A fast electromagnetic wave thermal acoustic imaging noise reduction method and device

Abstract

This invention discloses a fast electromagnetic wave thermoacoustic imaging denoising method and apparatus. The method includes: vectorizing multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by an acquisition system and merging and recombining them into a two-dimensional matrix; constructing a robust principal component analysis (PCA) theoretical model based on the two-dimensional matrix; constructing an optimization problem using PCA; iteratively solving the optimization problem; stopping the iteration when the optimization problem meets the convergence condition; extracting the last column of the low-rank matrix obtained through iteration, expanding it, and using it as a new electromagnetic wave thermoacoustic signal for the current frame; and obtaining an electromagnetic wave thermoacoustic image using an imaging algorithm. This invention fully utilizes the correlation between multi-frame electromagnetic wave thermoacoustic signals to obtain prior information about useful signals, and can adapt to moving targets and extract useful signals from them. Therefore, this invention can effectively improve the background noise and external interference problems in electromagnetic wave thermoacoustic imaging, and improve imaging quality.

Description

Technical Field

[0001] This invention belongs to the field of electromagnetic wave thermoacoustic imaging, including microwave thermoacoustic imaging and photoacoustic imaging, and particularly relates to a fast electromagnetic wave thermoacoustic imaging denoising method and apparatus. Background Technology

[0002] Electromagnetic wave thermoacoustic imaging, including microwave thermoacoustic imaging and photoacoustic imaging, is a novel medical imaging technology. It combines the high contrast of electromagnetic wave imaging with the high resolution of ultrasound imaging, thus showing promising applications in medical fields such as breast cancer, cerebral hemorrhage, fatty liver, microwave ablation, vascular tumors, and blood oxygen saturation measurement. Electromagnetic wave thermoacoustic imaging uses electromagnetic waves as an excitation source. Under the influence of electromagnetic waves, biological tissue undergoes thermoelastic expansion, generating ultrasound waves. These ultrasound waves are received by an array of ultrasound probes, and then imaging algorithms are used to create images.

[0003] Electromagnetic wave thermoacoustic imaging frequently encounters noise interference. This includes both thermal noise from the medium and electronic noise from the sensor, collectively known as background noise in electromagnetic wave thermoacoustic images. Background noise is typically mixed with the received signal. Compared to the signal amplitude, this noise is relatively small and randomly distributed. On the other hand, the transducer itself, subjected to radiation, is also affected by electromagnetic waves, generating strong interference signals. The amplitude of this interference-induced noise far exceeds that of the useful signal. Interference and background noise severely impact the quality of electromagnetic wave thermoacoustic images, thus affecting doctors' interpretation of images and diagnosis. Therefore, removing electromagnetic interference and reducing the impact of background noise in electromagnetic wave thermoacoustic imaging is crucial.

[0004] Existing denoising methods for removing background noise, including frequency domain filtering, time-frequency methods, and deep learning methods, are mostly based on single-frame signal denoising and cannot remove strong interference. Solutions for electromagnetic interference include frame averaging, frequency domain filtering, and some time-frequency methods. Among these, filtering methods require that the spectra of the signal and interference do not overlap, while time-frequency methods require prior information about the signal profile. However, these conditions are often not met in practical electromagnetic thermoacoustic imaging. Frame averaging is worth mentioning; it utilizes multi-frame information to enhance the signal and suppress interference. However, it also has limitations. First, since interference is not completely random, it can only reduce the impact of interference, not truly eliminate it. Second, frame averaging reduces the frame rate; increasing the system frame rate can lead to microwave energy accumulation, posing a safety risk. Finally, real biological bodies experience rapid changes in electromagnetic thermoacoustic image results due to respiration, heartbeat, and movement. In such cases, frame averaging can introduce significant errors.

[0005] A study proposed a singular value decomposition (SVD)-based method to address laser interference in photoacoustic imaging. This method performs SVD on the single-frame photoacoustic image matrix, manually truncating the singular values ​​to separate the interference and useful signal. However, this method is only applicable to regular interference during parallel acquisition. Furthermore, the SVD method has two well-known limitations. First, it requires empirical manual truncating of singular values, which corresponds to the rank selection of the signal and interference subspaces, a choice that SVD is sensitive to. Second, SVD has poor robustness. When a large amount of random noise is present in the signal, SVD easily fails.

[0006] Therefore, it is crucial to develop an electromagnetic wave thermoacoustic imaging denoising method that can effectively suppress noise, remove strong interference, adapt to biological movement, require no manual parameter adjustment, and operate at high speed. Summary of the Invention

[0007] Based on the aforementioned problems in the existing technology, this invention provides a fast electromagnetic wave thermoacoustic imaging denoising method and apparatus, which can effectively suppress noise, remove strong interference, adapt to biological movement, eliminate the need for manual parameter adjustment, and operate at high speed.

[0008] The fast electromagnetic wave thermoacoustic imaging denoising method proposed in this invention includes the following steps:

[0009] Step 1: Vectorize the multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by the acquisition system, and merge and reassemble them into a two-dimensional matrix M;

[0010] Step 2: Construct a robust principal component analysis theoretical model based on the two-dimensional matrix. Treat the useful signal in the two-dimensional matrix as the low-rank part and the noise or interference in the two-dimensional matrix as the sparse part, and obtain the useful signal matrix L with low-rank characteristics and the noise matrix S with sparse characteristics, as follows:

[0011] M = L + S(1)

[0012] Step 3: Construct the optimization problem using robust principal component analysis theory and solve iteratively. Stop the iteration when the optimization problem satisfies the convergence condition, and obtain the final useful signal matrix L. ＊ ;

[0013] Step 4: Extract the final useful signal matrix L obtained through iteration. * The last column is expanded and serves as a new electromagnetic wave thermoacoustic signal for the current frame;

[0014] Step 5: Use an imaging algorithm to obtain an image of electromagnetic wave thermoacoustic sound.

[0015] Preferably, the multi-frame, multi-channel electromagnetic wave thermoacoustic signal has a current frame and n-1 past frames, each frame being an s×T matrix, and the merged and recombined two-dimensional matrix...

[0016] Where n represents the number of electromagnetic wave thermoacoustic signal frames used, s represents the number of ultrasonic probes used in the acquisition system to collect electromagnetic wave thermoacoustic signals, T represents the number of sampling points of electromagnetic wave thermoacoustic signals, and each column of the two-dimensional matrix M is a vector stretched from each frame of electromagnetic wave thermoacoustic signal matrix.

[0017] As a preferred approach, the robust principal component analysis problem is transformed into a convex optimization problem using the principal component pursuit method, as follows:

[0018]

[0019] Where ρ represents the regularization parameter, ||·|| * and ||·||1 represent the nuclear norm and l1 norm of the matrix, respectively. They are the rank and convex relaxation of the l0 norm of the matrix. The l0 norm is the number of non-zero elements in the matrix. st is an abbreviation for subject to, and the part after st represents the constraint condition.

[0020] The unconstrained convex optimization problem is constructed using the augmented Lagrange multiplier method, as follows:

[0021]

[0022] in, Let ||·|| denote the constructed augmented Lagrangian function, Λ denote the Lagrange multiplier, and <·> denote the inner product. F Let Frobenius norm of the matrix be denoted, μ denote the adaptive adjustment penalty factor, and := denote the definition as follows;

[0023] Then, the alternating direction method is used to construct the solution function for the convex optimization problem and iteratively solve it. The symmetric alternating direction augmented Lagrangian method and adaptive adjustment of the penalty factor are used to obtain the iterative optimal solution, that is, the final useful signal matrix L. ＊ .

[0024] As a preferred approach, the iterative solution process is as follows:

[0025] (·) k This represents the k-th iteration process; it updates the low-rank characteristic useful signal matrix L of the (k+1)-th iteration. k+1 :

[0026]

[0027] Among them, S kLet Λ represent the sparse property noise matrix of the k-th iteration. k Let μ be the Lagrange multiplier for the k-th iteration. k Let denot be the penalty factor for the k-th iteration. Solve this optimization problem using the singular value threshold operator:

[0028]

[0029] in, Singular value threshold operator:

[0030]

[0031] Where M-S+μΛ=U∑V T Let represent the singular value decomposition of matrix M-S+μΛ, where U represents the left singular vector matrix, ∑ represents the singular value matrix, and V represents the right singular vector matrix. T Z represents the transpose of a matrix. 1 / μ (·) denotes the soft threshold operator:

[0032]

[0033] Where sgn(·) represents the sign function, max represents the maximum value, and x represents each element in the matrix M-S+μΛ; the first update of the Lagrange multipliers

[0034]

[0035] Update the sparse property noise matrix S in the (k+1)th iteration. k+1 :

[0036]

[0037] The optimization problem is solved using the soft threshold operator:

[0038]

[0039]

[0040] Where sgn(·) represents the sign function, max represents the maximum value, and x′ represents each element in the matrix M-L+μΛ. The soft threshold operator operates on each element of the matrix.

[0041] Second update of Lagrange multipliers Λ k+1 :

[0042]

[0043] Update the penalty factor μ in the (k+1)th iteration. k+1:

[0044]

[0045] Where η represents the parameter for adaptively adjusting the penalty factor;

[0046] Iteration relative error The iteration terminates when the value is less than the error limit γ or when the maximum number of iterations K is reached, yielding the final useful signal matrix L. * .

[0047] Preferably, the new one in the current frame

[0048] Preferably, the imaging algorithm can be any electromagnetic wave thermoacoustic imaging algorithm, such as delay superposition algorithm, back projection algorithm, time reversal algorithm, model-based algorithm, and deep learning algorithm.

[0049] The fast electromagnetic wave thermoacoustic imaging denoising device proposed in this invention is applied to the fast electromagnetic wave thermoacoustic imaging denoising method proposed in this invention. It includes an electromagnetic wave thermoacoustic signal acquisition module, a model building module, a fast robust principal component analysis denoising module, and a real-time electromagnetic wave thermoacoustic imaging module.

[0050] The electromagnetic wave thermoacoustic signal acquisition module is used to receive multi-channel electromagnetic wave thermoacoustic signals of the current frame in real time, digitally and in parallel acquire multi-channel electromagnetic wave thermoacoustic signals, and save the digital acquisition results in real time.

[0051] The model building module is used to vectorize the multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by the acquisition system, merge and reassemble them into a two-dimensional matrix; construct a robust principal component analysis theoretical model based on the two-dimensional matrix; model the useful signals in the two-dimensional matrix as low-rank parts, and treat the noise or interference in the signal matrix as sparse parts;

[0052] The fast robust principal component analysis denoising module is used to construct an optimization problem using robust principal component analysis theory, iteratively solve the optimization problem, and stop iterating when the optimization problem meets the iterative convergence condition, thereby obtaining the low-rank matrix and sparse matrix after matrix decomposition.

[0053] The real-time electromagnetic wave thermoacoustic imaging module is used to extract the last column of the low-rank matrix obtained by iteration, expand it, and use it as a new electromagnetic wave thermoacoustic signal for the current frame; the imaging algorithm is driven by the real-time electromagnetic wave thermoacoustic imaging device to obtain an electromagnetic wave thermoacoustic image.

[0054] An electronic device includes a memory and a processor, the memory storing a computer program executable on the processor, the processor implementing the steps of a fast electromagnetic wave thermoacoustic imaging denoising method when executing the computer program.

[0055] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of a fast electromagnetic wave thermoacoustic imaging denoising method.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The fast electromagnetic wave thermoacoustic imaging denoising method proposed in this invention vectorizes the multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by the acquisition system, merges and recombines them into a two-dimensional matrix, makes full use of the correlation between the multi-frame electromagnetic wave thermoacoustic signals, and can make full use of the prior information of the signal and noise, thereby achieving a better denoising effect.

[0058] 2. This invention constructs a robust principal component analysis theoretical model based on a two-dimensional matrix, treating the useful signal in the two-dimensional matrix as a low-rank part and the noise or interference in the signal matrix as a sparse part. While utilizing multiple frames of electromagnetic wave thermoacoustic signals, the useful signal is modeled as a low-rank matrix in the spatiotemporal domain, which effectively models the temporal offset of electromagnetic wave thermoacoustic signals caused by biological movement, avoiding the errors that occur in the frame averaging method on moving targets.

[0059] 3. This invention uses robust principal component analysis theory to construct an optimization problem, iteratively solves the optimization problem, and stops iterating when the optimization problem meets the convergence condition. The low-rank matrix and sparse matrix after matrix decomposition are obtained. The alternating direction method is used to construct a convex optimization problem solution function for iterative solution. The symmetric alternating direction augmented Lagrange method and adaptive penalty factor are used to obtain the iterative optimal solution, reducing the computational cost of the iteration process and making the noise removal process extremely time-consuming.

[0060] 4. This invention does not directly perform matrix decomposition on electromagnetic wave thermoacoustic images. Instead, it performs matrix decomposition on multiple frames of the original electromagnetic wave thermoacoustic signals before imaging. This maximizes the preservation of useful signals and noise interference characteristics in the original signals, achieving more efficient denoising. This method can completely remove the interference of electromagnetic waves on the imaging system in electromagnetic wave thermoacoustic imaging, effectively reducing the influence of background noise such as thermal and electronic noise in electromagnetic wave thermoacoustic images, resulting in clearer and more comprehensive electromagnetic wave thermoacoustic images. It solves the problems existing in the prior art and provides important assistance to doctors in interpreting electromagnetic wave thermoacoustic images and diagnosing diseases.

[0061] 5. The rapid electromagnetic wave thermoacoustic imaging denoising device proposed in this invention achieves rapid electromagnetic wave thermoacoustic imaging denoising by dividing the system into an electromagnetic wave thermoacoustic signal acquisition module, a model building module, a rapid robust principal component analysis denoising module, and a real-time electromagnetic wave thermoacoustic imaging module. The modular approach allows each module to work independently, facilitating management. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart of the fast electromagnetic wave thermoacoustic imaging noise reduction method in an embodiment of the present invention;

[0064] Figure 2 This is a schematic diagram of the microwave thermoacoustic imaging results in an embodiment of the present invention;

[0065] Figure 3 This is a schematic diagram of the photoacoustic imaging results in an embodiment of the present invention;

[0066] Figure 4 This is a structural diagram of the fast electromagnetic wave thermoacoustic imaging noise reduction device in an embodiment of the present invention;

[0067] Figure 5 This is a structural diagram of the electronic device in an embodiment of the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0069] Example 1

[0070] Figure 1 This is a flowchart of a fast electromagnetic wave thermoacoustic imaging denoising method according to Embodiment 1 of the present invention, including the following steps:

[0071] S1. Vectorize the multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by the acquisition system and merge and reassemble them into a two-dimensional matrix;

[0072] The multi-frame, multi-channel electromagnetic wave thermoacoustic signal consists of the current frame and n-1 past frames, each frame being an s×T matrix. The merged and recombined two-dimensional matrix...

[0073] Where n represents the number of electromagnetic wave thermoacoustic signal frames used, s represents the number of ultrasonic probes used to collect electromagnetic wave thermoacoustic signals, T represents the number of sampling points of electromagnetic wave thermoacoustic signals, and each column of the M matrix is ​​a vector stretched from each frame of electromagnetic wave thermoacoustic signal matrix.

[0074] S2. Construct a robust principal component analysis theoretical model based on the two-dimensional matrix, treating the useful signal in the two-dimensional matrix as the low-rank part and the noise or interference in the two-dimensional matrix as the sparse part.

[0075] The robust principal component analysis theoretical model constructed based on the two-dimensional matrix M is as follows:

[0076] M = L + S (1)

[0077] Where L represents the useful signal matrix with low-rank characteristics, and S represents the noise matrix with sparse characteristics.

[0078] S3. Construct an optimization problem using robust principal component analysis theory, and iteratively solve the optimization problem. Stop the iteration when the optimization problem satisfies the convergence condition, and obtain the final useful signal matrix L. ＊ The details are as follows:

[0079] This embodiment uses the principal component tracking method to transform the robust principal component analysis problem into a convex optimization problem, as follows:

[0080]

[0081] Where ρ represents the regularization parameter, ||·|| * and ||·||1 represent the nuclear norm and l1 norm of the matrix, respectively. They are the rank and convex relaxation of the l0 norm of the matrix. The l0 norm is the number of non-zero elements in the matrix. st is an abbreviation for subject to, and the part after st represents the constraint condition.

[0082] The unconstrained convex optimization problem is constructed using the augmented Lagrange multiplier method, as follows:

[0083]

[0084] in, Let ||·|| denote the constructed augmented Lagrangian function, Λ denote the Lagrange multiplier, and <·> denote the inner product. F Let Frobenius norm of the matrix be denoted, μ denote the adaptive adjustment penalty factor, and := denote the definition as follows;

[0085] Then, the alternating direction method is used to construct the solution function of the convex optimization problem and solve iteratively. The symmetric alternating direction augmented Lagrangian method and adaptive adjustment of the penalty factor are used to obtain the iterative optimal solution, thereby reducing the computational cost of the iteration process.

[0086] Specifically, (·) k This represents the k-th iteration process; it updates the low-rank characteristic useful signal matrix L of the (k+1)-th iteration. k +1 :

[0087]

[0088] Among them, S k Let Λ represent the sparse property noise matrix of the k-th iteration. k Let μ be the Lagrange multiplier for the k-th iteration. k Let denot be the penalty factor for the k-th iteration. Solve this optimization problem using the singular value threshold operator:

[0089]

[0090] in, Singular value threshold operator:

[0091]

[0092] Where M-S+μΛ=U∑V T Let represent the singular value decomposition of matrix M-S+μΛ, where U represents the left singular vector matrix, ∑ represents the singular value matrix, and V represents the right singular vector matrix. T Z represents the transpose of a matrix. 1 / μ (·) denotes the soft threshold operator:

[0093]

[0094] Where sgn(·) represents the sign function, max represents the maximum value, and x represents each element in the matrix M-S+μΛ; the soft threshold operator operates on each element of the matrix.

[0095] First update on Lagrange multipliers

[0096]

[0097] Update the sparse property noise matrix S in the (k+1)th iteration. k+1 :

[0098]

[0099] The optimization problem is solved using the soft threshold operator:

[0100]

[0101]

[0102] Where sgn(·) represents the sign function, max represents the maximum value, and x′ represents each element in the matrix M-L+μΛ. The soft threshold operator operates on each element of the matrix.

[0103] Second update of Lagrange multipliers Λ k+1 :

[0104]

[0105] Update the penalty factor μ in the (k+1)th iteration. k+1 :

[0106] μ k+1 :=η×μ k (13)

[0107] Where η represents the parameter for adaptively adjusting the penalty factor;

[0108] Iteration relative error The iteration terminates when the value is less than the error limit γ or when the maximum number of iterations K is reached, yielding the final useful signal matrix L. * .

[0109] S4. Extract the last column of the low-rank matrix obtained through iteration, expand it, and use it as the new electromagnetic wave thermoacoustic signal for the current frame; specifically, the new...

[0110] S5. Use the delay superposition algorithm on the new electromagnetic wave thermoacoustic signal matrix to obtain the electromagnetic wave thermoacoustic image.

[0111] Example 2

[0112] Figure 2 This is a schematic diagram of the results of microwave thermoacoustic imaging according to Embodiment 2 of the present invention.

[0113] This embodiment demonstrates real-time monitoring of microwave ablation results using microwave thermoacoustic imaging, wherein the microwave thermoacoustic imaging algorithm employs a delay superposition algorithm. Figure 2 A is the original microwave thermoacoustic image. It can be seen that due to the interference of the continuous microwave source of microwave ablation, the microwave thermoacoustic image shows irregularly distributed interference, the amplitude of which is much greater than the useful signal, making it impossible to accurately monitor the ablation process and accurately assess the ablation lesion. Figure 2b is the image result after denoising by the rapid electromagnetic wave thermoacoustic imaging denoising method described in this invention. It can be seen that the method described in this invention completely removes the interference of the microwave source, making the microwave thermoacoustic image clearly visible, so that doctors can accurately assess the size of the ablation lesion.

[0114] Example 3

[0115] Figure 3 This is a schematic diagram of the photoacoustic imaging result according to Embodiment 3 of the present invention.

[0116] This embodiment shows the results of real-time monitoring of blood vessels in the arm using photoacoustic imaging, wherein the photoacoustic imaging algorithm uses a delay superposition algorithm; Figure 2 A is the original photoacoustic image. It can be seen that there is some electronic noise in the acquisition system. The photoacoustic image shows linear noise, and the interface between tissue and blood vessels is not clear, resulting in a low signal-to-noise ratio. Figure 2 b is the denoised imaging result of the rapid electromagnetic wave thermoacoustic imaging denoising method described in this invention. It can be seen that the method described in this invention removes electronic noise and enhances the signal amplitude, making the interface between tissue and blood vessels clearer and greatly improving the signal-to-noise ratio.

[0117] Example 4

[0118] Figure 4 This is a structural diagram of a fast electromagnetic wave thermoacoustic imaging denoising device according to Embodiment 4 of the present invention, applied to the fast electromagnetic wave thermoacoustic imaging denoising method proposed in this invention, including an electromagnetic wave thermoacoustic signal acquisition module, a model building module, a fast robust principal component analysis denoising module, and a real-time electromagnetic wave thermoacoustic imaging module:

[0119] The electromagnetic wave thermoacoustic signal acquisition module is used to receive multi-channel electromagnetic wave thermoacoustic signals of the current frame in real time, digitally and in parallel acquire multi-channel electromagnetic wave thermoacoustic signals, and save the digital acquisition results in real time.

[0120] The model building module is used to vectorize the multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by the acquisition system, merge and reassemble them into a two-dimensional matrix; construct a robust principal component analysis theoretical model based on the two-dimensional matrix; model the useful signals in the two-dimensional matrix as low-rank parts, and treat the noise or interference in the signal matrix as sparse parts;

[0121] The fast robust principal component analysis denoising module is used to construct an optimization problem using robust principal component analysis theory, iteratively solve the optimization problem, and stop iterating when the optimization problem meets the iterative convergence condition, thereby obtaining the low-rank matrix and sparse matrix after matrix decomposition.

[0122] The real-time electromagnetic wave thermoacoustic imaging module is used to extract the last column of the low-rank matrix obtained by iteration, expand it, and use it as a new electromagnetic wave thermoacoustic signal for the current frame; the imaging algorithm is driven by the real-time electromagnetic wave thermoacoustic imaging device to obtain an electromagnetic wave thermoacoustic image.

[0123] Example 5

[0124] Figure 5 This is a structural diagram of an electronic device according to Embodiment 5 of the present invention.

[0125] The electronic device includes a memory and a processor, the memory storing steps for implementing a fast electromagnetic wave thermoacoustic imaging denoising method on the processor.

[0126] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fast electromagnetic wave thermoacoustic imaging noise reduction method, characterized in that, Includes the following steps: Step 1: Vectorize the multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by the acquisition system, and merge and reassemble them into a two-dimensional matrix M; Step 2: Construct a robust principal component analysis theoretical model based on the two-dimensional matrix. Treat the useful signal in the two-dimensional matrix as the low-rank part and the noise or interference in the two-dimensional matrix as the sparse part, and obtain the useful signal matrix L with low-rank characteristics and the noise matrix S with sparse characteristics, as follows: M = L + S (1) Step 3: Construct the optimization problem using robust principal component analysis theory and solve iteratively. Stop the iteration when the optimization problem satisfies the convergence condition, and obtain the final useful signal matrix L. ＊ ; Step 4: Extract the final useful signal matrix L obtained through iteration. ＊ The last column is expanded and serves as a new electromagnetic wave thermoacoustic signal for the current frame; Step 5: Use an imaging algorithm to obtain an image of electromagnetic wave thermoacoustic sound.

2. The rapid electromagnetic wave thermoacoustic imaging noise reduction method according to claim 1, characterized in that, The multi-frame, multi-channel electromagnetic wave thermoacoustic signal has a current frame and n-1 past frames. Each frame is an s×T matrix, and the merged and recombined two-dimensional matrix... Where n represents the number of electromagnetic wave thermoacoustic signal frames used, s represents the number of ultrasonic probes used in the acquisition system to collect electromagnetic wave thermoacoustic signals, T represents the number of sampling points of electromagnetic wave thermoacoustic signals, and each column of the two-dimensional matrix M is a vector stretched from each frame of electromagnetic wave thermoacoustic signal matrix.

3. The rapid electromagnetic wave thermoacoustic imaging noise reduction method according to claim 2, characterized in that, Step 3 is as follows: The robust principal component analysis problem is transformed into a convex optimization problem using the principal component pursuit method, as follows: Where ρ represents the regularization parameter, ||·|| * and ||·||1 represent the nuclear norm and l1 norm of the matrix, respectively. They are the rank and convex relaxation of the l0 norm of the matrix. The l0 norm is the number of non-zero elements in the matrix. st is an abbreviation for subject to, and the part after st represents the constraint condition. The unconstrained convex optimization problem is constructed using the augmented Lagrange multiplier method, as follows: in, Let ||·|| denote the constructed augmented Lagrangian function, Λ denote the Lagrange multiplier, and <·> denote the inner product. F Let Frobenius norm of the matrix be denoted, μ denote the adaptive adjustment penalty factor, and := denote the definition as follows; Then, the alternating direction method is used to construct the solution function for the convex optimization problem and iteratively solve it. The symmetric alternating direction augmented Lagrangian method and adaptive adjustment of the penalty factor are used to obtain the iterative optimal solution, that is, the final useful signal matrix L. ＊ .

4. The rapid electromagnetic wave thermoacoustic imaging noise reduction method according to claim 3, characterized in that, The iterative solution process is as follows: (·) k This represents the k-th iteration process; it updates the low-rank characteristic useful signal matrix L of the (k+1)-th iteration. k+1 : Among them, S k Let Λ represent the sparse property noise matrix of the k-th iteration. k Let μ be the Lagrange multiplier for the k-th iteration. k Let denot be the penalty factor for the k-th iteration. Solve this optimization problem using the singular value threshold operator: in, Singular value threshold operator: Where M-S+μΛ=U∑V T Let represent the singular value decomposition of matrix M-S+μΛ, where U represents the left singular vector matrix, ∑ represents the singular value matrix, and V represents the right singular vector matrix. T Z represents the transpose of a matrix. 1 / μ (·) denotes the soft threshold operator: Where sgn(·) represents the sign function, max represents the maximum value, and x represents each element in the matrix M-S+μΛ; First update on Lagrange multipliers Update the sparse property noise matrix S in the (k+1)th iteration. k+1 : The optimization problem is solved using the soft threshold operator: Where sgn(·) represents the sign function, max represents the maximum value, and x′ represents each element in the matrix M-L+μΛ. The soft threshold operator operates on each element of the matrix. Second update of Lagrange multipliers Λ k+1 : Update the penalty factor μ in the (k+1)th iteration. k+1 : m k+1 :=h×m k (13) Where η represents the parameter for adaptively adjusting the penalty factor; Iteration relative error The iteration terminates when the error limit γ is less than the maximum number of iterations K, yielding the final useful signal matrix L*.

5. The rapid electromagnetic wave thermoacoustic imaging noise reduction method according to claim 4, characterized in that, The new electromagnetic wave thermoacoustic signal matrix of the current frame in step 4 6. The rapid electromagnetic wave thermoacoustic imaging noise reduction method according to claim 4, characterized in that, The imaging algorithm is an electromagnetic wave thermoacoustic imaging algorithm, specifically a delay superposition algorithm, a back projection algorithm, a time reversal algorithm, a model-based algorithm, and a deep learning algorithm.

7. A rapid electromagnetic wave thermoacoustic imaging noise reduction device, characterized in that, The device is applied to the fast electromagnetic wave thermoacoustic imaging denoising method according to any one of claims 1-6, and includes an electromagnetic wave thermoacoustic signal acquisition module, a model building module, a fast robust principal component analysis denoising module, and a real-time electromagnetic wave thermoacoustic imaging module: The electromagnetic wave thermoacoustic signal acquisition module is used to receive multi-channel electromagnetic wave thermoacoustic signals of the current frame in real time, digitally and in parallel acquire multi-channel electromagnetic wave thermoacoustic signals, and save the digital acquisition results in real time. The model building module is used to vectorize the multi-frame, multi-channel electromagnetic wave thermoacoustic signals acquired by the acquisition system and merge and reassemble them into a two-dimensional matrix. A robust principal component analysis theoretical model is constructed based on a two-dimensional matrix. The useful signal in the two-dimensional matrix is ​​modeled as a low-rank part, and the noise or interference in the signal matrix is ​​regarded as a sparse part. The fast robust principal component analysis denoising module is used to construct an optimization problem using robust principal component analysis theory, iteratively solve the optimization problem, and stop iterating when the optimization problem meets the iterative convergence condition, thereby obtaining the low-rank matrix and sparse matrix after matrix decomposition. The real-time electromagnetic wave thermoacoustic imaging module is used to extract the last column of the low-rank matrix obtained by iteration, expand it, and use it as a new electromagnetic wave thermoacoustic signal for the current frame; the imaging algorithm is driven by the real-time electromagnetic wave thermoacoustic imaging device to obtain an electromagnetic wave thermoacoustic image.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement the steps of the fast electromagnetic wave thermoacoustic imaging denoising method as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the fast electromagnetic wave thermoacoustic imaging denoising method according to any one of claims 1-6.

Images