CN101238992A - Ultrasonic imaging system self-adaption beam former based on correlation analysis - Google Patents

Abstract

An adaptive beamforming device belongs to medicine ultrasound imaging field. The invention provides an adaptive beamforming technique based on correlativity analysis, directing towards the difficulty that the traditional beamforming device can not completely restrain sidelobe effect resulted from interchannel random phase error and not affect the focusing effect along with the dynamic alignment of detection position. The technique divides the echo signal obtained from each channel into several segments, direct towards each segment, firstly the reference signal is obtained by a classical beam forming method, and then the signal of each channel is mutually correlated with the reference signal, furthermore weight of channel in each segment is dynamically adjusted using correlation coefficient, finally signals of each channel are synthesized to obtain final beam output using adjusted weight. The adaptive beamforming device adaptively detects the variance of enviroment using dynamic apodization and restrains the sidelobe effect resulted from interchannel random phase error. The adaptive beamforming device has wide application prospect in ultrasound imaging field.

Description

Ultrasonic imaging system self-adaption Beam-former based on correlation analysis

Technical field

The present invention relates to the medical ultrasound imaging technical field, relate in particular to a kind of adaptive beam former.

Background technology

Ultra sonic imaging is the very wide diagnostic method of a kind of scope of application, has not damaged, no ionizing radiation, advantage such as easy to use.It is link the most key in the ultrasonic image-forming system that wave beam forms, and directly affects the quality of ultra sonic imaging.Low side lobe levels is the target of pursuing in the Beam-former design always.From the sixties in 20th century to the end of the eighties, ultrasonic image-forming system mainly adopts analog beam formation technology, and is bulky and focusing accuracy is lower.Early 1990s, along with large scale integrated circuit and at a high speed, the development of high-resolution modulus (A/D) transducer, digital beam forms technology and has occupied leading position gradually; It utilizes digital storage technique echo-signal is sampled and to note, and the mode that adds up by digital delay realizes focusing on then.

The focusing accuracy of digital beam formation device is relevant with the sampling rate of A/D converter, and the sampling rate of business PC use at present is generally 40MHz, can reach 10ns through focusing accuracy after the interpolation processing.Yet in real system, owing to reasons such as the restriction of manufacturing process and passage mismatches, have phase error inevitably between probe array element, the anisotropism of tested tissue also can be brought phase error simultaneously, and the phase error of these random distributions makes beam side lobe increase.Usually phase error improves sampling rate without any meaning again much larger than focusing accuracy.In such cases,, can only increase the port number of system in order further to reduce secondary lobe, to improve signal to noise ratio, however port number whenever double, signal to noise ratio improves 3dB at the most, cost performance is quite low.

Aperture change mark is an important technology during wave beam forms, and its ultimate principle is exactly to give bigger weight coefficient to the echo-signal that center array element receives, and weight coefficient reduces gradually to both sides, so that suppress beam side lobe and graing lobe.Traditional beam-forming technology is realized becoming mark by apply Gaussian window, Hanning window etc. on receiving aperture, and when channel phases error random distribution, fixedly apodizing function also changes for the inhibition ability of secondary lobe thereupon.In addition, after apodizing function is determined, the ability of its suppressed sidelobes changes with investigation depth: can suppress secondary lobe outside the axle well at the effective scanning depth bounds of change mark, and the ability of suppressed sidelobes weakens beyond the effective depth bounds of change mark, even secondary lobe is increased.This shows that traditional Beam-former can not suppress fully for the side lobe effect that the interchannel random phase error causes, and can not be with the change dynamics adjustment of detecting location, thereby focusing effect influenced.

Summary of the invention

The present invention is directed to traditional Beam-former can not fully suppress the side lobe effect that the interchannel random phase error causes and can not dynamically adjust the difficulty that influences focusing effect with detecting location, the beam-forming technology based on correlation analysis in a kind of ultra sonic imaging is provided, the dynamic apodization method of this technology utilization adapts to the variation of acquisition environment, when sonde configuration and system parameter variations, the weight of each passage is according to adjusting adaptively the time of advent of echo-signal, even there is random phase error, also can in whole ultrasonic scanning scope, obtain Sidelobe Suppression effect preferably.This self adaptation dynamically becomes the difference detection position of mark technology for different detected object and same target, can both reach focusing effect preferably, the lateral resolution of image and contrast resolution can have comparatively significantly raising with respect to traditional beam-forming technology.

The adaptive beam formation method that the present invention proposes is based on correlation analysis.In signal analysis and handling, dependency is an important concept.So-called " being correlated with " is meant a kind of linear relationship between the variable.For deterministic signal, the available functions that concerns between two variablees is described, and both have relation one to one.And, then can only relation between them be described by correlation analysis for stochastic signal.In ultrasonic system, probe phase error and system noise all are stochastic signals usually, make echo-signal also have certain randomness.

In ultrasound echo signal, the dependency that comprises between the useful signal of organizational information is very strong, and between the different passage because phase error and noise have random distribution nature, dependency a little less than.Therefore can utilize relative coefficient to come the weight of each passage is done dynamically to adjust, so that suppressed sidelobes better improves signal to noise ratio.

Beam-former provided by the invention is realized by following steps:

The first step, the echo-signal that each passage is obtained is divided into plurality of sections, at each section, at first utilizes classical wave beam formation method to obtain a reference signal, then the signal and the reference signal of each passage is done cross-correlation, calculates relative coefficient.

Suppose that the sampled point number in each segmentation is K, n passage receive with degree of depth k ₀For the section echo-signal at center can be expressed as:

$x_n\left(k_0\right)={[x_n(k_0-\frac{K}{2}),x_n(k_0-\frac{\mathrm{<figure-callout\; id="2"\; label="K"\; filenames="A20081006396600121.png"\; state="\{\{state\}\}">K</figure-callout>}}{2}+1),\&CenterDot;\&CenterDot;\&CenterDot;,x_n({\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="A20081006396600111.png,A20081006396600121.png"\; state="\{\{state\}\}">k</figure-callout>}}_0+\frac{K}{2})]}^T---\left(1\right)$

Reference signal is obtained by the wave beam formation method of classics:

$x_b\left(k_0\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{a}_n{x}_n\left(k_0\right)---\left(2\right)$

A wherein _nBe the weight of n passage during classical wave beam forms, as long as apodizing function is determined a _nFixing thereupon; N is the receive path number.The present invention adopts the windowed function of Gaussian window in forming as classical wave beam.

With every section echo-signal of n passage therewith the reference signal of section do computing cross-correlation, obtain correlation coefficient and be:

${\mathrm{\&rho;}}_{n,b}\left(k_0\right)=\frac{\mathrm{cov}(x_n\left(k_0\right),x_b\left(k_0\right))}{\sqrt{D\left(x_n\left(k_0\right)\right)D\left(x_b\left(k_0\right)\right)}}---\left(3\right)$

Wherein $\mathrm{cov}(x_n\left(k_0\right),x_b\left(k_0\right))=\frac{1}{K}\underset{k=k_0-K/2}{\overset{k=k_0+K/2}{\mathrm{\&Sigma;}}}[x_n\left(k\right)-\stackrel{\&OverBar;}{x_n\left(k\right)}]\&times;[x_b\left(k\right)-\stackrel{\&OverBar;}{x_b\left(k\right)}],$

$D\left(x_n\left(k_0\right)\right)=\frac{1}{K}\underset{k=k_0-K/2}{\overset{k=k_0+K/2}{\mathrm{\&Sigma;}}}{[x_n\left(k\right)-\stackrel{\&OverBar;}{x_n\left(k\right)}]}^2,$

$D\left(x_b\left(k_0\right)\right)=\frac{1}{K}\underset{k=k_0-K/2}{\overset{k=k_0+K/2}{\mathrm{\&Sigma;}}}{[x_b\left(k\right)-\stackrel{\&OverBar;}{x_b\left(k\right)}]}^2$

Here reference signal is equivalent to one poor priori source.If the echo-signal and the reference signal of certain passage are close more, corresponding relative coefficient is just big more, and the weight of this passage in adaptive beam forms is also big more.That is to say,, it is strengthened if the signal of certain passage and useful signal dependency are stronger; Otherwise, if with the useful signal dependency a little less than, then it is suppressed.

In second step, utilize relative coefficient that the weight of each passage in this segmentation done dynamically to adjust synthetic echo signal.The dynamic adjustment of each passage weight realizes by weight coefficient.Each passage multiplies each other original weight coefficient and correlation coefficient, obtains new weight coefficient:

a′ _n(k ₀)＝ρ _n，b(k ₀)a _n (4)

So degree of depth k ₀The new synthetic echo signal at place is:

$x\left(k_0\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{a}_n^{\&prime;}\left(k_0\right)x_n\left(k_0\right)=\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{\&rho;}}_{n,b}\left(k_0\right)a_n{x}_n\left(k_0\right)---\left(5\right)$

As can be seen, the echo-signal that each passage array element receives has been passed through dynamic matched filtering again and has been handled through after the Gaussian window weighting and adding up, and makes reference signal all have higher signal to noise ratio near field and far field.Dynamically the output of matched filter is reference signal.When systematic parameter and acquisition environment variation, the weight of each passage also changes adaptively along with the time of advent of echo-signal, thereby makes that the secondary lobe at each investigation depth place can both be suppressed preferably.

According to above analysis, the reference signal that is formed by traditional beam-forming technology is actually a feedback of beam feature, and adaptive beam former proposed by the invention just is equivalent to a closed loop feedback control system.Traditional beam-forming technology is regarded beam forming process as a space-filtering operation, with superimposed, out-phase decay, and adaptive beam formation method proposed by the invention is considered more random factor, utilize relative coefficient that each passage weight coefficient is carried out dynamic optimal and distribute, thereby guarantee in the environment that changes, to obtain best wave beam.

Focusing delay parameter among the present invention can calculate by following method: the corresponding relation curve of at first setting up the ultrasonic attenuation characteristic and the velocity of sound, search the instant velocity of sound that obtains measured target according to the attenuation characteristic of ultrasound wave in measured target by the relation curve of setting up then, just can calculate the definite focusing delay parameter of current measured target based on this instant velocity of sound.The curve is here set up, is searched with delay parameter computational process and can both realize on computers.

Adaptive beam former of the present invention can adopt field programmable gate array (FPGA) to realize.Utilize the on-line reconfiguration characteristic of FPGA, the work process of system can be divided into parameter identification and normal two stages of scanning.The parameter cognitive phase will be separated with I/Q and be in harmonious proportion the parameter recognition logic that plural self correlation is estimated as core and dispose to FPGA, and for same detection position, it is just passable only need to carry out the primary parameter identifying.The normal scanning stage forms logic configuration with acoustic beam and gives FPGA, and is identical with traditional Beam-former on the structure, has only the delay parameter of focusing to obtain through the parameter identifying.Owing to utilized the reconfigurable characteristics of FPGA, parameter identification and wave beam formation can time sharing shared same fpga chips, so the present invention compares with traditional beam-forming device, hardware cost does not have any increase.

Description of drawings

Fig. 1 is the adaptive beam former theory diagram.

Fig. 2 is the adaptive beam former functional block diagram.

It is 0 that Fig. 3 obeys average for phase error between array element, when variance is the normal distribution of 0.064 π, the beam position figure under the different beams formation method.The adaptive beam formation method that proposes among the corresponding the present invention of solid line, the wave beam formation method that dotted line is corresponding traditional.

It is 0 that Fig. 4 obeys average for phase error between array element, when variance is the normal distribution of 0.64 π, the beam position figure under the different beams formation method.The adaptive beam formation method that proposes among the corresponding the present invention of solid line, the wave beam formation method that dotted line is corresponding traditional.

Among Fig. 1:

101 first in first out (FIFO) buffer

102 multipliers 1

103 accumulators 1

104 matched filters 1

105 computing cross-correlation devices

106 weight adjustors

107 multipliers 2

108 accumulators 2

109 matched filters 2

Among Fig. 2:

201 A/D sampling arrays

202 field programmable gate arrays (FPGA)

203 parameter recognition logics

204 acoustic beams form logic

205 FPGA configuration circuits

206 pci interfaces

207 computers (PC)

The specific embodiment

Below in conjunction with accompanying drawing the specific embodiment of the present invention is described in further detail.

Whole adaptive beam former theory diagram as shown in Figure 1.Echo-signal enters behind the adaptive beam former at first buffer memory in first in first out (FIFO) buffer 101, being divided into three the tunnel behind the buffer memory handles: the one tunnel utilizes traditional method to carry out wave beam forms, one the tunnel makes computing cross-correlation with reference signal, and another road is waited for and carried out adaptive beam formation adjusted passage weight coefficient calculations is come out after.Form link at traditional wave beam, each passage echo-signal at first multiplies each other by multiplier 102 with the window coefficient of fixed window, add up by accumulator 103 then, add up and be the composite value of traditional wave beam formation method, and then through dynamic filter link 104, carry out matched filtering, eliminate out-of-band noise.Signal behind the dynamic filter is reference signal, and each channel signal and reference signal are carried out cross-correlation calculation by computing cross-correlation device 105, obtains cross-correlation coefficient.Then, utilize cross-correlation coefficient that original window coefficient is adjusted by weight adjustor 106, each passage echo-signal of adjusted weight coefficient and buffer memory is multiplied each other by multiplier 107, then by accumulator 108 summation that adds up, after adding up and carry out dynamic filter by matched filter 109 again, wave filter output is the resulting synthetic wave beam of adaptive beam former.

The adaptive beam former functional block diagram as shown in Figure 2, ultrasound echo signal enters field programmable gate array (FPGA) 202 through A/D sampling array 201, the concrete XC3S4000 that adopts Xilinx company, the A/D converter that it is 8bit that each passage is joined a resolution specifically adopts AD9057.At the parameter cognitive phase, computer 207 downloads to the parameter recognition logic among the SRAM of FPGA configuration circuit 205 by PCII interface 206, CPLD in the FPGA configuration circuit 205 reads logical code and gives FPGA 202 by the configuration of certain time sequence rule from SRAM then, and FPGA 202 realizes parameter recognition logics 203 at this moment.In the normal scanning stage, computer 207 forms acoustic beam among the SRAM that logic downloads to configuration circuit 205 by pci interface 206, CPLD in the configuration circuit 205 reads logical code and gives FPGA 202 by the configuration of certain time sequence rule from SRAM then, and this moment, FPGA 202 realization acoustic beams formed logic 204.

Fig. 3 has shown interchannel, and exist to obey average be 0, when variance is the normal distribution random phase error of 0.064 π, the performance comparison of two kinds of wave beam formation methods.The adaptive beam formation method that proposes among the corresponding the present invention of solid line, the wave beam formation method that dotted line is corresponding traditional.When not adopting when becoming the mark technology, beam side lobe is bigger, and the main lobe characteristic is also bad, has had a strong impact on the horizontal minute ratio and the contrast resolution of image.Traditional wave beam formation method that receiving aperture is added fixed window is suppressed sidelobes to a certain extent, and incident is the increase of main lobe width, but Sidelobe Suppression effect and unsatisfactory.By contrast, adaptive beam formation method is suppressed sidelobes fully, has improved the lateral resolution of image.

Fig. 4 has shown interchannel, and exist to obey average be 0, when variance is the normal distribution random phase error of 0.64 π, the performance comparison of two kinds of wave beam formation methods.At this moment, when not adopting any change mark technology, the beam feature severe exacerbation is difficult to distinguish main lobe and secondary lobe, and focus beam does not almost have directivity.As shown in the figure, traditional beam-forming technology is suppressed sidelobes to a certain extent, but side lobe levels is still very big, almost with main lobe on an order of magnitude, adaptive approach then can be suppressed at the secondary lobe amplitude a lower level, guarantees the lateral resolution and the contrast resolution of image to a certain extent.The side lobe effect among the figure and the deterioration of beam feature are that inevitably this is owing to the too big reason of phase error.

Claims (5)

1. dynamic self-adapting Beam-former that is used for the digital ultrasound imaging, it is characterized in that: the echo-signal that each passage is obtained is divided into plurality of sections, at each section, at first utilize classical wave beam formation method to obtain a reference signal, then the signal and the reference signal of each passage are done cross-correlation, utilize relative coefficient that the weight of passage in this segmentation done dynamically to adjust again, utilize adjusted weight that each channel signal is synthesized at last and obtain final wave beam output.

2. adaptive beam former according to claim 1, it is characterized in that: adopt the I/Q demodulation module that the radio frequency echo data is carried out quadrature demodulation, orthogonal reference is sinusoidal identical with the ultrasound emission frequency with the frequency of cosine signal, obtain the data of homophase I and two passages of quadrature Q after the demodulation, I, Q data are evenly divided into plurality of sections, between section and the section overlapping is arranged, so that make the result more level and smooth, the data after the segmentation are carried out plural self correlation and are estimated.

3. adaptive beam former according to claim 1, it is characterized in that: utilize classical wave beam formation method to obtain a reference signal, this reference signal is equivalent to one poor priori source, for the echo-signal passage close with reference signal, big relative coefficient is set, the weight of this passage in adaptive beam forms is also bigger, otherwise little relative coefficient then is set, and the weight of this passage in adaptive beam forms is also less.

4. adaptive beam former according to claim 1, it is characterized in that: utilize relative coefficient that the weight of passage in this segmentation done dynamically to adjust, the weight method of adjustment is that original weight coefficient and correlation coefficient are multiplied each other, and calculates new weight coefficient.

5. adaptive beam former according to claim 1, it is characterized in that: the host computer of system is a computer, the all scanning control parameters of logical code and system that have FPGA in this computer, computer and ultrasonic system swap data, at work, computer disposes logical code to FPGA by configuration circuit earlier, and then that some are necessary scanning control parameter downloads is to FPGA, after one two field picture scanning was finished, the computer reading images also showed.

Images