CN102835975A - MV (Minimum variance) wave beam formation and MV-based CF (correlation factor) fusion method - Google Patents

Abstract

The invention discloses an MV (minimum variance) wave beam formation and MV-based CF (correlation factor) fusion method. The method comprises the following steps of: constructing an MV constraint matrix by utilizing a direction vector corresponding to a group of maximal value output by a traditional wave beam forming device; acquiring the output of an MV wave beam formation by an MV wave beam forming algorithm, and replacing the correlation part of the CF by the output formed by the MV wave beam so as to form a CF with high resolution; and weighting the correction coefficient with high resolution and the output formed by the MV wave beam, thereby obtaining an output model formed by the wave beam finally. According to the invention, by utilizing the high resolution of MV, the correlation part of CF is replaced by the output formed by the MV wave beam so as to form the CF with high resolution, and the CF with high resolution and the output formed by the MV wave beam are subjected to weighting and fusion; and the simulation result verification of a point scattering target and a sound absorption target shows that compared with the prior art, the method provided by the invention can improve the resolution, contrast and robustness of images well.

Description

The minimum variance wave beam forms and the method that merges based on the minimum variance coherence factor

Technical field

The present invention relates to diagnose the ultra sonic imaging of usefulness, particularly a kind of minimum variance wave beam forms and the method that merges based on the minimum variance coherence factor.

Background technology

The minimum variance wave beam forms (MV) and is proposed in 1969 by Capon the earliest, and its basic thought is to make noise and from the power minimum that any interference on the non-information source direction is contributed, keep the signal power on the information source direction constant simultaneously.But this wave beam formation method only is applicable to far field, arrowband unrelated signal, and ultrasound data has near field, broadband and very strong characteristics such as dependency, has limited its application in medical ultrasound image.To these problems, Synnevag etc. adopt the space smoothing method, have not only removed the dependency of echo-signal, and have obtained a good covariance matrix [8].Li etc. utilize the diagonal angle loading technique to improve the robustness of MV algorithm.Though the MV algorithm can improve imaging space resolution effectively, contrast is not improved.

How can further improve resolution and contrast and become the hot issue of industry, still, also not have a kind of molding the minimum variance wave beam is formed and the medical ultrasound image method that merges based on the minimum variance coherence factor at present.

Summary of the invention

In view of this, the purpose of this invention is to provide a kind of minimum variance wave beam and form and the method that merges based on the minimum variance coherence factor, compare with existing method, this method can better improve resolution, contrast and the robustness of image.

The objective of the invention is to realize through following technical scheme:

This minimum variance wave beam forms and the method that merges based on the minimum variance coherence factor; It is the corresponding direction vector structure minimum variance constraint square formation of one group of maximum of utilizing traditional Beam-former output; Obtain the output that the MV wave beam forms through the minimum variance beamforming algorithm; The phase stem portion that the output of MV wave beam formation is replaced coherence factor; Form a high-resolution coherence factor, weighting is carried out in the output of high-resolution coherence factor that obtains and the formation of minimum variance wave beam, finally obtain the output model that wave beam forms;

Said minimum variance beamforming algorithm may further comprise the steps:

1) keeping making the output energy minimization of array under the constant condition of user's desired orientation gain, its mathematic(al) representation is:

$\underset{w}{\min }{w}^H{R}_{i+n}w,\mathrm{subject\; to}{w}^{H}a=1,$

Wherein, wherein, R _I+nBe the covariance matrix of M * M interference plus noise, a is for getting complete 1 direction vector, and w is an optimum weighted value to be asked, w ^HBe the conjugate transpose of w,

$\underset{w}{\min }{w}^H{R}_{i+n}w,\mathrm{subject\; to}{w}^{H}a=1$

Promptly ought satisfy w ^HDuring a=1 constraints, make w ^HR _I+nW is w hour;

2) utilize lagrange's method of multipliers, solve the optimum weighting vector and be:

$w_{\mathrm{opt}}=\frac{R_{i+n}^{-1}a}{a^H{R}_{i+n}^{-1}a};$

3) with step 2) in covariance matrix replace with sample covariance matrix, sample covariance matrix is expressed as:

${\tilde{R}=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_d\left(n\right)x_d\left(n\right)}^H;$

4) adopt the space smoothing method to remove the dependency of echo-signal, this method is divided into the submatrix that array element number is L to M array element, estimates the correlation matrix of each submatrix respectively, and making even then all obtains:

${\tilde{R}=\frac{1}{M-L+1}\underset{l=1}{\overset{M-L+1}{\mathrm{\Σ}}}{x}_d^l\left(k\right)x_d^l\left(k\right)}^H;$ X wherein ^l _d(k) be the output vector of l submatrix,

$x_d^{l+1}\left(k\right),\·\·\·,{x_d}^{l+L-1}\left(k\right)];$

5) through step 2) calculate MV optimum weighting vector with the formula of step 4), finally being output as of MV wave beam formation:

$y_{\mathrm{mv}}\left(k\right)=\frac{1}{M-L+1}\underset{l=1}{\overset{M-L+1}{\mathrm{\Σ}}}{w}^H\left(k\right)x_d^l\left(k\right);$

The expression formula of high-resolution coherence factor is:

$\mathrm{HRCF}\left(k\right)=\frac{M{\left|y_{\mathrm{mv}}\left(k\right)\right|}^2}{\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{\left|x_d(m,k)\right|}^2};$

Calculate the high-resolution coherence factor through above-mentioned expression formula, and carry out weighting, obtain finally being output as of wave beam formation with the output that the minimum variance wave beam forms:

y _hrcf+mv(k)＝HRCF(k)y _mv(k)；

Wherein, the selection of submatrix length should guarantee that covariance matrix is reversible, is limited to M/2 on the L;

Further, L is set at equals M/4, and the diagonal angle is set loads Δ=1/ (10L).

The invention has the beneficial effects as follows: method of the present invention makes full use of the high-resolution of MV; The phase stem portion of the output replaced C F that forms with the minimum variance wave beam forms a high-resolution coherence factor; And carry out weighting fusion with the output that the minimum variance wave beam forms; Simulation result checking through point scattering target and sound absorption target shows: compare with existing method, method of the present invention can better improve resolution, contrast and the robustness of image.

Other advantages of the present invention, target and characteristic will be set forth in description subsequently to a certain extent; And to a certain extent; Based on being conspicuous to those skilled in the art, perhaps can from practice of the present invention, obtain instruction to investigating of hereinafter.Target of the present invention and other advantages can realize and obtain through following description and claims.

Description of drawings

In order to make the object of the invention, technical scheme and advantage clearer, will combine accompanying drawing that the present invention is made further detailed description below, wherein:

Fig. 1 is the various imaging contrast of a point target sketch map;

Fig. 2 is lateral resolution contrast sketch map;

Fig. 3 is the reconstructed image sketch map of sound absorption target.

The specific embodiment

Below will carry out detailed description to the preferred embodiments of the present invention with reference to accompanying drawing.Should be appreciated that preferred embodiment has been merely explanation the present invention, rather than in order to limit protection scope of the present invention.

The sensor signal model is normally set up in the following manner: suppose that M the equidistantly sensor array of array element arranged, and have some point scattering targets at the near-field region of array.The output that wave beam forms can be expressed as:

$y\left(k\right)=w^H\left(k\right)x_d\left(k\right)=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{w}_i\left(k\right)x_i(k-{\mathrm{\Δ}}_i)---\left(1\right)$

Wherein, x _d(k) for through focusing on the signal after delaying time, be expressed as x _d(k)=[x ₁(k-Δ ₁) ..., x _M(k-Δ _M)] ^T, w (k)=[w ₁(k),

, w _M(k)] ^TBe complex value weighting parameters, Δ _iAmount of delay for each passage.X (k) can be expressed as:

x(k)＝s(k)+i(k)+n(k) (2)

Wherein, s (k), i (k), n (k) representes desired signal, interfering signal, noise signal respectively.

The design object of Beam-former is exactly to seek best weighting parameters former echo-signal is handled at present, accurately estimates desired signal.

Method of the present invention is to utilize the corresponding direction vector structure minimum variance constraint square formation of one group of maximum of traditional Beam-former output; Obtain the output that the MV wave beam forms through the minimum variance beamforming algorithm; The phase stem portion that the output of MV wave beam formation is replaced coherence factor; Form a high-resolution coherence factor, weighting is carried out in the output of high-resolution coherence factor that obtains and the formation of minimum variance wave beam, finally obtain the output model that wave beam forms;

Wherein, the core of minimum variance beamforming algorithm is exactly to seek the weight vectors w an of the best, is keeping making the output energy minimization of array under the constant condition of user's desired orientation gain.Specifically comprised following steps:

1) keeping making the output energy minimization of array under the constant condition of user's desired orientation gain, its mathematic(al) representation is:

$\underset{w}{\min }{w}^H{R}_{i+n}w,\mathrm{subject\; to}{w}^{H}a=1,$

Wherein, R _I+nBe the covariance matrix of M * M interference plus noise, a is for getting complete 1 direction vector, and w is an optimum weighted value to be asked, w ^HBe the conjugate transpose of w,

$\underset{w}{\min }{w}^H{R}_{i+n}w,\mathrm{subject\; to}{w}^{H}a=1$

Promptly ought satisfy w ^HDuring a=1 constraints, make w ^HR _I+nW is w hour;

2) utilize lagrange's method of multipliers, solve the optimum weighting vector and be:

$w_{\mathrm{opt}}=\frac{R_{i+n}^{-1}a}{a^H{R}_{i+n}^{-1}a};$

3) with step 2) in covariance matrix replace with sample covariance matrix, sample covariance matrix is expressed as:

${\tilde{R}=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_d\left(n\right)x_d\left(n\right)}^H;$

Need to prove that standard MV beamforming algorithm is only applicable to arrowband, unrelated signal.For narrow band signal, direction vector a is one group of plural number relevant with signal frequency.Yet for the broadband signal of ultra sonic imaging, direction vector a can not simply be expressed as one group of plural number again, and it is relevant to receive signal height.In order to overcome these constraints, all adopt dynamic focusing in the stage that transmits and receives, the response from focus can be approximated to a plane wave incident array element like this, thereby obtains a direction vector with frequency-independent, is expressed as complete 1 vector;

4) adopt the space smoothing method to remove the dependency of echo-signal, this method is divided into the submatrix that array element number is L to M array element, estimates the correlation matrix of each submatrix respectively, and making even then all obtains:

${\tilde{R}=\frac{1}{M-L+1}\underset{l=1}{\overset{M-L+1}{\mathrm{\Σ}}}{x}_d^l\left(k\right)x_d^l\left(k\right)}^H;$

X wherein ^l _d(k) be the output vector of l submatrix, ${x^l}_d\left(k\right)=[x_d^l\left(k\right),x_d^{l+1}\left(k\right),\·\·\·,{x_d}^{l+L-1}\left(k\right)];$

5) through step 2) calculate MV optimum weighting vector with the formula of step 4), finally being output as of MV wave beam formation:

$y_{\mathrm{mv}}\left(k\right)=\frac{1}{M-L+1}\underset{l=1}{\overset{M-L+1}{\mathrm{\Σ}}}{w}^H\left(k\right)x_d^l\left(k\right).$

Among the present invention, the expression formula of high-resolution coherence factor is:

$\mathrm{HRCF}\left(k\right)=\frac{M{\left|y_{\mathrm{mv}}\left(k\right)\right|}^2}{\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{\left|x_d(m,k)\right|}^2};$

Calculate the high-resolution coherence factor through above-mentioned expression formula, and carry out weighting, obtain finally being output as of wave beam formation with the output that the minimum variance wave beam forms:

y _hrcf+mv(k)＝HRCF(k)y _mv(k)；

Wherein, the selection of submatrix length should guarantee that covariance matrix is reversible, is limited to M/2 on the L.Work as L=1, be equivalent to the DAS wave beam and form, the resolution of image reduces, and robustness improves; Work as L=M/2, the resolution of image improves, and robustness but reduces.Weigh through between, L=M/4 is set, and the diagonal angle is set loads Δ=1/ (10L) and guarantee to obtain a more sane covariance matrix.

Need to prove that coherence factor is as the index of weighing focusing quality.It is defined as the energy of relevant direction and the ratio of gross energy:

$\mathrm{CF}\left(k\right)=\frac{{\left|\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{x}_d(m,k)\right|}^2}{M\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{\left|x_d(m,k)\right|}^2};$

Wherein, k is a time coefficient, and (m k) is the reception signal through m passage after the focusing time-delay to xd.

The value of coherence factor is between 0 to 1.The value of coherence factor shows that more greatly ultrasound field has the directivity of height, is a high-quality imaging; Show low directivity or a low-quality imaging and value is more for a short time.Yet for speckle scattering target, itself have inherent incoherence, Li proposes broad sense coherence factor (GCF).Its homophase responsibility number is similar, and just molecule is replaced by the energy from a certain special angle scope in the transmit beam direction, and CF only comprises the energy of DC part.The mathematic(al) representation of GCF is:

$\mathrm{GCF}\left(k\right)=\frac{\underset{n=-(K-1)/2}{\overset{(K-1)/2}{\mathrm{\Σ}}}{\left|X\left(n\right)\right|}^2}{M\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{\left|x_d(m,k)\right|}^2},$ Wherein $X\left(n\right)=\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{x}_d(m,k)e^{-j2\mathrm{\π\; Nm}/M}.$

Simulation result

Utilize the Field II to carry out emulation experiment; Through carrying out algorithm process to receiving data; And beamforming algorithm of being carried and conventional beamforming algorithm, minimum variance beamforming algorithm and minimum variance wave beam formed the algorithm that merges with coherence factor in resolution, contrast and robustness aspect compare.

The mode of operation that all emulation all adopt dynamic emission and dynamic focusing to receive adopts linear array, and effectively array element number is 96, and tranmitting frequency is 4MHZ, and the systematic sampling frequency is 20MHZ, and the array element Center Gap is a half wavelength.The white Gaussian noise that has added 60dB during signal simulation, display dynamic range are 60dB.

(1) point scattering target

Totally 7 pairs of target scattering points are spacedly distributed in the degree of depth is the zone of 40 ~ 70mm, and two scattering point horizontal spacings of same depth are 2mm.Fig. 1 is the imaging results of distinct methods to the different depth scattering point.Wherein, Fig. 1 (a) is the time-delay stack; Fig. 1 (b) forms for the minimum variance wave beam; Fig. 1 (c) is time-delay stack and coherence factor; Fig. 1 (d) forms and coherence factor for the minimum variance wave beam; Fig. 1 (e) is that high-resolution coherence factor and minimum variance wave beam form.

As can be seen from Figure 1, the DAS wave beam form resolution and contrast be the poorest; It is higher that the MV wave beam forms resolution, but higher secondary lobe grade is still arranged; CF is improving contrast, and it is obvious to reduce secondary lobe grade aspect effect; MV is combined with CF, and picture quality is further enhanced; And HRCF that this paper proposed and MV are combined in raising resolution, contrast and it is best to reduce secondary lobe grade aspect effect.

In order further to understand the influence situation of distinct methods to picture quality in depth, Fig. 2 has provided the lateral resolution of point target at 40mm and 70mm place.Can know that from Fig. 2 the main lobe width of HRCF and the bonded method of MV is the narrowest, the secondary lobe grade is minimum, be MV and the bonded method of CF secondly, is followed successively by the bonded method of DAS and CF at last, MV and DAS.

(2) sound absorption target

The imaging depth of sound absorption target is 32 ~ 45mm.The scattering target is to be centered close to the 36mm degree of depth, and radius is the circular sound absorber of 3mm, and scattering object is Gauss distribution.Fig. 3 is the reconstructed image of different beams formation method; Wherein, Fig. 3 (a) is traditional time-delay stack, and Fig. 3 (b) forms for the minimum variance wave beam, and Fig. 3 (c) is traditional time-delay stack and coherence factor; Fig. 3 (d) forms and coherence factor for the minimum variance wave beam, and Fig. 3 (e) is that high-resolution coherence factor and minimum variance wave beam form.

In order to assess the contrast resolution that wave beam forms intuitively, introduce contrast (CR), the mean power that its mean power that is defined as center circle and external context are regional poor.Table 1 is listed the contrast of various formation methods.

Table 1 sound absorption objective body contrast

Can find out that from table 1 contrast that the HRCF that this paper proposes combines with MV to obtain is the highest.Secondly be that CF combines with MV to combine with CF with DAS, this is because CF algorithm effect on the contrast that improves image is obvious.Being higher than DAS as for the bonded contrast of CF and MV combines with CF; The reason that MV is higher than DAS is owing to L in the MV algorithm is set to M/4; The algorithm robustness is good, can be in the mean intensity and the approximately equalised while of DAS that keep the external context zone, and the mean intensity of center circle is reduced significantly.

Explanation is at last; Above embodiment is only unrestricted in order to technical scheme of the present invention to be described; Although with reference to preferred embodiment the present invention is specified, those of ordinary skill in the art should be appreciated that and can make amendment or be equal to replacement technical scheme of the present invention; And not breaking away from the aim and the scope of present technique scheme, it all should be encompassed in the middle of the claim scope of the present invention.

Claims (2)

1. the minimum variance wave beam forms and the method that merges based on the minimum variance coherence factor; It is characterized in that: the corresponding direction vector structure minimum variance constraint square formation of one group of maximum of utilizing traditional Beam-former output; Obtain the output that the MV wave beam forms through the minimum variance beamforming algorithm; The phase stem portion that the output of MV wave beam formation is replaced coherence factor; Form a high-resolution coherence factor, weighting is carried out in the output of high-resolution coherence factor that obtains and the formation of minimum variance wave beam, finally obtain the output model that wave beam forms;

Said minimum variance beamforming algorithm may further comprise the steps:

1) keeping making the output energy minimization of array under the constant condition of user's desired orientation gain, its mathematic(al) representation is:

Wherein, R _I+nBe the covariance matrix of M * M interference plus noise, a is for getting complete 1 direction vector, and w is an optimum weighted value to be asked, w ^HBe the conjugate transpose of w,

$\underset{W}{\min }{w}^H{R}_{i+n}w,\mathrm{subject\; to}{w}^{H}a=1$

Promptly ought satisfy w ^HDuring a=1 constraints, make w ^HR _I+nW is w hour;

2) utilize lagrange's method of multipliers, solve the optimum weighting vector and be:

$w_{\mathrm{opt}}=\frac{R_{i+n}^{-1}a}{a^H{R}_{i+n}^{-1}a};$

3) with step 2) in covariance matrix replace with sample covariance matrix, sample covariance matrix is expressed as:

${\tilde{R}=\frac{1}{N}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{x}_d\left(n\right)x_d\left(n\right)}^H;$

4) adopt the space smoothing method to remove the dependency of echo-signal, this method is divided into the submatrix that array element number is L to M array element, estimates the correlation matrix of each submatrix respectively, and making even then all obtains: ${\tilde{R}=\frac{1}{M-L+1}\underset{l=1}{\overset{M-L+1}{\mathrm{\Σ}}}{x}_d^l\left(k\right)x_d^l\left(k\right)}^H;$ X wherein ^l _d(k) be the output vector of l submatrix,

$x_d^{l+1}\left(k\right),\·\·\·,{x_d}^{l+L-1}\left(k\right)];$

5) through step 2) calculate MV optimum weighting vector with the formula of step 4), finally being output as of MV wave beam formation:

$y_{\mathrm{mv}}\left(k\right)=\frac{1}{M-L+1}\underset{l=1}{\overset{M-L+1}{\mathrm{\Σ}}}{w}^H\left(k\right)x_d^l\left(k\right);$

The expression formula of said high-resolution coherence factor is:

$\mathrm{HRCF}\left(k\right)=\frac{M{\left|y_{\mathrm{mv}}\left(k\right)\right|}^2}{\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}{\left|x_d(m,k)\right|}^2};$

Calculate the high-resolution coherence factor through above-mentioned expression formula, and carry out weighting, obtain finally being output as of wave beam formation with the output that the minimum variance wave beam forms:

y _hrcf+mv(k)＝HRCF(k)y _mv(k)；

Wherein, the selection of submatrix length should guarantee that covariance matrix is reversible, is limited to M/2 on the L.

2. minimum variance wave beam according to claim 1 forms and the method that merges based on the minimum variance coherence factor, it is characterized in that: L is set at equals M/4, and the diagonal angle is set loads Δ=1/ (10L).

Images