the angular frequency of the shear wave; k is the extracted approximate wave value.

[0029] A further improvement of the method of the present invention is that

[0030] In the step of obtaining the elasticity, viscosity and fluidity data of breast tissue according to the shear wave velocity at different frequencies, the shear wave velocity is fitted using the KVFD model to obtain the elasticity, viscosity and fluidity data of breast tissue.

[0031] A further improvement of the method of the present invention is that

[0032] In the step of obtaining and processing the displacement data of the breast tissue to be imaged, a breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging obtains the displacement data of the breast tissue; Fourier transform is performed on the displacement data of the breast tissue through an autocorrelation function to obtain frequency domain information;

[0033] Among them, the breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging includes a low-frequency resonance structure and an ultrasonic detection structure; the low-frequency resonance structure includes a signal generator, a power amplifier and a speaker; one end of the power amplifier is connected to the signal generator, and the other end of the power amplifier is used to connect to the breast tissue to be tested; the speaker is used to be arranged on the breast tissue to be tested; the ultrasonic detection structure is used to connect to the breast tissue to be tested; when in use, the signal generator is used to generate a continuous sinusoidal signal, and the power amplifier drives the speaker to vibrate to generate shear waves and transmit them to the breast tissue to be tested; the ultrasonic detection structure is used to obtain radio frequency data by emitting ultrasonic waves multiple times and receiving reflected signals, and perform beam synthesis on the radio frequency data to obtain displacement data of the breast tissue.

[0034] A further improvement of the method of the present invention is that

[0035] The expression of the autocorrelation function is:

[0036] R(x,x)=E(X(t1)*X(t2));

[0037] Where E is the expectation, X(t1) and X(t2) are the signal strengths at different times, and R(x, x) represents the linear correlation between the values ​​of t1 and t2 at any different times.

[0038] In a second aspect, the present invention provides a low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN, comprising:

[0039] A data acquisition module is used to acquire displacement data of breast tissue to be imaged and process it to obtain frequency domain information; based on the frequency domain information, shear wave velocities at different frequencies are acquired, and breast tissue elasticity, viscosity and fluidity data are acquired according to the shear wave velocities at different frequencies;

[0040] An imaging module is used to perform imaging processing based on the obtained breast tissue elasticity, viscosity and fluidity data using the trained FPINN model to obtain the imaging results of the breast tissue;

[0041] In the training process of the FPINN model, the input is the sample data of breast tissue elasticity, viscosity and fluidity, and the output is the breast tissue imaging result obtained according to the viscoelastic properties of different positions inside the breast tissue. The parameters are updated by minimizing the defined loss function, and the training is completed after reaching the preset convergence condition.

[0042] The loss function L is defined as fPINN The expression is:

[0043] L fPINN =L fPDE +β1L data+β2L data2 ;

[0044] Where, L fPDE is the model-driven loss, L data1 , L data2 They are the first experimental data driven loss and the second experimental data driven loss respectively; β1 and β2 are L data1 , L data2 The weight coefficient of

[0045]

[0046]

[0047] In the formula, is the equivalent Lame coefficient; u is the particle vibration displacement; t is time; f idc is the reverberation excitation source; μ and λ are the Lame coefficients; α is the fractional order; ω is the vibration angular frequency; k is the wave number; x is the spatial position vector; i is the imaginary unit; A j is the excitation amplitude of the jth excitation source;

[0048]

[0049] Where u1 is the displacement field predicted by the neural network in the FPINN model; u1 * displacement fields estimated for ultrafast ultrasound plane wave imaging;

[0050]

[0051] In the formula, c s is the shear wave velocity derived from the Lame coefficient predicted by the neural network in the FPINN model; c s * is the shear wave velocity obtained by the reverberation method.

[0052] A further improvement of the system of the present invention is that:

[0053] In the step of acquiring shear wave velocities at different frequencies based on the frequency domain information executed by the data acquisition module, an approximate wave value is first extracted from the frequency domain information using finite differences, and then the shear wave velocity is calculated based on the extracted approximate wave value;

[0054] The calculation expression of the approximate wave value is:

[0055]

[0056] In the formula, is the approximate wave number; C is equal to 10 / (Δx 2Bvv(0)) is a constant; Δx is the value of autocorrelation in a certain period of time; Bvv(0) and Bvv(Δx) are the autocorrelation coefficients at Δt=0 and Δt=x;

[0057] The calculation expression of shear wave velocity is:

[0058]

[0059] In the formula, C s is the shear wave velocity; ω1 is the angular frequency of the shear wave; k is the extracted approximate wave value.

[0060] A further improvement of the system of the present invention is that:

[0061] In the step of obtaining breast tissue elasticity, viscosity and fluidity data according to shear wave velocities at different frequencies executed by the data acquisition module, shear wave velocity fitting is performed using a KVFD model to obtain breast tissue elasticity, viscosity and fluidity data.

[0062] A further improvement of the system of the present invention is that:

[0063] In the step of acquiring and processing the displacement data of the breast tissue to be imaged, the data acquisition module acquires the displacement data of the breast tissue based on the breast tissue imaging device of low-frequency vibration and reverberation shear wave imaging; and performs Fourier transform on the displacement data of the breast tissue through the autocorrelation function to obtain frequency domain information;

[0064] Among them, the breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging includes a low-frequency resonance structure and an ultrasonic detection structure; the low-frequency resonance structure includes a signal generator, a power amplifier and a speaker; one end of the power amplifier is connected to the signal generator, and the other end of the power amplifier is used to connect to the breast tissue to be tested; the speaker is used to be arranged on the breast tissue to be tested; the ultrasonic detection structure is used to connect to the breast tissue to be tested; when in use, the signal generator is used to generate a continuous sinusoidal signal, and the power amplifier drives the speaker to vibrate to generate shear waves and transmit them to the breast tissue to be tested; the ultrasonic detection structure is used to obtain radio frequency data by emitting ultrasonic waves multiple times and receiving reflected signals, and perform beam synthesis on the radio frequency data to obtain displacement data of the breast tissue.

[0065] A further improvement of the system of the present invention is that:

[0066] The expression of the autocorrelation function is:

[0067] R(x,x)=E(X(t1)*x(t2));

[0068] Where E is the expectation, X(t1) and X(t2) are the signal strengths at different times, and R(x, x) represents the linear correlation between the values ​​of t1 and t2 at any different times.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] In the low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method based on FPINN proposed by the present invention, the trained FPINN model is used for imaging processing, and the input breast tissue elasticity, viscosity, and flow data can be quickly converted into imaging results. Compared with the existing traditional imaging methods, it does not require complex manual feature extraction and lengthy calculation processes. The present invention uses the parallel computing ability of neural networks and the learned mapping relationship to efficiently generate imaging results, thereby improving the real-time and accuracy of imaging. Explanatory, in the training phase of the FPINN model, a large amount of breast tissue sample data (including known elasticity, viscosity, and flow data and their corresponding imaging result labels) is used for training, and the model can learn the complex mapping relationship between breast tissue data and imaging results. This data-driven learning method can make full use of sample information and improve the adaptability of the model to different situations. The FPINN model itself integrates physical information, can better understand and process the mechanical property data of breast tissue, and can more accurately reflect the actual structure and state of breast tissue during the imaging process by combining physical principles with neural networks, avoiding imaging results that do not conform to physical laws that may occur due to pure data drive, and further improving the accuracy of imaging. To further explain, the fractional wave equation is embedded in the neural network as the first loss function L fPDE ; The displacement field u1 estimated by ultrafast ultrasound plane wave imaging * The displacement field u1 predicted by the neural network is used as the second loss function; the shear wave velocity c obtained by the reverberation method is used s * The shear wave velocity c derived from the Lame coefficient predicted by the neural network s As the third loss function; by embedding the fractional-order wave equation into the neural network as the loss function L fPDE, which can ensure that the prediction results of the neural network are consistent with the laws of physics. The fractional-order wave equation can more accurately describe the wave phenomenon in complex media. Therefore, this embedding method helps to improve the accuracy of the prediction results. The second loss function compares the displacement field estimated by ultrafast ultrasonic plane wave imaging with the displacement field predicted by the neural network. Since the ultrafast ultrasonic plane wave imaging technology itself has high accuracy, this comparison can prompt the neural network to learn more accurate displacement field prediction capabilities, thereby improving the accuracy of imaging. The third loss function further constrains the prediction results of the neural network by comparing the shear wave velocity obtained by the reverberation method with the shear wave velocity derived from the Lame coefficient predicted by the neural network, making it more consistent with physical reality, thereby improving the accuracy of imaging. By optimizing the three loss functions at the same time, the neural network can learn more generalized feature representations between different tasks. This multi-task learning method helps to improve the generalization ability of the neural network, so that it can maintain high accuracy under different data sets and conditions. In addition, by embedding physical information into the loss function, the neural network can learn the inherent laws of the data faster during the training process, thereby accelerating the convergence process, which means that the neural network can achieve higher accuracy in a shorter time, which helps to improve the real-time performance of imaging.

[0071] In the preferred technical solution of the present invention, the displacement data of breast tissue is obtained by a specific technology, the displacement data is processed to obtain frequency domain information, and the shear wave velocities at different frequencies are obtained based on the frequency domain information. This process requires an accurate physical model and calculation method. According to physical principles, the shear wave velocity is related to the mechanical properties of the tissue. The shear wave velocities at different frequencies are calculated through precise mathematical formulas and algorithms, thereby providing support for accurately obtaining the elasticity, viscosity, and fluidity data of breast tissue.

[0072] In the preferred technical solution of the present invention, a breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging is proposed. A signal generator is set to generate a continuous sinusoidal signal, and a power amplifier is used to drive the speaker to vibrate to generate shear waves that are transmitted to the breast; at the same time, a high-frame rate ultrasound probe is used to transmit ultrasonic waves multiple times and receive reflected signals to collect radio frequency data, which is convenient for subsequent observation of the displacement of breast tissue under shear wave excitation. Specifically, the signal generator and the power amplifier are connected so that they can generate the vibration frequency required for the experiment. The speaker is fixed on the breast tissue, and the ultrasound probe is placed above the breast. The low-frequency resonance device undergoes multiple reflections in the breast tissue to generate reverberation shear waves that cause the breast tissue to produce a small displacement. At the same time, the ultrasound probe emits a high-frame-rate planar ultrasound wave and receives the echo signal to obtain the radio frequency data. The radio frequency data is beamformed to obtain the breast tissue displacement data, and the breast tissue displacement data is processed to obtain the frequency domain information. The shear wave velocity at different frequencies is obtained based on the frequency domain information. The elasticity, viscosity, and flow data of the breast tissue are obtained according to the shear wave velocity at different frequencies. The device uses the reverberation signal generated by the low-frequency vibration signal to significantly improve the sensitivity of the shear wave in the detection of deep micro-lesion tissue, and realizes the imaging of the viscoelastic flow parameters of the breast tissue, providing all-round and high-resolution imaging of the breast tissue. The device has higher safety, and the low-frequency vibration has less impact on the lesion tissue, which is suitable for large-scale clinical promotion and application. In addition, the low-frequency vibration massage function of the device can not only effectively stimulate the shear wave signal, but also further enhance the comfort experience of detection, ensuring accurate screening while paying more attention to the optimization of patient experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below; obviously, the drawings described below are some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0074] Figure 1 It is a flow chart of a method for ultrafast low-frequency reverberation breast ultrasound viscoelastic imaging based on FPINN in an embodiment of the present invention;

[0075] Figure 2 is a schematic structural diagram of a breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging in an embodiment of the present invention;

[0076] Figure 3 It is a detailed flow chart of a breast tissue imaging method based on low-frequency vibration and reverberation shear wave imaging in an embodiment of the present invention;

[0077] Figure 4is a schematic diagram of a FPINN model in an embodiment of the present invention;

[0078] Figure 5 is a schematic diagram of the particle vibration results at the maximum displacement at different frequencies in an embodiment of the present invention; wherein, Figure 5 (a) is a schematic diagram at 100Hz. Figure 5 (b) is a schematic diagram at 200Hz. Figure 5 (c) is a schematic diagram at 300Hz. Figure 5 (d) is a schematic diagram at 400Hz. Figure 5 (e) is a schematic diagram at 500Hz. Figure 5 (f) is a schematic diagram at 600Hz. Figure 5 (g) is a schematic diagram at 700Hz. Figure 5 (h) is a schematic diagram at 800Hz. Figure 5 (i) is a schematic diagram at 900 Hz. Figure 5 (j) is a schematic diagram at 1000 Hz;

[0079] Figure 6 is a schematic diagram of shear wave imaging results of deep micro-abnormal tissue at different frequencies in an embodiment of the present invention; wherein, Figure 6 (a) is a schematic diagram at 100Hz. Figure 6 (b) is a schematic diagram at 200Hz. Figure 6 (c) is a schematic diagram at 300Hz. Figure 6 (d) is a schematic diagram at 400Hz. Figure 6 (e) is a schematic diagram at 500Hz. Figure 6 (f) is a schematic diagram at 600Hz. Figure 6 (g) is a schematic diagram at 700Hz. Figure 6 (h) is a schematic diagram at 800Hz. Figure 6 (i) is a schematic diagram at 900 Hz. Figure 6 (j) is a schematic diagram at 1000 Hz;

[0080] Figure 7 is a schematic diagram of a low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN in an embodiment of the present invention;

[0081] The explanations of the reference numerals in the figures are as follows:

[0082] 1. Signal generator; 2. Power amplifier; 3. Speaker; 4. Ultrasonic detection structure. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention; it is obvious that the described embodiments and technical solutions are only part of the embodiments of the present invention, not all of the embodiments.

[0084] All other embodiments obtained by those of ordinary skill in the art without creative work based on the technical solutions disclosed in the embodiments of the present invention belong to the scope of protection of the present invention. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0085] See also Figure 1 The embodiment of the present invention provides a low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method based on FPINN, comprising the following steps:

[0086] Step 1, obtaining displacement data of breast tissue to be imaged and processing it to obtain frequency domain information; obtaining shear wave velocities at different frequencies based on the frequency domain information, and obtaining breast tissue elasticity, viscosity and fluidity data according to the shear wave velocities at different frequencies;

[0087] Step 2: Based on the obtained breast tissue elasticity, viscosity and fluidity data, the trained FPINN model is used to perform imaging processing to obtain the imaging results of the breast tissue;

[0088] In the training process of the FPINN model, the input is the sample data of breast tissue elasticity, viscosity and fluidity, and the output is the breast tissue imaging result obtained according to the viscoelastic properties of different positions inside the breast tissue. The parameters are updated by minimizing the defined loss function, and the training is completed after reaching the preset convergence condition.

[0089] The loss function L is defined as fPINN The expression is:

[0090] L fPINN =L fPDE +β1L data1 +β2L dat ;

[0091] Where, L fPDE is the model-driven loss, L data1 , L data2 They are the first experimental data driven loss and the second experimental data driven loss respectively; β1 and β2 are L data1 , L data2The weight coefficient of

[0092]

[0093] In the formula, is the equivalent Lame coefficient; u is the particle vibration displacement; t is time; f idc is the reverberation excitation source; μ and λ are the Lame coefficients; α is the fractional order; ω is the vibration angular frequency; k is the wave number; x is the spatial position vector; i is the imaginary unit; A j is the excitation amplitude of the jth excitation source;

[0094]

[0095] Where u1 is the displacement field predicted by the neural network in the FPINN model; u1 * displacement fields estimated for ultrafast ultrasound plane wave imaging;

[0096]

[0097] In the formula, c s is the shear wave velocity derived from the Lame coefficient predicted by the neural network in the FPINN model; c s * is the shear wave velocity obtained by the reverberation method.

[0098] In practical applications, the prediction speed of neural networks is usually much faster than traditional physical simulation methods. By embedding physical information into neural networks and training them, more efficient calculations can be achieved while maintaining high accuracy, which helps to improve the real-time performance of imaging and enable it to meet the needs of real-time monitoring and diagnosis.

[0099] See also Figure 1 and Figure 2 The embodiment of the present invention provides a low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method based on FPINN, comprising the following steps:

[0100] Obtain breast tissue displacement data, process the breast tissue displacement data to obtain frequency domain information, obtain shear wave velocities at different frequencies based on the frequency domain information, and obtain breast tissue elasticity, viscosity, and fluidity data according to the shear wave velocities at different frequencies;

[0101] In the exemplary technical solution, breast tissue displacement data is obtained based on a breast tissue imaging device using low-frequency vibration and reverberation shear wave imaging. The structure of the breast tissue imaging device is shown in FIG. Figure 2As shown, it includes a low-frequency resonance structure and an ultrasonic detection structure 4; the low-frequency resonance structure includes a signal generator 1, a power amplifier 2 and a speaker 3; one end of the power amplifier 2 is connected to the signal generator 1, and the other end of the power amplifier 2 is connected to the breast tissue to be tested; the speaker 3 is arranged on the breast tissue to be tested; and the ultrasonic detection structure 4 is connected to the breast tissue to be tested. The signal generator 1 and the power amplifier 2 are correctly connected so that they can generate the vibration frequency required for the experiment, and the ultrasonic probe of the ultrasonic detection structure 4 is connected to the position above the breast tissue to be tested through a coupling agent. The low-frequency resonance frequency is between 100Hz and 1000Hz, with a step size of 100Hz; there are several speakers 3, and several speakers 3 are evenly arranged on the breast tissue to be tested. In the embodiment of the present invention, four speakers 3 are fixed in the four directions of up, down, left and right of the breast tissue for experiment. Since the breast tissue is relatively soft and the surface is irregular, a soft material is used to fit and fix the speaker 3 to the breast tissue to ensure that the contact surface fully transmits vibration, including but not limited to the use of silicone pads or medical tapes, flexible straps, flexible supports, gel patches, etc. The gain of the power amplifier is set to X2 or X10. The power amplifier has a high input impedance of 10kΩ and a complete output protection circuit (output overcurrent protection, internal temperature abnormality protection) to ensure the stable, reliable and safe operation of the instrument. It has the characteristics of small size, light weight and easy use, and is suitable for various environments. The frequency range of the speaker is 80Hz~20KHz, the diameter is 30mm, and the total weight does not exceed 50g. It is small and light. It is a speaker that can produce low-frequency resonance and multi-functional vibration massage therapy. It is suitable for various environments. The speaker is actually a low-frequency resonance vibration speaker with a multi-functional massage effect. It can play a certain massage role when the patient is tested, which can effectively relieve the patient's tension and relax him, which is conducive to data collection. The signal generator is an arbitrary waveform generator with an output power less than or equal to 60MHz, a vertical resolution of 14bits, and a sampling rate of 200Msa / s. The signal generator can generate high-quality arbitrary waveforms to ensure that the signal changes in the time domain are smoother and rich in details.

[0102] To explain the principle, the signal generator 1 generates a continuous sinusoidal signal, which drives the speaker 3 to vibrate through the power amplifier 2 to generate shear waves that are transmitted to the breast tissue. The ultrasonic detection structure 4 then uses multiple transmissions of ultrasonic waves and receives reflected signals to obtain radio frequency data, which is convenient for subsequent observations of the displacement of the breast tissue under the excitation of the shear wave. To further explain, when a low-frequency vibrating speaker is fixed on the breast tissue for a shear wave experiment, the key is to ensure good contact and stable fixation between the vibration source and the tissue to avoid falling off or large positional displacement during the vibration process, so as to effectively transmit the vibration without causing damage to the tissue or affecting the normal experimental results.

[0103] The radio frequency data is beam synthesized to obtain breast tissue displacement data, the breast tissue displacement data is processed to obtain frequency domain information, the shear wave velocities at different frequencies are obtained based on the frequency domain information, and the elasticity, viscosity and fluidity data of the breast tissue are obtained according to the shear wave velocities at different frequencies.

[0104] The breast tissue displacement data is processed to obtain frequency domain information, specifically:

[0105] The breast tissue displacement data is Fourier transformed through the autocorrelation function to obtain frequency domain information;

[0106] The autocorrelation function is expressed as follows:

[0107] R(x,x)=E(X(t1)*x(t2));

[0108] Where E is the expectation, X(t1) and X(t2) are the signal strengths at different times, and R(x, x) represents the linear correlation between the values ​​of t1 and t2 at any different times.

[0109] In the embodiment of the present invention, the shear wave velocity at different frequencies is obtained based on the frequency domain information, specifically: firstly, an approximate wave value is extracted from the frequency domain information using finite differences, and then the shear wave velocity is calculated based on the extracted approximate wave value;

[0110] The calculation expression of the approximate wave value is:

[0111]

[0112] In the formula, is the approximate wave number; C is equal to 10 / (Δx 2 Bvv(0)) is a constant; Δx is the value of autocorrelation in a certain period of time; Bvv(0) and Bvv(Δx) are the autocorrelation coefficients at Δt=0 and Δt=x;

[0113] The calculation expression of shear wave velocity is:

[0114]

[0115] In the formula, C s is the shear wave velocity; ω1 is the angular frequency of the shear wave; k is the extracted approximate wave value.

[0116] The elasticity, viscosity and fluidity data of breast tissue were obtained according to the shear wave velocity at different frequencies. Specifically, the shear wave velocity was fitted using the KVFD model (Kelvin-Voigt Fractional Derivative Model) to obtain the elasticity, viscosity and fluidity data of breast tissue.

[0117] Combine the following Figure 3 and Figure 4 A detailed description is given as follows: In the embodiment of the present invention, the low-frequency resonance device undergoes multiple reflections in the breast tissue, generating reverberation shear waves to cause a small displacement of the breast tissue. Specifically, a high-frame rate imaging method is used to track the displacement changes inside the tissue, and a Loupas displacement estimator is used to estimate the phase change of the ultrasonic reflection echo to calculate the axial displacement, thereby obtaining an estimated value of the small displacement; at the same time, the probe of the ultrasonic detection structure transmits a high-frame rate planar ultrasonic wave, and receives the echo signal to obtain radio frequency data, and the radio frequency data is filtered and beam synthesized. The beam synthesis method uses the classic algorithm DAS (Delay-and-Sum) to obtain a synthesized beam; breast tissue displacement data is obtained according to the synthesized beam;

[0118] The breast tissue displacement data is Fourier transformed through the autocorrelation function to obtain frequency domain information;

[0119] Finite differences are used to approximate the wavenumbers extracted from the frequency domain information. According to the approximate wave number The shear wave velocity is calculated, and low-frequency sinusoidal electrical signals of different frequencies are used to excite the low-frequency resonance device to generate reverberant shear wave fields at different frequencies, and the shear wave propagation velocities at different frequencies are obtained. The above steps are repeated using low-frequency sinusoidal electrical signals of different frequencies to obtain the shear wave velocities at different frequencies.

[0120] Using the KVFD model to fit the shear wave velocity can obtain the multi-parameter tissue mechanical properties such as elasticity, viscosity, and fluidity. Its constitutive equation is: σ(t)=E0ε(t)+ξD α [ε(t)]; This formula can describe the constitutive relationship of soft tissue, where σ(t) and ε(t) represent the stress and strain under the action of external force, E0 represents the Young's modulus of the tissue, that is, elasticity, and ξ represents the viscosity of the tissue, that is, viscosity, and the unit is Pa·S a , D a represents the fractional derivative, a represents the fluidity of the organization, and Figure 4 Formula and KVFD equation, we can get

[0121]

[0122] This formula can be used to fit multi-frequency shear wave velocities, thereby achieving viscoelastic flow parametric imaging. By calculating the viscosity, elasticity, and flow properties of tissues, it overcomes the limitation of existing imaging that can only perform breast tissue elasticity, which is of great significance for the screening and diagnosis of early breast tumors.

[0123] In the embodiment of the present invention, the trained FPINN model is used to perform imaging processing based on the elasticity, viscosity and fluidity data of breast tissue to obtain breast tissue imaging results. Compared with the integer order model, the fractional order model can more accurately describe the physical properties of soft matter, so the fractional order Kelvin-Voigt model (KVFD) is used to describe the constitutive equation. Figure 4 As shown in the figure, the fractional-order physical information neural network FPINN is a method that combines deep learning with fractional-order physical modeling to solve fractional-order partial differential equations (PDEs). The key idea of ​​FPINN is to embed fractional-order physical constraints into neural networks, so that the network can learn the behavior of physical systems and satisfy physical equations. The steps to obtain the FPINN model include: defining the physical model, defining the physical equations and constraints of the system; building a neural network, building a neural network structure containing physical information; designing a loss function to train the network: training the network by minimizing the loss function through an optimization algorithm so that it approximates the physical equations and satisfies the constraints. Solving problems: Use the trained network to solve physical problems.

[0124] The FPINN model of the present invention uses position x, y, z and time t as inputs of the neural network, presents the training input data to the neural network, and uses a loss function to compare the resulting output with the training output data. The loss returned by the function is used to adjust the weights of the network through back propagation to reduce the loss.

[0125] The FPINN of the present invention uses a custom loss function, which includes an additional loss component to constrain the neural network to produce outputs that meet the modeled differential equation, predict the numerical solution of the partial differential equation under this condition, and produce the corresponding displacement u as output. Training FPINN to meet the requirements of the differential equation requires outputting the corresponding fractional derivatives and second-order derivatives, that is, and These derivatives can be obtained in TensorFlow and PyTorch through the automatic differentiation functions of each platform.

[0126] The technical solution of the embodiment of the present invention is characterized by a new design of the Loss function, which includes three parts:

[0127] Part I fPDE Drive the loss function for the model:

[0128]

[0129] The second part is the loss function driven by ultrasonic experimental data:

[0130]

[0131] The third part is the loss function driven by the KVFD model and reverberation experimental data:

[0132]

[0133] The loss function of the entire FPINN neural network is the weighted sum of three loss functions:

[0134] L fPINN =L fPDE +β1L data1 +β2L data2 ;

[0135] The technical solution of the embodiment of the present invention utilizes FPINN one-time training to realize multi-frequency mechanical parameter learning, and utilizes the KVFD model to realize viscoelastic flow parameter imaging; further, when the network is trained, reverberation excitation sources of multiple frequencies are given, and the trained network can realize physical parameter learning of reverberation source excitations of multiple frequencies, thereby greatly improving the learning efficiency.

[0136] On this basis, using Fit the viscoelastic flow parameters to achieve breast viscoelastic flow parameter imaging using FPINN.

[0137] Figure 5 The vibration displacement diagram of the particle at the maximum displacement at different frequencies is shown. Figure 5 (a) to (j) in the figure show the time domain response of the vibration system with a frequency of 100 to 1000 Hz respectively; the results show that the initial displacement of the particle is large, and the amplitude gradually decays over time, showing typical damped vibration characteristics, which is consistent with the propagation law of the attenuation characteristics of vibration signals in soft tissues.

[0138] Figure 6 The shear wave imaging results of deep micro-abnormal tissues at different frequencies are shown. Figure 6 It can be clearly seen in (a) that when the vibration frequency is 100 Hz, the shear wave imaging result is significantly poor, and the shear wave velocity of deep and small abnormal tissues cannot be accurately calculated; Figure 6 (b) and Figure 6 It can be clearly seen in (c) that when the vibration frequency is 200Hz and 300Hz, the shear wave velocity of the deep abnormal tissue can be accurately calculated, but a strong interference signal is generated in the top area of ​​the soft tissue; Figure 6 Middle (d) and Figure 6 It can be clearly seen in (h) that when the vibration frequency is 400Hz~800Hz, the shear wave velocity of deep small abnormal tissue can be accurately calculated, and the surrounding interference signal is weak; when the low-frequency vibration signal is further enhanced, such as Figure 6 (i) and Figure 6As shown in (j), when the frequency is 900Hz and 1000Hz, interference signals are generated in the surroundings, and the shear wave velocity of deep micro-abnormal tissue cannot be accurately calculated. Therefore, from the experimental results, low-frequency vibrations of 400Hz to 800Hz can effectively enhance the reverberation field strength, and thus effectively distinguish the shear wave velocity of deep micro-abnormal tissue and surrounding normal tissue.

[0139] Specifically and exemplarily, the breast tissue imaging method based on the FPINN model of the present invention includes:

[0140] Step (1), obtaining and processing displacement data of breast tissue, includes:

[0141] Data acquisition: Ultrafast ultrasound imaging technology is used to scan breast tissue and obtain displacement data of tissue at different time points. This data can be expressed as a displacement field of a time series, i.e., u(x,t), where x is the spatial position and t is the time.

[0142] Frequency domain information processing: Perform Fourier transform on the acquired displacement data to convert it from the time domain to the frequency domain to obtain the frequency domain information U(k,ω), where k is the wave number and ω is the vibration angular frequency.

[0143] Shear wave velocity calculation: Based on frequency domain information, the shear wave velocity c_s(ω) at different frequencies is calculated using wave theory. The shear wave velocity is closely related to the elasticity, viscosity and fluidity characteristics of the tissue.

[0144] Obtain breast tissue characteristic data: Based on the shear wave velocity c_s(ω), combined with the known wave equation and material mechanics model, the elasticity, viscosity and fluidity data of breast tissue are inverted. These data can be expressed as a function of spatial position, namely E(x), ν(x), η(x), etc., where E is the elastic modulus, ν is the Poisson's ratio, and η is the viscosity coefficient.

[0145] Step (2): Using the FPINN model to perform imaging processing, including:

[0146] FPINN model training: Input data: collect a batch of breast tissue elasticity, viscosity and fluidity sample data as the input of the FPINN model; Output data: obtain the corresponding breast tissue imaging results according to the viscoelastic properties of different positions inside the breast tissue as the output of the FPINN model; by minimizing the loss function, use the optimization algorithm (such as gradient descent method) to update the parameters of the FPINN model. When the loss function reaches the preset convergence condition, the training of the FPINN model is completed.

[0147] Imaging processing: The acquired breast tissue elasticity, viscosity and fluidity data are input into the trained FPINN model. The FPINN model predicts the imaging results of breast tissue based on the input data, combined with physical laws and experimental data. The imaging results can be expressed as a function of spatial position, i.e., I(x), where I is the imaging intensity or imaging parameter.

[0148] Specific example parameters, frequency domain information processing: Fourier transform uses the fast Fourier transform (FFT) algorithm, and the frequency domain range is selected as 0 to 1000Hz; shear wave velocity calculation: the wave equation is used to calculate the shear wave velocity, considering the influence of fractional derivatives, and the equivalent Lame coefficient is calculated using the formula. FPINN model training: input data: 1000 sets of breast tissue elasticity, viscosity and fluidity sample data; output data: corresponding 1000 sets of breast tissue imaging results; loss function weight coefficient: β1 = 0.5, β2 = 0.5; optimization algorithm: use the Adam optimization algorithm to update parameters; convergence condition: the loss function value is less than 0.01. Imaging processing: the acquired breast tissue characteristic data is input into the FPINN model, and the FPINN model outputs the imaging results of the breast tissue. The imaging results are displayed in the form of grayscale images, where the grayscale value represents the imaging intensity. Through the above steps and parameter settings, accurate imaging of breast tissue can be achieved, which can provide strong support for the diagnosis and treatment of breast diseases.

[0149] The following are device embodiments of the present invention, which can be used to implement the method embodiments of the present invention. For details not disclosed in the device embodiments, please refer to the method embodiments of the present invention.

[0150] See also Figure 7 In an embodiment of the present invention, a low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN is provided, comprising:

[0151] A data acquisition module is used to acquire displacement data of breast tissue to be imaged and process it to obtain frequency domain information; based on the frequency domain information, shear wave velocities at different frequencies are acquired, and breast tissue elasticity, viscosity and fluidity data are acquired according to the shear wave velocities at different frequencies;

[0152] An imaging module is used to perform imaging processing based on the obtained breast tissue elasticity, viscosity and fluidity data using the trained FPINN model to obtain the imaging results of the breast tissue;

[0153] In the training process of the FPINN model, the input is the sample data of breast tissue elasticity, viscosity and fluidity, and the output is the breast tissue imaging result obtained according to the viscoelastic properties of different positions inside the breast tissue. The parameters are updated by minimizing the defined loss function, and the training is completed after reaching the preset convergence condition.

[0154] The loss function L is defined as fPINN The expression is:

[0155] L fPINN =L fPDE +β1L data1 +β2L da ;

[0156] Where, L fPD E is the model-driven loss, L data , L data2 They are the first experimental data driven loss and the second experimental data driven loss respectively; β1 and β2 are L data , L data2 The weight coefficient of .

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method based on FPINN, characterized in that: The following steps are involved: Acquiring and processing the displacement data of the breast tissue to be imaged to obtain frequency domain information; Based on the frequency domain information, shear wave velocities at different frequencies are obtained, and breast tissue elasticity, viscosity and fluidity data are obtained according to the shear wave velocities at different frequencies; Based on the elasticity, viscosity and fluidity data of breast tissue, the trained FPINN model is used for imaging processing to obtain the imaging results of breast tissue; In the training process of the FPINN model, the input is the sample data of breast tissue elasticity, viscosity and fluidity, and the output is the breast tissue imaging result obtained according to the viscoelastic properties of different positions inside the breast tissue. The parameters are updated by minimizing the defined loss function, and the training is completed after reaching the preset convergence condition. The loss function L is defined as fPINN The expression is: L fPINN =L fPDE +β1L data1 +β2L data2 ; Where, L fPDE is the model-driven loss, L data1 , L data2 They are the first experimental data driven loss and the second experimental data driven loss respectively; β1 and β2 are L data1 , L data2 The weight coefficient of In the formula, is the equivalent Lame coefficient; u is the particle vibration displacement; t is time; f idc is the reverberation excitation source; μ and λ are the Lame coefficients; α is the fractional order; ω is the vibration angular frequency; k is the wave number; x is the spatial position vector; i is the imaginary unit; A j is the excitation amplitude of the jth excitation source; Where u1 is the displacement field predicted by the neural network in the FPINN model; u1 * displacement fields estimated for ultrafast ultrasound plane wave imaging; In the formula, c s is the shear wave velocity derived from the Lame coefficient predicted by the neural network in the FPINN model; c s * is the shear wave velocity obtained by the reverberation method.

2. The FPINN-based low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method according to claim 1, characterized in that: In the step of obtaining the shear wave velocity at different frequencies based on the frequency domain information, firstly, an approximate wave value is extracted from the frequency domain information using finite differences, and then the shear wave velocity is calculated based on the extracted approximate wave value; The calculation expression of the approximate wave value is: In the formula, is the approximate wave number; C is equal to 10 / (Δx 2 Bvv(0)) is a constant; Δx is the value of autocorrelation in a certain period of time; Bvv(0) and Bvv(Δx) are the autocorrelation coefficients at Δt=0 and Δt=x; The calculation expression of shear wave velocity is: In the formula, C s is the shear wave velocity; ω1 is the angular frequency of the shear wave; k is the extracted approximate wave value.

3. The FPINN-based low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method according to claim 1, characterized in that: In the step of obtaining the elasticity, viscosity and fluidity data of breast tissue according to the shear wave velocity at different frequencies, the shear wave velocity is fitted using the KVFD model to obtain the elasticity, viscosity and fluidity data of breast tissue.

4. The FPINN-based low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method according to claim 1, characterized in that: In the step of obtaining and processing the displacement data of the breast tissue to be imaged, a breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging obtains the displacement data of the breast tissue; Fourier transform is performed on the displacement data of the breast tissue through an autocorrelation function to obtain frequency domain information; The breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging comprises a low-frequency resonance structure and an ultrasonic detection structure (4); the low-frequency resonance structure comprises a signal generator (1), a power amplifier (2) and a loudspeaker (3); one end of the power amplifier (2) is connected to the signal generator (1), and the other end of the power amplifier (2) is used to connect to the breast tissue to be tested; the loudspeaker (3) is used to be arranged on the breast tissue to be tested; the ultrasonic detection structure (4) is used to connect to the breast tissue to be tested; when in use, the signal generator (1) is used to generate a continuous sinusoidal signal, and the power amplifier (2) drives the loudspeaker (3) to vibrate to generate a shear wave that is transmitted to the breast tissue to be tested; the ultrasonic detection structure (4) is used to obtain radio frequency data by emitting ultrasonic waves multiple times and receiving reflected signals, and to perform beam synthesis on the radio frequency data to obtain displacement data of the breast tissue.

5. The FPINN-based low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging method according to claim 4, characterized in that: The expression of the autocorrelation function is: R(x,x)=E(X(t1)*X(t2)); Where E is the expectation, X(t1) and X(t2) are the signal strengths at different times, and R(x,x) represents the linear correlation between the values ​​of t1 and t2 at any different times.

6. A low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN, characterized in that: include: A data acquisition module is used to acquire and process the displacement data of the breast tissue to be imaged to obtain frequency domain information; Based on the frequency domain information, shear wave velocities at different frequencies are obtained, and breast tissue elasticity, viscosity and fluidity data are obtained according to the shear wave velocities at different frequencies; An imaging module is used to perform imaging processing based on the obtained breast tissue elasticity, viscosity and fluidity data using the trained FPINN model to obtain the imaging results of the breast tissue; In the training process of the FPINN model, the input is the sample data of breast tissue elasticity, viscosity and fluidity, and the output is the breast tissue imaging result obtained according to the viscoelastic properties of different positions inside the breast tissue. The parameters are updated by minimizing the defined loss function, and the training is completed after reaching the preset convergence condition. The loss function L is defined as fPINN The expression is: L fPINN =L fPDE +β1L data1 +β2L dat ; Where, L fPDE is the model-driven loss, L data1 , L data2 They are the first experimental data driven loss and the second experimental data driven loss respectively; β1 and β2 are L dat , L data2 The weight coefficient of In the formula, is the equivalent Lame coefficient; u is the particle vibration displacement; t is time; f idc is the reverberation excitation source; μ and λ are the Lame coefficients; α is the fractional order; ω is the vibration angular frequency; k is the wave number; x is the spatial position vector; i is the imaginary unit; A j is the excitation amplitude of the jth excitation source; Where u1 is the displacement field predicted by the neural network in the FPINN model; u1 * displacement fields estimated for ultrafast ultrasound plane wave imaging; In the formula, c s is the shear wave velocity derived from the Lame coefficient predicted by the neural network in the FPINN model; c s * is the shear wave velocity obtained by the reverberation method.

7. The low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN according to claim 6, characterized in that: In the step of acquiring shear wave velocities at different frequencies based on the frequency domain information executed by the data acquisition module, an approximate wave value is first extracted from the frequency domain information using finite differences, and then the shear wave velocity is calculated based on the extracted approximate wave value; The calculation expression of the approximate wave value is: In the formula, is the approximate wave number; C is equal to 10 / (Δx 2 Bvv(0)) is a constant; Δx is the value of autocorrelation in a certain period of time; Bvv(0) and Bvv(Δx) are the autocorrelation coefficients at Δt=0 and Δt=x; The calculation expression of shear wave velocity is: In the formula, C s is the shear wave velocity; ω1 is the angular frequency of the shear wave; k is the extracted approximate wave value.

8. The low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN according to claim 6, characterized in that: In the step of obtaining breast tissue elasticity, viscosity and fluidity data according to shear wave velocities at different frequencies executed by the data acquisition module, shear wave velocity fitting is performed using a KVFD model to obtain breast tissue elasticity, viscosity and fluidity data.

9. The low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN according to claim 6, characterized in that: In the step of acquiring and processing the displacement data of the breast tissue to be imaged, the data acquisition module acquires the displacement data of the breast tissue based on the breast tissue imaging device of low-frequency vibration and reverberation shear wave imaging; and performs Fourier transform on the displacement data of the breast tissue through the autocorrelation function to obtain frequency domain information; The breast tissue imaging device based on low-frequency vibration and reverberation shear wave imaging comprises a low-frequency resonance structure and an ultrasonic detection structure (4); the low-frequency resonance structure comprises a signal generator (1), a power amplifier (2) and a loudspeaker (3); one end of the power amplifier (2) is connected to the signal generator (1), and the other end of the power amplifier (2) is used to connect to the breast tissue to be tested; the loudspeaker (3) is used to be arranged on the breast tissue to be tested; the ultrasonic detection structure (4) is used to connect to the breast tissue to be tested; when in use, the signal generator (1) is used to generate a continuous sinusoidal signal, and the power amplifier (2) drives the loudspeaker (3) to vibrate to generate a shear wave that is transmitted to the breast tissue to be tested; the ultrasonic detection structure (4) is used to obtain radio frequency data by emitting ultrasonic waves multiple times and receiving reflected signals, and to perform beam synthesis on the radio frequency data to obtain displacement data of the breast tissue.

10. The low-frequency reverberation breast ultrasound viscoelastic ultrafast imaging system based on FPINN according to claim 9, characterized in that: The expression of the autocorrelation function is: R(x,x)=E(X(t1)*X(t2)); Where E is the expectation, X(t1) and X(t2) are the signal strengths at different times, and R(x,x) represents the linear correlation between the values ​​of t1 and t2 at any different times.

Images