Ultrasound microvascular imaging method based on rank decomposition filtering

Abstract

Provided in the present application is an ultrasound microvascular imaging method based on rank decomposition filtering, the method comprising: step one, acquiring ultrasound ultrafast plane wave data; step two, reading the number of decomposed rank-1 tensors, and performing CP decomposition on the ultrasound ultrafast plane wave data; step three, searching for a curvature inflection point for a rank-n tensor curve obtained by means of CP decomposition, determining whether the curvature inflection point meets a preset limiting condition, and if the curvature inflection point meets the preset limiting condition, executing step four, otherwise, executing step two; step four, truncating a low-rank signal before the curvature inflection point in the rank-n tensor curve, and reconstructing IQ signal data; and step five, filtering the IQ signal data, and performing fixed-lag correlation calculation on the filtered IQ signal data, so as to obtain an ultrasound Doppler blood flow signal. In the present application, by using a matrix decomposition method, three-dimensional data is converted into a tensor, and rank decomposition is directly performed; and a low rank obtained by means of decomposition is truncated to filter out noise, and ultrasound blood flow information is obtained by means of fixed-lag correlation calculation.

Description

A method for ultrasonic micro-blood flow imaging based on rank decomposition filtering

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to Chinese patent application number 2023109588056, filed with the China Patent Office on August 1, 2023, entitled “A method for ultrasonic micro-blood flow imaging based on rank decomposition filtering,” the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of ultrasonic imaging technology, and in particular to an ultrasonic micro-blood flow imaging method based on rank decomposition filtering. Background Art

[0004] Ultrasound micro-flow imaging is a high-resolution ultrasound technique that can be used to assess blood flow in tiny blood vessels within human tissue. It provides detailed information about blood flow velocity, volume, and direction within microvessels, which is crucial for the diagnosis and treatment of many diseases. Combined with ultrasound's advantages of being non-invasive, non-invasive, and radiation-free, ultrasound micro-flow imaging has gained widespread clinical application. Common clinical applications of ultrasound micro-flow imaging include tumor assessment, cardiovascular disease detection, providing detailed information about cardiac function and myocardial contraction, neurological applications, and assessment of human microcirculation. Overall, ultrasound micro-flow imaging technology has been applied in many medical fields, including oncology, cardiology, neurology, and microcirculation. It provides detailed blood flow information, helping doctors more accurately diagnose diseases, formulate treatment plans, and monitor treatment effectiveness.

[0005] Currently, the most widely used filtering method in ultrasound micro-blood flow imaging is singular value decomposition (SVD). Unlike the infinite impulse response (IIR) or finite impulse response (FIR) filtering in traditional ultrasound blood flow imaging technology, the SVD filter does not convert the time domain signal to the frequency domain for frequency blocking, but instead combines the time and space information of the signal into the product of three matrices, that is, a matrix A is represented as the product of U, Σ and V, where U and V are orthogonal matrices and Σ is a diagonal matrix. The features in the diagonal matrix are used to extract singular values ​​and denoise the signal. The singular value decomposition formula is as follows: A = UΣV T

[0006] Among them, U is a matrix of scale mxm; Σ is a matrix of scale mxn, with all elements except the elements on the main diagonal being 0, and each element on the main diagonal is called a singular value; V is a matrix of scale nxn, as shown in Figure 1.

[0007] Existing Chinese patent CN115969413A discloses an ultrasonic color flow filtering method based on singular value decomposition. This method uses SVD filter decomposition and reconstruction to compare with the original signal to set different thresholds for judgment, dynamically selecting the SVD filtering range. Compared to traditional color signal processing methods, this method can adaptively determine the most appropriate target threshold based on real-time signal characteristics, more accurately filtering out clutter and improving blood flow signal accuracy.

[0008] In addition, for multi-angle plane wave ultrasound micro-blood flow imaging, it is necessary to process the ultrasound sequence data of a specific time length to convert it into a spatiotemporal joint matrix, and convert the three-dimensional matrix into a two-dimensional matrix for SVD filtering. However, when a large number of data points are collected, this dimensional conversion and SVD filtering will consume a considerable amount of time, which poses a major obstacle to improving the real-time frame rate in ultrasound ultrafast sequences. Common signal processing optimization methods generally use block sampling or random sampling combined with SVD to improve the calculation speed of this step. Existing patent CN112914611A discloses an ultrasound microvascular imaging method based on random singular value decomposition and weighted L1 norm. It uses random singular value decomposition to randomly extract the spatiotemporal joint reconstruction matrix and then construct a sparse matrix to increase the external weight of the microvessels. The projected gradient descent method is then used to optimize the cost function to obtain ultrasound microvascular imaging results with the addition of a total variation regularizer. This method can increase the external weight of microvessels to better suppress the external noise of microvessels, improve the signal to noise ratio (SNR) of microvascular imaging, and improve the calculation speed.

[0009] It can be seen that the methods used in the existing technology have defects such as high implementation complexity and inaccurate blood flow information.

[0010] Summary of the Invention

[0011] One of the purposes of this application is to provide an ultrasonic micro-blood flow imaging method based on rank decomposition filtering. In order to solve the problem of long time-consuming SVD filtering processing in ultrafast imaging duplex mode, the three-dimensional data is converted into a tensor and directly rank decomposition is performed by adopting the matrix decomposition method. The low rank decomposed is truncated to filter out noise, and the ultrasonic blood flow information is obtained through fixed delay correlation calculation.

[0012] The present application provides an ultrasonic micro-blood flow imaging method based on rank decomposition filtering, comprising:

[0013] Step 1: Acquire ultrasonic ultrafast plane wave data;

[0014] Step 2: Read the number of decomposed rank 1 tensors and perform CP decomposition on the ultrasonic ultrafast plane wave data;

[0015] Step 3: Find the curvature inflection point for the rank n tensor curve decomposed by CP; determine whether the curvature inflection point meets the preset constraint conditions. If yes, execute step 4; otherwise, execute step 2.

[0016] Step 4: truncating the low-rank signal before the curvature inflection point in the rank n tensor curve, and reconstructing IQ signal data;

[0017] Step 5: Filter the IQ signal data, and perform fixed-delay correlation calculation on the filtered IQ signal data to obtain an ultrasonic Doppler blood flow signal.

[0018] [Corrected 26.07.2024 according to Rule 91] Optionally, the dimensions of the ultrasonic ultrafast plane wave data are m x n x t, where m is the number of longitudinal sampling points, n is the number of transverse sampling points, and t is the data length in the slow-time direction.

[0019] Optionally, the conversion formula between the measured Doppler velocity and the transmit pulse repetition frequency in the t direction after multi-angle composite slow time is:

[0020] Among them, PRF is the repetition frequency of the transmitted pulse in the slow time t direction after multi-angle compounding, Scale max is the user's measurement range of the Doppler signal of fine blood flow, and λ is the emission wavelength.

[0021] Optionally, in step 2, the number of rank-1 tensors is set to the total length in the slow-time t direction, which is between [20, 128].

[0022] Optionally, in step 2, CP decomposition decomposes the tensor into a sum of rank 1 tensors, which can be expressed as:

[0023] in, Indicates the outer product, χ is a tensor, a r 、b r 、c r Represent three vectors of three dimensions respectively, R is the rank in CP decomposition, and the three-dimensional tensor can be regarded as a r 、b r 、c r The sum of outer products at different ranks;

[0024] At the same time, the tensor is matrixed, all a vectors are combined into an A matrix for operation, and A = [a1, a2, ..., aR] is obtained, and then extended to the B and C matrices; the matrixed tensor CP decomposition is expressed as: X (1) ≈A(B☉C) T X (2) ≈B(C☉A) T X (3) ≈C(B☉A) T

[0025] Where ⊙ represents the Khatri-Rao product, X (i) The matrix representing the matrix after the tensor χ is matrixed modulo i;

[0026] [Corrected 26.07.2024 according to Rule 91] After CP decomposition, assuming that the column vectors of matrices A, B and C are normalized vectors, the extracted weights are merged into the vector λ∈R. The CP decomposition expression is:

[0027] Optionally, in step three, when searching for the curvature inflection point of the rank n tensor curve decomposed by CP, the extracted weight is incorporated into the vector λ and the vector curve is extended from the high-rank interval to the low-rank interval to perform least squares curve fitting to find the curvature inflection point.

[0028] Optionally, in step three, the restriction condition is that the minimum value of the curvature inflection point coordinate index is greater than or equal to 2.

[0029] Optionally, truncating the low-rank signal before the curvature inflection point in the rank-n tensor curve and reconstructing IQ signal data includes:

[0030] The vector λ before the curvature inflection point in the rank n tensor curve is truncated to 0, and the IQ signal data is reconstructed according to the CP decomposition expression.

[0031] [Corrected 26.07.2024 according to Rule 91] Optionally, the dimensions of the IQ signal data are mxnxt.

[0032] Optionally, when performing fixed delay correlation calculation on the filtered IQ signal data, a calculation formula is:

[0033] Wherein, WFIQ represents the filtered IQ data, EL represents the length of the ensemble, which is the length t in the slow-time direction in step 1; i and j represent the row and column coordinates in space, respectively; R0 represents zero-delay correlation, R1 represents fixed-delay correlation, and the fixed delay is taken as 1; * is the conjugate complex number symbol; the inverse tangent function calculation and velocity conversion are performed on the obtained R1 to obtain the Doppler blood flow velocity micro-flow image, and the modulus and threshold compression of R0 are performed to obtain the Doppler blood flow intensity signal micro-flow image.

[0034] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present application. The purposes and other advantages of the present application can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0035] The technical solution of the present application is further described in detail below through the accompanying drawings and examples. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings are used to provide a further understanding of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings:

[0037] FIG1 is a flow chart of an ultrasonic micro-blood flow imaging method based on rank decomposition filtering provided in an embodiment of the present application;

[0038] [Corrected 26.07.2024 according to Rule 91] FIG2 is a schematic diagram of the SVD decomposition of the matrix A of scale M x N in an embodiment of the present application;

[0039] FIG3 is a schematic diagram of CP decomposition of the tensor χ in an embodiment of the present application;

[0040] FIG4 is a curve of different length ranks N in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The preferred embodiments of the present application are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application and are not used to limit the present application.

[0042] Please refer to Figures 1 and 2 in conjunction, wherein Figure 1 is a flow chart of a method for ultrasonic micro-blood flow imaging based on rank decomposition filtering provided in an embodiment of the present application, and Figure 2 is a schematic diagram of the SVD decomposition of an M×N matrix A in an embodiment of the present application. The method specifically includes the following steps:

[0043] Step 1: Acquire ultrasonic ultrafast plane wave data;

[0044] Step 2: Read the number of decomposed rank 1 tensors and perform CP decomposition on the ultrasonic ultrafast plane wave data;

[0045] Step 3: Find the curvature inflection point for the rank n tensor curve decomposed by CP; determine whether the curvature inflection point meets the preset constraint conditions. If yes, execute step 4; otherwise, execute step 2.

[0046] Step 4: truncating the low-rank signal before the curvature inflection point in the rank n tensor curve, and reconstructing IQ signal data;

[0047] Step 5: Filter the IQ signal data, and perform fixed-delay correlation calculation on the filtered IQ signal data to obtain an ultrasonic Doppler blood flow signal.

[0048] [Corrected 26.07.2024 according to Rule 91] The dimensions of the ultrasonic ultrafast plane wave data are mxnxt, where m is the number of longitudinal sampling points, n is the number of transverse sampling points, and t is the data length in the slow-time direction.

[0049] The conversion formula between the measured Doppler velocity and the transmitted pulse repetition frequency in the t direction after multi-angle composite slow time is:

[0050] Among them, PRF is the repetition frequency of the transmitted pulse in the slow time t direction after multi-angle compounding, Scale max is the user's measurement range of the Doppler signal of fine blood flow, and λ is the emission wavelength.

[0051] In the step 2, the number of rank 1 tensors is set to the total length in the slow time t direction, which is between [20, 128].

[0052] In step 2, CP decomposition decomposes the tensor into a sum of rank 1 tensors, which can be expressed as:

[0053] in, Indicates the outer product, χ is a tensor, a r 、b r 、c r Represent three vectors of three dimensions respectively, R is the rank in CP decomposition, and the three-dimensional tensor can be regarded as a r 、b r 、c r The sum of outer products at different ranks;

[0054] Matrix the tensor, merge all a vectors into an A matrix for operation, and get A = [a1, a2, ..., aR], and expand it to the B and C matrices; the matrixed tensor CP decomposition is expressed as: X (1) ≈A(B☉C) TX (2) ≈B(C☉A) T X (3) ≈C(B☉A) T

[0055] Where ⊙ represents the Khatri-Rao product, X (i) The matrix representing the matrix after the tensor χ is matrixed modulo i;

[0056] After CP decomposition, assuming that the column vectors of matrices A, B, and C are normalized vectors, the extracted weights are merged into the vector λ∈R. The CP decomposition expression is:

[0057] In the step three, when searching for the curvature inflection point of the rank n tensor curve decomposed by CP, the extracted weight is incorporated into the vector λ and the vector curve is extended from the high-rank interval to the low-rank interval to perform least squares curve fitting to find the curvature inflection point.

[0058] In step three, the restriction condition is that the minimum value of the curvature inflection point coordinate index is greater than or equal to 2.

[0059] Step 4: truncating the low-rank signal before the curvature inflection point in the rank n tensor curve and reconstructing IQ signal data, including:

[0060] The vector λ before the curvature inflection point in the rank n tensor curve is truncated to 0, and the IQ signal data is reconstructed according to the CP decomposition expression.

[0061] [Corrected 26.07.2024 according to Rule 91] The dimensions of the IQ signal data are both mxnxt.

[0062] The calculation formula for performing fixed delay correlation calculation on the filtered IQ signal data is:

[0063] Wherein, WFIQ represents the filtered IQ data, EL represents the length of Ensemble, which is the length t in the slow-time direction in step 1; i and j represent the row and column coordinates in space, respectively; R0 represents zero-delay correlation, R1 represents fixed-delay correlation, and the fixed delay is taken as 1; * is the conjugate complex number symbol; the arc tangent function calculation and velocity conversion of the obtained R1 can obtain the Doppler blood flow velocity micro-flow image, and the modulus and threshold compression of R0 can obtain the Doppler blood flow intensity signal micro-flow image.

[0064] The following is an introduction to the specific implementation method of the ultrasonic micro-blood flow imaging method based on rank decomposition filtering provided in the embodiment of the present application:

[0065] [Corrected 26.07.2024 according to Rule 91] Acquire multi-angle composited ultrasonic ultrafast plane wave data with dimensions m x n x t, where m is the number of longitudinal sampling points, n is the number of transverse sampling points, and t is the data length in the slow time direction. For the multi-angle composited transmitted pulse repetition frequency (PRF) in the slow time t direction, the conversion formula between the measured Doppler velocity and PRF is:

[0066] Among them, Scale max is the user's measurement range of the Doppler signal of fine blood flow, and λ is the emission wavelength.

[0067] The number of decomposed rank-1 tensors is read by the preset parameters, and CP decomposition is performed on the three-dimensional data.

[0068] The number of preset rank-1 tensors is set to the total length of the slow-time t direction, which is generally between [20, 128]. CP decomposition was first proposed by FL Hitchcock in The Expression of a Tensor or a Polyadic as a Sum of Products, FL Hitchcock, 1927, Journal of Mathematics and Physics. Its essence is to decompose a tensor into a sum of rank-1 tensors, which can be expressed as:

[0069] in, Indicates the outer product, χ is a tensor, a r 、b r 、c r Represent three vectors of three dimensions respectively, R is the rank in CP decomposition, and the three-dimensional tensor can be regarded as a r 、b r 、c r The sum of outer products at different ranks. The schematic diagram of CP decomposition is shown in Figure 3.

[0070] At the same time, the tensor can be matrixed, and all a vectors can be combined into an A matrix for operation, and A=[a1,a2,…,aR]. This theory can be extended to B and C matrices. Therefore, the CP decomposition of the matrixed tensor can be expressed as: X (1) ≈A(B☉C) T X (2) ≈B(C☉A) T X (3) ≈C(B☉A) T

[0071] Where ⊙ represents the Khatri-Rao product, X (i) The matrix representing the matrix after the tensor χ is matrixed modulo i.

[0072] Usually after CP decomposition, it is assumed that the column vectors of matrices A, B, and C are normalized vectors, and the extracted weights are merged into the vector λ∈R. Finally, the CP decomposition expression can be:

[0073] Find the curvature inflection point for the decomposed rank n tensor curve. If the curvature inflection point does not meet the preset constraints, extend the number of rank n tensors and return to repeat step 2.

[0074] The extracted weights are incorporated into the vector λ, and the vector curve is extended from the high-rank interval to the low-rank interval to perform least square curve fitting to find the curvature inflection point. The minimum value of the curvature inflection point coordinate index is greater than or equal to 2. If the restriction is not met, the length k is increased to obtain the CP decomposition rank n+k, and step 2 is repeated until the curvature inflection point coordinate index p is found and the restriction is met. In this application, k can be 12. Figure 4 describes rank N curves of different lengths, namely N=10, N=24 and N=48, and the vertical axis is the amplitude of the extracted weights.

[0075] [Corrected 26.07.2024 according to Rule 91] The low-rank signal before the selected curvature inflection point is truncated and the three-dimensional IQ signal is reconstructed to obtain data with the same dimension of mxnxt.

[0076] The vector λ before the curvature inflection point is truncated to 0, and the three-dimensional IQ signal is reconstructed according to the above CP decomposition expression.

[0077] Perform fixed delay correlation calculation on the filtered IQ data to obtain the ultrasonic Doppler blood flow signal. The delay correlation calculation method is as follows:

[0078] Wherein, WFIQ represents the IQ data after filtering, and EL represents the length of Ensemble (i.e., the length of t in the slow-time direction in step one). i and j represent the row and column coordinates in space, respectively. R0 represents zero-delay correlation, and R1 represents fixed-delay correlation. In this application, the fixed delay is taken as 1. * is a conjugate complex number symbol. By performing inverse tangent function calculation and velocity conversion on the obtained R1, a Doppler blood flow velocity micro-blood flow image can be obtained. By performing modulus and threshold compression on R0, a Doppler blood flow intensity signal micro-blood flow image can be obtained. In summary, the ultrasonic micro-blood flow imaging method based on rank decomposition filtering provided in this embodiment solves the problem of long time-consuming SVD filtering processing in the ultrafast imaging duplex mode. By adopting a matrix decomposition method, the three-dimensional data is converted into a tensor and directly rank decomposition is performed. The decomposed low rank is truncated and filtered to remove noise, and the ultrasonic blood flow information is obtained by fixed-delay correlation calculation. Compared with the prior art, the implementation complexity is reduced, and the accuracy of the obtained ultrasonic blood flow information is improved.

[0079] In addition, an embodiment of the present application further provides an ultrasonic micro-blood flow imaging device based on rank decomposition filtering, which can execute the method described in any of the above embodiments. The device may include an acquisition module, a decomposition module, a detection module, a reconstruction module, and a calculation module.

[0080] an acquisition module configured to acquire ultrasonic ultrafast plane wave data;

[0081] a decomposition module configured to read the number of decomposed rank-1 tensors and perform CP decomposition on the ultrasonic ultrafast plane wave data;

[0082] a detection module configured to search for a curvature inflection point of the rank n tensor curve decomposed by the CP, and determine whether the curvature inflection point satisfies a preset constraint condition;

[0083] a reconstruction module configured to, when the detection result of the detection module is yes, cut off the low-rank signal before the curvature inflection point in the rank-n tensor curve and reconstruct the IQ signal data;

[0084] The calculation module is configured to filter the IQ signal data and perform fixed-delay correlation calculation on the filtered IQ signal data to obtain an ultrasonic Doppler blood flow signal.

[0085] In a possible implementation, the dimension of the ultrasonic ultrafast plane wave data is m×n×t, where m is the number of longitudinal sampling points, n is the number of transverse sampling points, and t is the data length in the slow-time direction.

[0086] In one possible implementation, the conversion formula between the measured Doppler velocity and the transmit pulse repetition frequency in the t direction after multi-angle composite slow time is:

[0087] Among them, PRF is the repetition frequency of the transmitted pulse in the slow time t direction after multi-angle compounding, Scale max is the user's measurement range of the Doppler signal of fine blood flow, and λ is the emission wavelength.

[0088] The technical effects of the ultrasonic micro-blood flow imaging device based on rank decomposition filtering provided in this embodiment can refer to the technical effects of the methods described in any of the above implementation modes, and will not be repeated here.

[0089] In addition, embodiments of the present application further provide an electronic device, which may be a computer device, a server, or the like, and includes a storage medium and a processor. The storage medium and processor components are electrically connected to each other, directly or indirectly, to enable data transmission or interaction. For example, these components may be electrically connected to each other via one or more communication buses or signal lines.

[0090] The processor is used to read / write data or programs stored in the storage medium and execute the ultrasonic micro-blood flow imaging method based on rank decomposition filtering provided by any embodiment of the present invention.

[0091] The technical effects of the electronic device described in this embodiment can refer to the technical effects of the methods described in any of the above implementation methods, and will not be repeated here.

[0092] An embodiment of the present application further provides a computer-readable storage medium storing machine-executable instructions. When the machine-executable instructions are executed, the ultrasonic micro-blood flow imaging method based on rank decomposition filtering provided in the above embodiment is implemented.

[0093] The technical effects of the computer-readable storage medium described in this embodiment can refer to the technical effects of the methods described in any of the above implementation methods, and will not be repeated here.

[0094] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations. Industrial Applicability

[0095] In summary, the present application provides an ultrasonic micro-blood flow imaging method based on rank decomposition filtering. Through the matrix decomposition method, the three-dimensional data is converted into a tensor and directly rank decomposition is performed. The decomposed low-rank is truncated and filtered to remove noise, and the ultrasonic blood flow information is obtained through fixed delay correlation calculation, which reduces the implementation complexity and improves the accuracy of the obtained ultrasonic blood flow information.

Claims

1. A method for ultrasonic micro-blood flow imaging based on rank decomposition filtering, characterized in that: include: Step 1: Acquire ultrasonic ultrafast plane wave data; Step 2: reading the number of decomposed rank 1 tensors, and performing CP decomposition on the ultrasonic ultrafast plane wave data; Step 3: Find the curvature inflection point for the rank n tensor curve decomposed by CP; determine whether the curvature inflection point meets the preset restriction condition, if yes, execute step 4, otherwise, execute step 2; Step 4: truncating the low-rank signal before the curvature inflection point in the rank n tensor curve, and reconstructing IQ signal data; Step 5: Filter the IQ signal data, and perform fixed-delay correlation calculation on the filtered IQ signal data to obtain an ultrasonic Doppler blood flow signal.

2. [Corrected 26.07.2024 in accordance with Rule 91] A method for ultrasonic micro-blood flow imaging based on rank decomposition filtering as claimed in claim 1, characterized in that: The dimension of the ultrasonic ultrafast plane wave data is mxnxt, where m is the number of longitudinal sampling points, n is the number of transverse sampling points, and t is the data length in the slow time direction.

3. The method for ultrasonic micro-blood flow imaging based on rank decomposition filtering as claimed in claim 2, characterized in that: The conversion formula between the measured Doppler velocity and the transmission pulse repetition frequency in the t direction after multi-angle composite slow time is: Among them, PRF is the repetition frequency of the transmitted pulse in the slow time t direction after multi-angle compounding, Scale max is the user's measurement range of the Doppler signal of fine blood flow, and λ is the emission wavelength.

4. The method for ultrasonic micro-blood flow imaging based on rank decomposition filtering according to any one of claims 1 to 3, characterized in that: In step 2, the number of rank 1 tensors is set to the total length in the slow time t direction, which is between [20, 128].

5. [Corrected 26.07.2024 in accordance with Rule 91] A method for ultrasonic micro-blood flow imaging based on rank decomposition filtering as described in any one of claims 1 to 4, characterized in that: In step 2, CP decomposition decomposes the tensor into the sum of rank 1 tensors, which can be expressed as: in, Indicates the outer product, χ is a tensor, a r , b r 、c r Represent three vectors of three dimensions respectively, R is the rank in CP decomposition, and the three-dimensional tensor can be regarded as a r , b r 、c r The sum of outer products at different ranks; Matrix the tensor, merge all a vectors into an A matrix for operation, and get A = [a1, a2, ..., aR], which is extended to the B and C matrices; the matrixed tensor CP decomposition is expressed as: X (1) ≈A(B☉C) T X (2) ≈B(C☉A) T X (3) ≈C(B☉A) T Where ⊙ represents the Khatri-Rao product, X (i) The matrix representing the matrix after the tensor χ is matrixed modulo i; After CP decomposition, assuming that the column vectors of matrices A, B, and C are standardized vectors, the extracted weights are merged into the vector λ∈R, and the CP decomposition expression is:

6. The method for ultrasonic micro-blood flow imaging based on rank decomposition filtering according to claim 5, characterized in that: In the step three, when searching for the curvature inflection point of the rank n tensor curve decomposed by CP, the extracted weight is incorporated into the vector λ and the vector curve is extended from the high-rank interval to the low-rank interval to perform least squares curve fitting to find the curvature inflection point.

7. The method for ultrasonic micro-blood flow imaging based on rank decomposition filtering according to any one of claims 1 to 6, characterized in that: In step three, the restriction condition is that the minimum value of the curvature inflection point coordinate index is greater than or equal to 2.

8. The method for ultrasonic micro-blood flow imaging based on rank decomposition filtering according to claim 5, characterized in that: The step of truncating the low-rank signal before the curvature inflection point in the rank n tensor curve and reconstructing IQ signal data includes: The vector λ before the curvature inflection point in the rank n tensor curve is truncated to 0, and the IQ signal data is reconstructed according to the CP decomposition expression.

9. [Corrected 26.07.2024 in accordance with Rule 91] A method for ultrasonic micro-blood flow imaging based on rank decomposition filtering as described in any one of claims 1 to 8, characterized in that: The dimension of the IQ signal data is mxnxt.

10. The method for ultrasonic micro-blood flow imaging based on rank decomposition filtering according to any one of claims 2 to 9, characterized in that: The calculation formula for performing fixed delay correlation calculation on the filtered IQ signal data is: Wherein, WFIQ represents the IQ data after filtering, EL represents the length of Ensemble, that is, the length t in the slow time direction in step one); i and j represent the row and column coordinates in space respectively; R0 represents zero delay correlation, R1 represents fixed delay correlation, and the fixed delay is taken as 1; * is the conjugate complex number symbol; the inverse tangent function calculation and velocity conversion are performed on the obtained R1 to obtain the Doppler blood flow velocity micro-blood flow image, and the R0 is modulo-calculated and threshold-compressed to obtain the Doppler blood flow intensity signal micro-blood flow image.

11. An ultrasonic micro blood flow imaging device based on rank decomposition filtering, characterized in that: The device comprises: an acquisition module configured to acquire ultrasonic ultrafast plane wave data; a decomposition module configured to read the number of decomposed rank-1 tensors and perform CP decomposition on the ultrasonic ultrafast plane wave data; A detection module is configured to find a curvature inflection point for the rank n tensor curve decomposed by CP, and determine whether the curvature inflection point meets a preset restriction condition; a reconstruction module, configured to, when the detection result of the detection module is yes, cut off the low-rank signal before the curvature inflection point in the rank n tensor curve, and reconstruct the IQ signal data; The calculation module is configured to filter the IQ signal data and perform fixed delay correlation calculation on the filtered IQ signal data to obtain an ultrasonic Doppler blood flow signal.

12. The ultrasonic micro blood flow imaging device based on rank decomposition filtering according to claim 11, characterized in that: The dimension of the ultrasonic ultrafast plane wave data is m×n×t, where m is the number of longitudinal sampling points, n is the number of transverse sampling points, and t is the data length in the slow time direction.

13. The ultrasonic micro blood flow imaging device based on rank decomposition filtering according to claim 11, characterized in that: The conversion formula between the measured Doppler velocity and the transmission pulse repetition frequency in the t direction after multi-angle composite slow time is: Among them, PRF is the repetition frequency of the transmitted pulse in the slow time t direction after multi-angle compounding, Scale max is the user's measurement range of the Doppler signal of fine blood flow, and λ is the emission wavelength.

14. An electronic device, characterized in that: The electronic device comprises one or more storage media and one or more processors communicating with the storage media, wherein the one or more storage media store machine executable instructions executable by the processors, and when the electronic device is running, the processor executes the machine executable instructions to execute the method described in any one of claims 1 to 10.

15. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are executed by a processor, the method according to any one of claims 1 to 10 is implemented.

Images