Synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection

Abstract

The invention relates to a synthetic aperture minimum variance ultrasonic imaging method based on frequency domain segmentation orthogonal projection, which belongs to the technical field of ultrasonic imaging and comprises the following steps: S1: delay processing is performed on the sampling signal received by the ultrasonic array element; S2: the frequency domain segmentation length is selected to convert the ultrasonic echo signal into the sub-band signal in the frequency domain; S3: the sample covariance matrix under the sub-band is obtained; S4: the time-frequency signal subspace under the current sub-band is obtained; S5: the time-frequency orthogonal projection subspace is obtained; S6: the output weight of the beamformer under each sub-band is calculated; S7: the frequency domain weight and the frequency domain output of the frequency domain orthogonal projection minimum variance beamformer are obtained; S8: the frequency domain output value is converted into the time domain output value, low-resolution images are obtained in sequence, and then high-resolution imaging results are obtained by synthesis; the invention can effectively improve the resolution and contrast of ultrasonic images.

Description

Technical Field

[0001] The invention belongs to the technical field of ultrasonic imaging, and relates to a synthetic aperture minimum variance ultrasonic imaging method based on frequency domain segmentation orthogonal projection. Background Art

[0002] The Delay And Sum (DAS) algorithm is the most widely used and simplest beamforming technology in ultrasound imaging. It first calculates the delay of the received wave according to the geometric position relationship between the array elements and the detection target, and then aligns and superimposes the delayed data. The traditional DAS algorithm has low complexity and fast imaging speed, but the main lobe width increases and the resolution is low due to its use of fixed window function weighting.

[0003] In recent years, in order to improve the contrast and resolution of beamforming algorithms, adaptive algorithms have been applied to the field of ultrasound imaging. Among them, the minimum variance (MV) beamforming algorithm is the most widely used adaptive algorithm. This method minimizes the array output energy while keeping the gain in the desired direction unchanged, thereby performing spatial filtering. By dynamically calculating the weighted vector of the focused delay signal and weighted summing the signal, the image resolution can be improved. However, the disadvantage of this algorithm is that the robustness is greatly reduced and it is easy to cause the desired signal to cancel, which has a greater impact on the image quality when the signal-to-noise ratio is low, and there is an obvious lack of contrast.

[0004] Therefore, there is still a lot of room for improvement in algorithm resolution, contrast and robustness based on the minimum variance algorithm. Subsequent scholars used the short-time Fourier transform method to introduce ultrasonic signals into the frequency domain for processing and proposed the frequency domain segmentation minimum variance beamforming algorithm (Minimum Variance Based on Frequency Domain Segmentation, FDSMV). Although the algorithm resolution has been improved to a certain extent, there is still room for further improvement in the performance of frequency domain algorithms. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection. Based on the theory of frequency domain segmented minimum variance algorithm, this method introduces the feature space method into frequency domain ultrasound imaging, and corrects the signal subspace through correlation factors to obtain the time-frequency orthogonal projection subspace, thereby further improving the resolution and contrast of ultrasound images. It effectively overcomes the problem of low contrast of traditional minimum variance beamforming algorithm and improves the overall imaging quality of the image.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection comprises the following steps:

[0008] S1: Delay processing is performed on the sampling signal received by the ultrasonic array element in the synthetic aperture imaging mode to obtain an ultrasonic echo signal;

[0009] S2: Selecting a frequency domain segment length suitable for the ultrasonic echo signal, and using short-time Fourier transform to convert the ultrasonic echo signal into a sub-band signal in the frequency domain;

[0010] S3: performing sub-array spatial smoothing and diagonal loading processing on the sub-band signals of the ultrasonic echo in sequence to obtain the sample covariance matrix under the sub-band;

[0011] S4: performing feature space decomposition on the sample covariance matrix of each time-frequency point under the sub-band to obtain the time-frequency signal subspace under the current sub-band;

[0012] S5: Introduce the eigenvector with the strongest correlation with the frequency domain direction vector for spatial compensation to obtain the time-frequency orthogonal projection subspace;

[0013] S6: Calculate the beamformer output weights in each sub-band according to the minimum variance beamformer principle;

[0014] S7: Projecting the sub-band minimum variance beamformer weights into the time-frequency orthogonal projection subspace to obtain the frequency domain weights of the frequency domain orthogonal projection minimum variance beamformer, and then obtaining the frequency domain output of the beamformer;

[0015] S8: using inverse fast Fourier transform to convert the frequency domain output value of the beamformer into a time domain output value, obtaining low-resolution images in sequence, and then synthesizing the low-resolution images obtained each time to obtain a final high-resolution imaging result.

[0016] Further, the step S2 specifically includes the following steps:

[0017] S21: Use short-time Fourier transform to perform frequency domain segmentation on the ultrasonic echo signal x(k), and the process is shown in the following formula:

[0018]

[0019] Where x(k) represents the ultrasonic echo time domain signal, and x(k) is expressed as x(k)=[x1(k),x2(k),…,x N(k)], where N represents the number of array elements of the ultrasonic array, x1(k) represents the echo signal of array element No. 1, k represents the sampling time corresponding to the sampling depth; z(k) represents a Hanning window with a window length and a short-time Fourier transform point number of 64 and no signal overlap, S(h,ω) represents the corresponding value of the x(k) signal in the time-frequency domain, h represents the time sequence number, ω represents the frequency domain sequence number under the segment sequence number h, i represents the imaginary unit, and e represents the natural constant, which is approximately equal to 2.71828;

[0020] S22: Based on step S21, the ultrasonic echo signal of each sensor array element is sequentially converted into a number of independent equally spaced sub-band signals, and the m-th narrowband sub-signal SP on the n-th aperture n(m) The expression is:

[0021] SP n(m) =[S n (m,1),S n (m,2),...,S n (m,ω),…,S n (m,W-1),S n (m,W)]

[0022] Where m = 1, 2, ..., M, M represents the number of segments into which the aperture signal is divided; ω represents the frequency domain number under the segment number m, ω = 1, 2, ..., W, W is the length of each sub-band signal; M × W represents the total length of the aperture signal in the time domain, S n (m,ω) represents the signal amplitude of the nth aperture at the point (m,ω) in the time-frequency domain.

[0023] Further, step S3 specifically includes the following steps:

[0024] S31: Extract the array information of each time-frequency point. The array signal XS(m,ω) of the time-frequency point (m,ω) is expressed as:

[0025] XS(m,ω)=[S1(m,ω),S2(m,ω)...,S N (m,ω)]

[0026] S32: Divide the array information of each time-frequency point into sub-arrays, and construct a frequency domain sample covariance matrix RS(m,ω) based on the array information after sub-array division:

[0027]

[0028] XS l (m,ω)=[S l (m,ω),S l+1 (m,ω),...,S l+L-1(m,ω)] represents the frequency domain forward smoothing vector of the lth subarray, l = 1, 2, ..., N-L + 1, N represents the total length of the ultrasound array, L represents the subarray length of the ultrasound array, XS l (m,ω) H For XS l The conjugate transpose of (m,ω);

[0029] S33: Perform diagonal loading on the frequency domain sample covariance matrix RS(m,ω) to obtain the covariance matrix after diagonal loading as follows:

[0030]

[0031] Among them, η=trace(RS(m,ω))·α, trace(RS(m,ω)) is the equivalent power of the signal, trace(·) is the function for finding the matrix trace, η is the ratio of spatial noise to signal power, and I is the identity matrix.

[0032] Further, in step S4, the time-frequency signal subspace under the current sub-band is obtained by the following formula:

[0033]

[0034] Among them, FE s (m,w)=[fe1,fe2,...fe i ,...,fe γ ] and FE n (m,w)=[fe γ+1 ,fe γ+2 ,...fe i ,...,fe L ] represent the signal subspace and noise subspace of the echo frequency domain sample covariance matrix at the time-frequency point (m, ω), respectively, fe i is the eigenvector of the frequency domain sample covariance matrix; i represents the eigenvalue of the frequency domain sample covariance matrix, FΛ s (m,w)=[λ1,λ2,...,λ i ,...,λ γ ] represents the vector of eigenvalues ​​of the signal subspace, FΛ n (m,w)=[λ γ+1 ,λ γ+2 ,...,λ i ,...,λ L ] represents the vector of eigenvalues ​​of the noise subspace, (·) H Represents the conjugate transpose operation.

[0035] Further, in step S5, the feature vector fe that is most strongly correlated with the frequency domain direction vector fa is introduced c-max The signal subspace FE corresponding to the current time-frequency point (m, ω) s (m,w) is compensated to obtain the optimized time-frequency orthogonal projection subspace MFE s (m,w), specifically including the following steps:

[0036] S51: Calculate each frequency domain feature vector fe respectively i Correlation coefficient with the fa direction vector in the frequency domain:

[0037]

[0038] Among them, fCS(i) is the i-th frequency domain feature vector fe i The correlation coefficient between fa and ||·||2 represents the vector bi-norm, which is used to normalize the eigenvector modulus, and |·| is the absolute value operation;

[0039] S52:fe c-max It is the eigenvector with the largest correlation coefficient with fa. Adding it into the frequency domain signal subspace, the time-frequency orthogonal projection subspace is obtained as follows:

[0040] MFE s (m,w)=[fe1,fe2,...fe i ,...,fe γ ,fe c-max ]

[0041] Among them, [fe1,fe2,...fe i ,...,fe γ ] is the eigenvector of the original space, [fe c-max ] is the complementary feature vector.

[0042] Further, in step S6, the beamformer output weights of each time-frequency point in each sub-band are calculated according to the minimum variance beamformer principle, as shown below:

[0043]

[0044] Among them, w FDS-MV is the minimum variance weight for frequency domain segmentation in the frequency domain, fa is the direction vector of the frequency domain array with all element values ​​1, fa H is the conjugate transpose of fa, is the covariance matrix The inverse matrix of .

[0045] Further, the step S7 specifically includes the following steps:

[0046] S71: Project the sub-band minimum variance beamformer weights into the time-frequency domain signal subspace to obtain the frequency domain weights of the frequency domain orthogonal projection minimum variance beamformer:

[0047] w STEBMV =MFE s (m,w)MFE s (m,w) H w FDS-MV

[0048] Among them, w STEBMV is the final beamformer weight corresponding to the time-frequency point (m, ω), w FDS-MV MFE is the minimum variance weight of frequency domain segmentation in traditional frequency domain. s (m,w) represents the time-frequency orthogonal projection subspace obtained at the time-frequency point (m,ω), MFE s (m,w) H Is MFE s The conjugate transpose of (m,w);

[0049] S72: Weight the echo data in the sub-bands to obtain the frequency domain output of the final beamformer:

[0050]

[0051] Among them, y STEBMV (m, ω) represents the frequency domain output value of the synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection at the time-frequency domain point (m, ω), w STEBMV represents the composite weight of the synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection, Indicates w STEBMV The conjugate transpose of XS l (m,ω) represents the frequency domain data corresponding to the lth sub-matrix at the time-frequency domain point (m,ω).

[0052] Further, the step S8 specifically includes the following steps:

[0053] S81: Convert the frequency domain output value of the beamformer into a time domain output value using inverse fast Fourier transform:

[0054]

[0055] Among them, y STEBMV(t) (k) represents the time domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at sampling time k during the tth transmission and reception process; ISTFT(·) represents the inverse short-time Fourier transform. STEBMV(m, ω) represents the frequency domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at the time-frequency domain point (m, ω);

[0056] S82: synthesize the low-resolution images obtained in each transmission and reception process to obtain the final high-resolution beamformer imaging result:

[0057]

[0058] Among them, y STEBMV(t) (k) represents the time domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at sampling time k during the t-th transmission and reception process; ISTFT(·) represents the inverse short-time Fourier transform; y STEBMV-ALL (k) represents the final time domain output value of the synthetic aperture mode of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at the sampling time k.

[0059] The beneficial effects of the present invention are as follows: the present invention adopts a synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection. The algorithm first uses short-time Fourier transform to divide the time domain ultrasound signal into short narrowband signals and introduce them into the time-frequency domain, and performs a minimum variance beamforming algorithm in the time-frequency domain; secondly, the feature space method is introduced into the frequency domain ultrasound imaging, and the signal subspace is corrected by the correlation factor to obtain the optimized time-frequency orthogonal projection subspace; finally, the output weights of the frequency domain beamformer are projected and optimized, thereby effectively improving the algorithm resolution and sidelobe suppression capabilities. Finally, the short-time Fourier inverse transform method is used to convert the time-frequency domain output into the time domain for imaging. Therefore, the method proposed by the present invention can further improve the resolution and contrast performance of the traditional ultrasound minimum variance algorithm, and has a higher effect on sidelobe suppression.

[0060] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0062] Figure 1 is a flow chart of the method of the present invention;

[0063] Figure 2The following are the point target imaging effects of four algorithms, among which (a) is the DAS algorithm, (b) is the MV algorithm, (c) is the FDSMV algorithm, and (d) is the STEBMV algorithm;

[0064] Figure 3 The lateral resolution curves of point targets at a depth of 50 mm for the four algorithms are shown below;

[0065] Figure 4 The following are the imaging effects of anechoic cysts simulated by four algorithms, among which (a) is the DAS algorithm, (b) is the MV algorithm, (c) is the FDSMV algorithm, and (d) is the STEBMV algorithm;

[0066] Figure 5 The following is a comparison of the Ats_wires imaging effects of the offline data of four algorithms, where (a) is the DAS algorithm, (b) is the MV algorithm, (c) is the FDSMV algorithm, and (d) is the STEBMV algorithm. DETAILED DESCRIPTION

[0067] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0068] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0069] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0070] like Figure 1 As shown, the present invention provides a synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection, comprising the following steps:

[0071] Step S1: Amplify, AD convert and delay focus the echo signal received by the ultrasonic array element to obtain an ultrasonic echo signal; obtain the signal x(k) after delay focus processing, where x(k) is expressed as x(k)=[x1(k),x2(k),…,x N (k)], where N represents the number of array elements of the ultrasound array, and k represents the sampling time corresponding to the sampling depth;

[0072] Step S2: Select a frequency domain segment length suitable for the ultrasonic echo signal, and use short-time Fourier transform to convert the ultrasonic echo signal into a sub-band signal in the frequency domain. Specifically, the following steps are included:

[0073] S21: Use short-time Fourier transform to perform frequency domain segmentation on the ultrasonic echo signal x(k), and the process is shown in the following formula:

[0074]

[0075] Where x(k) represents the ultrasonic echo time domain signal, and x(k) is expressed as x(k)=[x1(k),x2(k),…,x N (k)], where N represents the number of array elements of the ultrasonic array, x1(k) represents the echo signal of array element No. 1, k represents the sampling time corresponding to the sampling depth; z(k) represents a Hanning window with a window length and a short-time Fourier transform point number of 64 and no signal overlap, S(h,ω) represents the corresponding value of the x(k) signal in the time-frequency domain, h represents the time sequence number, ω represents the frequency domain sequence number under the segment sequence number h, i represents the imaginary unit, and e represents the natural constant, which is approximately equal to 2.71828;

[0076] S22: Based on the previous step, the ultrasonic echo signal of each sensor array element is converted into a number of independent equally spaced sub-band signals in sequence, and the m-th narrowband sub-signal SP on the n-th aperture n(m) The expression is:

[0077] SP n(m) =[S n (m,1),S n (m,2),...,S n (m,ω),…,S n (m,W-1),S n (m,W)]

[0078] Where m = 1, 2, ..., M, M represents the number of segments into which the aperture signal is divided; ω represents the frequency domain number under the segment number m. ω = 1, 2, ..., W, W is the length of each sub-band signal. M × W represents the total length of the aperture signal in the time domain. S n (m,ω) represents the signal amplitude of the nth aperture at the point (m,ω) in the time-frequency domain.

[0079] Step S3: extract the array information of each time-frequency point, perform sub-array spatial smoothing, diagonal loading and other processing on the ultrasonic echo data of the sub-frequency band in turn, and obtain the sample covariance matrix under the sub-frequency band. Specifically, it includes the following steps:

[0080] S31: Extract the array information of each time-frequency point. The array signal XS(m,ω) of the time-frequency point (m,ω) can be expressed as:

[0081] XS(m,ω)=[S1(m,ω),S2(m,ω)…,S N (m,ω)]

[0082] S32: Divide the array information of each time-frequency point into sub-arrays, and construct the frequency domain sample covariance matrix RS(m,ω) based on the array information after the sub-array division:

[0083]

[0084] XS l (m,ω)=[S l (m,ω),S l+1 (m,ω),...,S l+L-1 (m,ω)] represents the frequency domain forward smoothing vector of the lth subarray, and l = 1, 2, ..., N-L+1, N represents the total length of the ultrasound array, and L represents the subarray length of the ultrasound array. XS l (m,ω) H For XS l Conjugate transpose of (m,ω).

[0085] S33: Perform diagonal loading on the frequency domain sample covariance matrix RS(m,ω) to enhance the robustness of the algorithm and obtain the covariance matrix after diagonal loading as follows:

[0086]

[0087] Among them, η=trace(RS(m,ω))·α, trace(RS(m,ω)) is the equivalent power of the signal, trace(·) is the function for finding the matrix trace, η is the ratio of spatial noise to signal power, and I is the identity matrix.

[0088] Step S4: performing feature space decomposition on the sample covariance matrix under the sub-band to obtain the time-frequency signal subspace under the current sub-band;

[0089]

[0090] Among them, FE s (m,w)=[fe1,fe2,...fe i ,...,fe γ ] and FE n (m,w)=[fe γ+1 ,fe γ+2 ,...fe i ,...,fe L ] represent the signal subspace and noise subspace of the echo frequency domain sample covariance matrix at the time-frequency point (m, ω), respectively, fe i is the eigenvector of the frequency domain sample covariance matrix. i represents the eigenvalue of the frequency domain sample covariance matrix, FΛ s (m,w)=[λ1,λ2,...,λ i ,...,λ γ ] represents the vector of eigenvalues ​​of the signal subspace, FΛ n (m,w)=[λ γ+1 ,λ γ+2 ,...,λ i ,...,λ L ] represents the vector of eigenvalues ​​of the noise subspace, (·) H Represents the conjugate transpose operation.

[0091] Step S5: Introduce the feature vector fe that is most strongly correlated with the frequency domain direction vector fa c-max The signal subspace FE corresponding to the current time-frequency point (m, ω) s (m,w) is compensated to obtain the optimized time-frequency orthogonal projection subspace MFE s (m,w):

[0092] S51: Calculate each frequency domain feature vector fe respectively i The fa correlation coefficient with the frequency domain direction vector,

[0093]

[0094] Where fCS(i) is the correlation coefficient between the i-th frequency domain feature vector and fa, ||·||2 represents the vector bi-norm, which is used to normalize the feature vector, and |·| is the absolute value operation.

[0095] S52: The eigenvector with the largest correlation coefficient with fa is fe c-max , and add it into the frequency domain signal subspace to obtain the time-frequency orthogonal projection subspace as follows:

[0096] MFE s (m,w)=[fe1,fe2,...fe i ,...,fe γ ,fe c-max ]

[0097] Among them, [fe1,fe2,...fe i ,...,fe γ ] is the eigenvector of the original space, [fe c-max ] is the complementary feature vector.

[0098] Step S6: Calculate the beamformer output weights of each time-frequency point in each sub-band according to the minimum variance beamformer principle, as shown below:

[0099]

[0100] Among them, w FDS-MV is the minimum variance weight for frequency domain segmentation in the frequency domain, fa is the direction vector of the frequency domain array with all element values ​​1, fa H is the conjugate transpose of fa, is the covariance matrix The inverse matrix of .

[0101] Step S7: Project the sub-band minimum variance beamformer weights into the time-frequency domain signal subspace to obtain the frequency domain weights of the frequency domain orthogonal projection minimum variance beamformer, and weight the echo data in the sub-band to obtain the frequency domain output of the beamformer, as shown in the following formula:

[0102] S71: First, the sub-band minimum variance beamformer weights are projected into the time-frequency domain signal subspace to obtain the frequency domain weights of the frequency domain orthogonal projection minimum variance beamformer:

[0103] w STEBMV =MFE s (m,w)MFE s (m,w) H w FDS-MV

[0104] Among them, w STEBMV is the final beamformer weight corresponding to the time-frequency point (m, ω), w FDS-MV It is the minimum variance weight of frequency domain segmentation in traditional frequency domain. MFE s (m,w) represents the time-frequency orthogonal projection subspace obtained at the time-frequency point (m,ω).

[0105] S72: Secondly, the echo signal is weighted in the sub-band to obtain the frequency domain output of the final beamformer:

[0106]

[0107] Among them, y STEBMV (m, ω) represents the frequency domain output value of the synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection at the time-frequency domain point (m, ω), w STEBMV represents the composite weight of the synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection, and Indicates w STEBMV The conjugate transpose of XS l (m,ω) represents the frequency domain data corresponding to the lth sub-matrix at the time-frequency domain point (m,ω).

[0108] Step S8: Use the inverse fast Fourier transform to convert the frequency domain output value of the beamformer into the time domain output value, obtain low-resolution images in sequence, and then synthesize the low-resolution images obtained in each transmission and reception process to obtain the final high-resolution beamformer imaging result:

[0109] S81: Convert the frequency domain output value of the beamformer into a time domain output value using inverse fast Fourier transform:

[0110]

[0111] Among them, y STEBMV(t) (k) represents the time domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at sampling time k during the tth transmission and reception process. ISTFT(·) represents the inverse short-time Fourier transform. STEBMV (m, ω) represents the frequency domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at the time-frequency domain point (m, ω).

[0112] S82: synthesize the low-resolution images obtained in each transmission and reception process to obtain the final high-resolution beamformer imaging result:

[0113]

[0114] Among them, y STEBMV(t) (k) represents the time domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at sampling time k during the tth transmission and reception process. ISTFT(·) represents the inverse short-time Fourier transform. STEBMV-ALL(k) represents the final time domain output value of the synthetic aperture mode of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at the sampling time k.

[0115] Verification experiment:

[0116] Field II is an ultrasound experiment simulation platform developed by the Technical University of Denmark based on acoustic principles. It has been widely recognized and used in theoretical research. To verify the effectiveness of the proposed algorithm, Field II was used to image the point scattering targets and anechoic cyst simulation models commonly used in ultrasound imaging, as well as the offline data of the actual phantom experiment Ats_wires.

[0117] In the point target simulation experiment, 20 point targets with a double row longitudinal interval of 5mm and a lateral interval of 2mm are set, and the depth distribution is between 40mm and 85mm. The emission fixed-point focusing and receiving dynamic focusing methods are adopted. The emission focus is fixed at 50mm, and the imaging dynamic range of the image is set to 60dB. In the anechoic cyst simulation experiment, an anechoic cyst with a radius of 6mm and a center at (0mm, 40mm) is set. The synthetic aperture focusing method is used, and the imaging dynamic range of the image is set to 60dB. The amplitude ratio of the scattering point between the anechoic cyst and the background is 1000 times. In addition, the center frequency of the array element used in the experiment is 7MHz, the number of array elements is 64, the spacing is 0.24mm, the sampling frequency is 100MHz, and the sound speed is 1540m / s.

[0118] For the above two experimental objectives, comparative imaging experiments were carried out using the delay superposition algorithm (DAS), minimum variance algorithm (MV), frequency domain segmentation minimum variance algorithm (FDSMV), and synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection (STEBMV).

[0119] Figure 2 (a)-(d) in Figure 4 give the point target imaging results of four algorithms. Figure 2 It can be seen that the DAS algorithm has the worst imaging quality, with serious sidelobe artifacts. Compared with other algorithms, it has higher lateral artifacts and wider main lobe width, making it difficult to distinguish target points. The MV algorithm significantly narrows the main lobe width and significantly improves the lateral resolution, but still performs poorly in suppressing sidelobe artifacts.

[0120] Although the FDSMV algorithm further improves the resolution in the entire depth range, artifacts are still visible. The STEBMV algorithm further improves the imaging quality, and the main lobe width and side lobe level are the best among the four algorithms. And its far-field resolution has been significantly improved.

[0121] like Figure 3As shown in the figure, the lateral resolution curves of the four algorithms at the 50mm focus are given. Table 1 gives the main lobe width values ​​of the four imaging algorithms at -6dB in the case of point targets at depths of 40mm, 50mm, and 60mm. It can be seen that the main lobe width of STEBMV is the best, which means it has the best point target resolution. FDSMV is second, and the effect of DAS is poor. Compared with the DAS algorithm, STEBMV has a maximum improvement of 76% in the main lobe width of resolution; compared with the MV algorithm, STEBMV has a maximum improvement of 75% in the main lobe width of resolution, and the resolution improvement is obvious.

[0122] Table 1

[0123]

[0124] like Figure 4 As shown in (a)-(d) in the figure, four algorithms are given to simulate the imaging effect of anechoic cysts. Table 2 shows the comparison of imaging performance indicators of different imaging algorithms for multi-spot imaging simulation. It can be seen that compared with the DAS algorithm, the MV algorithm has a certain improvement in contrast CR, but the signal-to-noise ratio CNR decreases significantly. The improvement of the FDSMV algorithm in CR is only equal to that of the MV, while the CR value of the STEBMV reaches 45.27dB, which is 7.95dB higher than that of the MV algorithm. This shows that the STEBMV can effectively improve the imaging quality of anechoic cyst targets, and its CNR value is also significantly improved compared with the MV algorithm.

[0125] Table 2

[0126]

[0127] like Figure 5 As shown in (a)-(d) in the figure, the imaging effects of four algorithms for offline Ats_wires experimental data are compared. It can be seen that DAS has the lowest resolution, MV has improved resolution compared to DAS, and FDSMV has further improved resolution compared to MV. The STEBMV proposed in the present invention has the best resolution and contrast.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection, characterized in that: The following steps are involved: S1: Delay processing is performed on the sampling signal received by the ultrasonic array element in the synthetic aperture imaging mode to obtain an ultrasonic echo signal; S2: Selecting a frequency domain segment length suitable for the ultrasonic echo signal, and using short-time Fourier transform to convert the ultrasonic echo signal into a sub-band signal in the frequency domain; S3: performing sub-array spatial smoothing and diagonal loading processing on the sub-band signals of the ultrasonic echo in sequence to obtain the sample covariance matrix under the sub-band; S4: performing feature space decomposition on the sample covariance matrix of each time-frequency point under the sub-band to obtain the time-frequency signal subspace under the current sub-band; S5: Introduce the eigenvector with the strongest correlation with the frequency domain direction vector for spatial compensation to obtain the time-frequency orthogonal projection subspace; S6: Calculate the beamformer output weights in each sub-band according to the minimum variance beamformer principle; S7: Projecting the sub-band minimum variance beamformer weights into the time-frequency orthogonal projection subspace to obtain the frequency domain weights of the frequency domain orthogonal projection minimum variance beamformer, and then obtaining the frequency domain output of the beamformer; S8: using inverse fast Fourier transform to convert the frequency domain output value of the beamformer into a time domain output value, obtaining low-resolution images in sequence, and then synthesizing the low-resolution images obtained each time to obtain a final high-resolution imaging result.

2. The method for synthetic aperture minimum variance ultrasound imaging based on frequency domain segmentation orthogonal projection according to claim 1, characterized in that: The step S2 specifically includes the following steps: S21: Use short-time Fourier transform to perform frequency domain segmentation on the ultrasonic echo signal x(k), and the process is shown in the following formula: Where x(k) represents the ultrasonic echo time domain signal, and x(k) is expressed as x(k)=[x1(k),x2(k),…,x N (k)], where N represents the number of array elements of the ultrasonic array, x1(k) represents the echo signal of array element No. 1, k represents the sampling time corresponding to the sampling depth; z(k) represents a Hanning window with a window length and a short-time Fourier transform point number of 64 and no signal overlap, S(h,ω) represents the corresponding value of the x(k) signal in the time-frequency domain, h represents the time sequence number, ω represents the frequency domain sequence number under the segment sequence number h, i represents the imaginary unit, and e represents the natural constant, which is approximately equal to 2.71828; S22: Based on step S21, the ultrasonic echo signal of each sensor array element is sequentially converted into a number of independent equally spaced sub-band signals, and the m-th narrowband sub-signal SP on the n-th aperture n(m) The expression is: SP n(m) =[S n (m,1),S n (m,2),...,S n (m,ω),...,S n (m,W-1),S n (m,W)] Wherein, m=1,2,...,M, M represents the number of segments into which the signal of the nth aperture is divided; ω represents the frequency domain number under the segment number m, ω=1,2,...,W, W is the length of each sub-band signal; M×W represents the total length of the aperture signal in the time domain, S n (m, ω) represents the signal amplitude of the nth aperture at the point (m, ω) in the time-frequency domain.

3. The method for synthetic aperture minimum variance ultrasonic imaging based on frequency domain segmentation orthogonal projection according to claim 1, characterized in that: Step S3 specifically includes the following steps: S31: Extract array information of each time-frequency point. The array signal XS(m, ω) of the time-frequency point (m, ω) is expressed as: XS(m,ω)=[S l (m,ω), S2(m,ω)..., S N (m, ω)] S32: Divide the array information of each time-frequency point into sub-arrays, and construct a frequency domain sample covariance matrix RS(m, ω) based on the array information after sub-array division: XS l (m, ω) = [S l (m,ω),S l+1 (m, ω), ..., S l+L-1 (m, ω)] represents the frequency domain forward smoothing vector of the lth subarray, l = 1, 2, ..., N-L+1, N represents the total length of the ultrasound array, L represents the subarray length of the ultrasound array, XS l (m, ω) H For XS l The conjugate transpose of (m, ω); S33: Perform diagonal loading on the frequency domain sample covariance matrix RS(m, ω) to obtain the covariance matrix after diagonal loading as follows: Among them, η=trace(RS(m,ω))·α, trace(RS(m, ω)) is the equivalent power of the signal, trace(·) is the function for finding the matrix trace, η is the ratio of spatial noise to signal power, and I is the identity matrix.

4. The method for synthetic aperture minimum variance ultrasonic imaging based on frequency domain segmentation orthogonal projection according to claim 1, characterized in that: In step S4, the time-frequency signal subspace under the current sub-band is obtained by the following formula: in, FE s (m,ω)=[fe1,fe2,...fe i , ..., fe γ ] and FE n (m, ω) = [fe γ+1 ,fe γ+2 ,...fe i , ..., fe L ] represent the signal subspace and noise subspace of the echo frequency domain sample covariance matrix at the time-frequency point (m, ω), respectively, fe i is the eigenvector of the frequency domain sample covariance matrix; i represents the eigenvalue of the frequency domain sample covariance matrix, FΛ s (m,ω)=[λ1,λ2,...,λ i , ..., λ γ ] represents the vector of eigenvalues ​​of the signal subspace, FΛ n (m, ω) = [λ γ+1 ,λ γ+2 , ..., λ i , ..., λ L ] represents the vector of eigenvalues ​​of the noise subspace, (·) H Represents the conjugate transpose operation.

5. The method for synthetic aperture minimum variance ultrasonic imaging based on frequency domain segmentation orthogonal projection according to claim 1, characterized in that: In step S5, the eigenvector fe that is most strongly correlated with the frequency domain direction vector fa is introduced c-max The signal subspace FE corresponding to the current time-frequency point (m, ω) s (m, ω) is compensated to obtain the optimized time-frequency orthogonal projection subspace MFE s (m, ω), specifically including the following steps: S51: Calculate each frequency domain feature vector fe respectively i Correlation coefficient with the fa direction vector in the frequency domain: Where fCS(i) is the i-th frequency domain eigenvector fe i The correlation coefficient between fa and ||·||2 represents the vector bi-norm, which is used to normalize the eigenvector modulus, and |·| is the absolute value operation; S52: fe c-max Adding the frequency domain signal subspace, the time-frequency orthogonal projection subspace is obtained as follows: MFE s (m,ω)=[fe1,fe2,...fe i ..., oh γ , yes c-max ] Among them, [fe1, fe2, ...fe i , ..., fe γ ] is the eigenvector of the original space, [fe c-max ] is the complementary feature vector.

6. The method for synthetic aperture minimum variance ultrasound imaging based on frequency domain segmentation orthogonal projection according to claim 1, characterized in that: In step S6, the beamformer output weights of each time-frequency point in each sub-band are calculated according to the minimum variance beamformer principle, as shown below: Among them, w FDS-MV is the minimum variance weight for frequency domain segmentation in the frequency domain, fa is the direction vector of the frequency domain array with all element values ​​1, fa H is the conjugate transpose of fa, is the covariance matrix The inverse matrix of .

7. The method for synthetic aperture minimum variance ultrasonic imaging based on frequency domain segmentation orthogonal projection according to claim 1, characterized in that: The step S7 specifically includes the following steps: S71: Project the sub-band minimum variance beamformer weights into the time-frequency domain signal subspace to obtain the frequency domain weights of the frequency domain orthogonal projection minimum variance beamformer: w STEBMV =MFE s (m,ω)MFE s (m,ω) H w FDS-MV Among them, w STEBMV is the final beamformer weight corresponding to the time-frequency point (m, ω), w FDS-MV MFE is the minimum variance weight of frequency domain segmentation in traditional frequency domain. s (m, ω) represents the time-frequency orthogonal projection subspace obtained at the time-frequency point (m, ω), MFE s (m, ω) H Is MFE s The conjugate transpose of (m, ω); S72: Weight the echo signal in the sub-band to obtain the frequency domain output of the final beamformer: Among them, y STEBMV (m, ω) represents the frequency domain output value of the synthetic aperture minimum variance ultrasound imaging method based on frequency domain segmentation orthogonal projection at the time-frequency domain point (m, ω), Indicates w STEBMV The conjugate transpose of XS l (m, ω) represents the frequency domain data corresponding to the lth subarray at the time-frequency domain point (m, ω); N represents the total length of the ultrasonic array, and L represents the subarray length of the ultrasonic array.

8. The method for synthetic aperture minimum variance ultrasound imaging based on frequency domain segmentation orthogonal projection according to claim 1, characterized in that: The step S8 specifically includes the following steps: S81: Convert the frequency domain output value of the beamformer into a time domain output value using inverse fast Fourier transform: Among them, y STEBMV(t) (k) represents the time domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at sampling time k during the t-th transmission and reception process; ISTFT(·) represents the inverse short-time Fourier transform; y STEBMV (m, ω) represents the frequency domain output value of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at the time-frequency domain point (m, ω); S82: synthesize the low-resolution images obtained in each transmission and reception process to obtain the final high-resolution beamformer imaging result: Among them, y STEBMV-ALL (k) represents the final time domain output value of the synthetic aperture mode of the synthetic aperture minimum variance ultrasound imaging beamformer based on frequency domain segmentation orthogonal projection at the sampling time k.

Images