阅读说明：本技术 用于超声系统独立衰减系数估计的方法 (Method for ultrasound system independent attenuation coefficient estimation ) 是由 公平 P·宋 S·陈 J·D·奇萨思科 于 2018-06-22 设计创作，主要内容包括：描述了用于使用参考频率方法(“RFM”)估计正被成像的受试者或其他对象的感兴趣区域中的声学特性的系统和方法。使用此RFM技术,通过在不同频率(例如,相邻频率或其他频率)下采集的超声数据对在给定频率下采集的超声数据进行标准化,以便提供对独立于用于采集基础数据的超声系统的声学特性(例如,衰减系数、反向散射系数、或两者)的估计。例如,每个频率分量的振幅可通过功率谱中的不同频率进行标准化,以消除系统依赖的效应。因为本公开中描述的方法是系统独立的,所以它们可应用于任何换能器几何形状(例如,线性或曲线阵列)并且可使用任何波束图案(例如,聚焦的或非聚焦的)。(Systems and methods for estimating acoustic properties in a region of interest of a subject or other object being imaged using a reference frequency method ("RFM") are described. Using this RFM technique, ultrasound data acquired at a given frequency is normalized by ultrasound data acquired at different frequencies (e.g., adjacent frequencies or other frequencies) in order to provide an estimate of acoustic characteristics (e.g., attenuation coefficient, backscatter coefficient, or both) that are independent of the ultrasound system used to acquire the underlying data. For example, the amplitude of each frequency component may be normalized by a different frequency in the power spectrum to eliminate system-dependent effects. Because the methods described in this disclosure are system independent, they can be applied to any transducer geometry (e.g., linear or curved arrays) and can use any beam pattern (e.g., focused or unfocused).)

1. A method for estimating acoustic properties of a physical medium in a region of interest using an ultrasound system, the steps of the method comprising:

(a) acquiring ultrasound data from a plurality of depth locations along an ultrasound beam path within the region of interest, the ultrasound data including at least first ultrasound data acquired in response to transmitting ultrasound at a first frequency to the region of interest and second ultrasound data acquired in response to transmitting ultrasound at a second frequency different from the first frequency to the region of interest;

(b) calculating a plurality of spectral ratio values by calculating a ratio between the first ultrasound data and the second ultrasound data acquired at each of a plurality of different depth positions;

(c) estimating an acoustic property value of a physical medium in the region of interest from the plurality of spectral ratio values; and

(d) generating a report based on the estimated acoustic property value of the physical medium.

2. The method of claim 1, wherein the acoustic property value is an attenuation coefficient.

3. The method of claim 2, wherein estimating the attenuation coefficient comprises: calculating a logarithm value by calculating a logarithm of the spectral ratio values, forming a spectral ratio decay curve by arranging the logarithm values as a function of depth position, calculating a slope over a portion of the spectral ratio decay curve, and estimating the attenuation coefficient based on the slope.

4. The method of claim 3, wherein the slope is calculated based on a linear fit.

5. The method of claim 4, wherein the reliability of the attenuation coefficient is estimated based on a residual error of the linear fit.

6. The method of claim 4, wherein the portion of the spectral ratio decay curve is selected as a range of depth positions over which the spectral ratio decay curve follows a linear decline as a function at the first frequency.

7. The method of claim 4, wherein the slope is calculated based on a non-linear fit.

8. The method of claim 7, wherein the non-linear fit is calculated based in part on a variable projection (varro) method.

9. The method of claim 2, wherein estimating the attenuation coefficient comprises calculating a logarithm value that is a logarithm of the plurality of spectral ratios and fitting the logarithm value to a parametric model to estimate the attenuation coefficient.

10. The method of claim 9, wherein the parametric model is a single parametric model of spectral ratios, and estimating the attenuation coefficients comprises normalizing the plurality of spectral ratios by reference depth data prior to calculating the log values.

11. The method of claim 10, wherein the reference depth data is ultrasound data acquired at the first frequency at one or more reference depths.

12. The method of claim 9, wherein the parametric model is a two-parameter model of a frequency spectrum, and estimating the attenuation coefficient further comprises estimating a backscatter coefficient.

13. The method of claim 12, wherein step (a) is repeated using at least one of different first and second frequency values or different depth locations such that the plurality of spectral ratios correspond to at least one of a plurality of frequency measurements or a plurality of depth measurements.

14. The method of claim 9, wherein fitting the logarithmic values to the parametric model comprises using a least squares estimate.

15. The method of claim 9, wherein fitting the logarithmic values to the parametric model comprises constraining the fitting using a constraint.

16. The method of claim 15, wherein the constraint is a physical medium constraint that defines an expected range of attenuation coefficient values for the physical medium.

17. The method of claim 15, wherein the constraint is a data generation constraint that is adaptively determined based at least in part on the spectral ratio value.

18. The method of claim 1, wherein step (a) is repeated for a plurality of different first and second frequency values, step (b) is repeated for each of the plurality of different first and second frequency values, and step (c) comprises estimating a different value of the acoustic characteristic for each of the plurality of different first frequency values and estimating a final value of the acoustic characteristic based on a combination of the different values of the acoustic characteristic.

19. The method of claim 18, wherein the combining of the different acoustic property values comprises averaging the different acoustic property values.

20. The method of claim 19, wherein the different acoustic property values are averaged using a weighted average.

21. The method of claim 1, wherein the acoustic property value is a backscatter coefficient.

22. The method of claim 21, wherein estimating the backscatter coefficient comprises: calculating a logarithm value by calculating a logarithm of the spectral ratio values, forming a spectral ratio decay curve by arranging the logarithm values as a function of depth position, calculating a y-intercept of the spectral ratio decay curve, and estimating the backscatter coefficient based on the slope.

23. The method of claim 22, wherein arranging the plurality of spectral ratio values as a function of depth position comprises: providing transducer response data to the computer system, calculating transducer response ratios, and weighting the plurality of spectral ratio values based on the transducer response ratios.

24. The method of claim 23, further comprising: a slope of the spectral ratio decay curve is calculated and an attenuation coefficient value is estimated based on the slope.

25. The method of claim 1, wherein the first frequency is spectrally adjacent to the second frequency.

Background

Ultrasound attenuation coefficient estimation ("ACE") is useful for clinical applications. For example, estimates of ultrasound attenuation coefficients can be used to distinguish tumors and quantify fat content in the liver.

In the example of fat content detection, fat in the liver affects ultrasound propagation and is associated with an attenuation coefficient. Therefore, note that the coefficient may be used as a liver fat assessment factor. ACE provides a non-invasive and repeatable process compared to gold standard, liver biopsy.

The two most common ACE methods include spectral shift and reference volume mode. Estimating the downward shift of the center frequency along the depth by a frequency spectrum shift method; however, this approach ignores beam forming and diffraction effects, thereby limiting the estimation accuracy. On the other hand, in the reference phantom approach, a well-calibrated phantom is used to normalize all ultrasound system-dependent effects, such as focus, time gain compensation ("TGC"), and diffraction. However, in practice, well-calibrated phantoms are not always available, and their ultrasound properties may change over time as the materials in the phantom degrade or otherwise change over time.

Therefore, there remains a need for a method of estimating attenuation coefficient values that does not require a reference phantom for normalization while still being independent of ultrasound system effects.

Disclosure of Invention

The present disclosure addresses the above-described shortcomings by providing a method for estimating acoustic properties of a physical medium in a region of interest using an ultrasound system. Ultrasound data is acquired from a plurality of depth positions along an ultrasound beam path within the region of interest. The ultrasound data includes at least first ultrasound data acquired in response to transmitting ultrasound to the region of interest at a first frequency and second ultrasound data acquired in response to transmitting ultrasound to the region of interest at a second frequency different from the first frequency. Spectral ratio values are calculated by calculating a ratio between the first ultrasound data and the second ultrasound data acquired at each of the plurality of different depth positions, and acoustic property values of the physical medium in the region of interest are calculated from the plurality of spectral ratio values. A report is generated based on the estimated acoustic property value of the physical medium.

The above and other aspects and advantages of the present disclosure will become apparent from the following description. In the description, reference is made to the accompanying drawings which form a part hereof, and in which there is shown by way of illustration preferred embodiments. This embodiment does not necessarily represent the full scope of the invention, however, and reference is made therefore to the claims and herein for interpreting the scope of the invention.

Drawings

FIG. 1A shows three exemplary frequency components f _iOr 5, 6 and 7MHz, a frequency power spectrum taken from the natural logarithm as a function of depth.

Fig. 1B shows the ratio of the spectrum taken from the natural logarithm as a function of depth at the same frequency component. For better visualization, the spectral ratio decay curves for 6 and 7MHz are shifted up by 0.05 and 0.1, respectively.

Fig. 2 shows a histogram of possible attenuation coefficients. Data were collected from a commercial phantom with a nominal value of 0.5 dB/cm/MHz. The phantom is imaged with unfocused plane wave imaging at a 5MHz center frequency.

FIG. 3 is a flow chart setting forth the steps of an example method for estimating acoustic properties of a physical medium, such as tissue, from ultrasound data using a reference frequency method.

FIG. 4 is a block diagram of an example ultrasound system in which methods described in this disclosure may be implemented.

FIG. 5 is a block diagram of an example computer system that may implement the methods described in this disclosure.

Detailed Description

Described herein are systems and methods for estimating acoustic properties in a region of interest of a subject or other object being imaged using a reference frequency method ("RFM"), in which ultrasound data acquired at a given frequency is normalized by ultrasound data acquired at neighboring frequencies to provide an estimate of the acoustic properties (e.g., attenuation coefficient, backscatter coefficient, or both) independent of the ultrasound system used to acquire the underlying data. For example, the amplitude of each frequency component may be normalized by different frequencies in the power spectrum to eliminate system-dependent effects. Because the methods described in this disclosure are system independent, they can be applied to any transducer geometry (e.g., linear or curved arrays), and any beam pattern (e.g., focused or unfocused) can be used.

In ultrasound imaging, the power spectrum of a backscattered RF signal can be modeled as:

S(f _i,z _k)＝G(f _i)·TGC(z _k)·D(f _i,z _k)·BSC(f _i)·A(f _i,z _k) (1)；

wherein G (f) _i) Representing the transmit and receive transducer responses at frequency fi, where i is the frequency component index); TGC (z) _k) Is the time gain compensation ("TGC") as the depth z _kWhere k is the depth index); d (f) _i,z _k) Is focusing, beam forming and diffractingThe combined effect of the shots; BSC (f) _i) Is the backscatter coefficient assumed to be uniform in a region of interest ("ROI"); and A (f) _i,z _k) Is a frequency dependent attenuation that can be expressed as:

where a is the attenuation coefficient. In many cases, A (f) may be assumed _i,z _k) Is homogeneous in the ROI and has a linear frequency dependence. The shape of the ROI is flexible and in some cases may be rectangular, square, fan-shaped, etc. This type of model is generally applicable to all ultrasound systems regardless of the beam pattern (e.g., unfocused or focused).

To estimate the attenuation coefficient a, some multiplication terms as in equation (1) may be first eliminated. The teaching of the method described in this disclosure is that this can be achieved by grouping adjacent frequencies f _i-1Regarding as a reference frequency and calculating S (f) _i,z _k) And S (f) _i-1,z _k) The ratio between, as follows:

adjacent frequencies f in equation (3) _i-1But is only one non-limiting example. More generally, the adjacent frequency may be f _i±nWhere i and n are frequency indices that indicate (index) frequencies in a given frequency range. Although i and n are integers (e.g., i ≧ 0 and n ≧ 1), frequency f _i±nThe values of (d) need not be integers but may be any real-valued frequency. For example, in one example, f _iMay be equal to 5MHz, and f _i±nMay be equal to 5.1 MHz. In another example, f _iMay be equal to 5MHz, and f _i±nMay be equal to 6 MHz. It should therefore be understood that the methods described in this disclosure may be generalized to calculating a ratio between a signal acquired in response to a first transmit frequency and a signal acquired in response to a second transmit frequency different from the first transmit frequency. Thus, it is possible to provideAlthough for purposes of providing examples, adjacent frequencies are referred to as f in this disclosure _i-1However, it should be understood that f can be used _i-1The other frequencies serve as neighboring frequencies considered as reference frequencies.

It can be assumed in general that, at the same depth, f _iAnd f _i-1The difference in beam forming and diffraction effects between can be neglected, so that,

D(f _i,z _k)＝D(f _i-1,z _k) (4)。

furthermore, TGC (z) _k) Independent of frequency f _i. These two terms can therefore be eliminated after taking the ratio as in equation (3), which results,

after taking the natural logarithm of both sides of equation (5), the following linear relationship can be obtained:

ln(Rs(f _i,z _k))＝ln(G(f _i))-ln(G(f _i-1))+ln(BSC(f _i))-ln(BSC(f _i-1))-4a(f _i-f _i-1)z _k(6)。

relative to the independent variable z _kZ in equation (6) _kThe slope of (a) is such that,

slope-4 a _i(f _i-f _i-1) (7)；

And therefore the number of the first and second channels,

wherein a is _iIs at the frequency component f _iThe estimated attenuation coefficient of (d). It should be understood that although natural logarithms are used above and in the latter part of this disclosure, other logarithms, such as base-10 logarithms, may be implemented.

FIG. 1A shows three exemplary frequency components f at 5MHz, 6MHz, and 7MHz _iApplying the natural logarithm as a function of depth (i.e., ln (S (f) S) below _i,z _k) ) of the frequency spectrum of the power spectrum of the frequency spectrum. FIG. 1B shows the application of the natural logarithm (i.e., ln (Rs (f) as a function of depth) at the same frequency component _i,z _k) ) the ratio of the frequency spectrum after (c)). After taking the ratio, the frequency curve decays with similar slope for different frequencies until z _kIs 4 cm. However, the 6MHz frequency curve begins to rise at a depth of about 6cm, while the 7MHz frequency curve begins to rise at about 4 cm. This is because the higher frequency signal decays faster with depth than the lower frequency, and because electrical noise starts to dominate after a certain depth (i.e., resulting in an erroneously improved attenuation coefficient estimate). The linear fit for each frequency is represented by the dashed line, where the attenuation coefficient values are shown in the upper left corner of fig. 1B. The values are converted to units of dB/cm/MHz.

The linearity over depth (linear coverage depth) of the decay curve for different spectral ratios can be used as a quality control to determine the reliability of the attenuation estimate. For example, the residual error of the linear fit may be used as a metric to indicate the reliability of the attenuation estimate of the spectral ratio decay curve at a given frequency.

In another example of quality control, the maximum depth of an effective linear fit of a spectral ratio decay curve at a given frequency may be determined as the depth beyond which the curve no longer linearly decreases with depth. This information can then be used to determine the frequency components for the final attenuation estimate at a given depth. For example, as one example, estimates at different frequencies may be averaged.

In another example of quality control, a realistic boundary of ultrasound attenuation may be selected based on the expected range of attenuation for different soft tissues (e.g., based on reported ranges of values from literature), which may then be used to reject unrealistic estimates. For example, if the spectral ratio is at a certain frequency f _iWhen the lower side shows a positive slope, the corresponding attenuation a _iWill become negative, which is impractical and should therefore be eliminated from the final averaging process.

From a graph such as that shown in fig. 1B, various attenuation coefficient values at each frequency can be estimated by combining all the estimated slopes over the entire power spectrum. As one non-limiting example, the final attenuation coefficient a may be calculated by averaging all estimates at different frequencies, as follows,

where I is the total number of frequency components to be used in the attenuation coefficient estimation. In the example of equation (9), at frequency component f _iLower estimated attenuation coefficient a _iWith the same weight during combining. The attenuation coefficient value a may also be taken into account _iSuch as by applying greater weight to the center frequency and less weight to frequencies near the bandwidth limit, or by applying greater weight to estimates at frequencies that have a better linear fit with depth (e.g., those with smaller fit errors).

In another non-limiting example, the spectral ratio decay curves at different frequencies may be first averaged. The average may be a weighted average, or a non-weighted average. The slope of the averaged decay curve may then be estimated along the depth to calculate the final attenuation coefficient.

Spectral ratio Rs (f) as in equation (5) _i,z _k) If the transducer response G (f) at each frequency can be calibrated in advance _i) This term becomes a known parameter in the equation.

As one non-limiting example, the transducer response may be measured with hydrophones underwater. The power spectrum of the signal transmitted by the transducer and received by the hydrophone corresponds to a unidirectional (i.e., transmitting) transducer response at different frequency components. This spectrum can then be squared to calculate the bi-directional transducer response by assuming that the received response of the transducer is the same as the transmitted response.

For calibrating transducer response G (f) _i) Another non-limiting example of (a) is that the transducer emits directly into the air, or the transducer is strong into the waterThe reflective interface transmits and then receives echoes. The power spectrum of the echo received by the transducer itself accounts for both the transmit and receive transducer responses.

In the case of a calibrated transducer response, the backscattering coefficient BSC (f) in equation (5) can be estimated simultaneously with the attenuation coefficient _i) Frequency dependence of (d). The backscatter coefficients can be modeled as,

wherein b is a constant coefficient and n _bIs frequency dependent. Substituting (10) into equation (5) yields,

wherein R is _G(f _i) Is the transducer frequency response ratio G (f) _i)/G(f _i-1). Taking the natural logarithm of both sides of equation (11) results,

ln(Rs(f _i,z _k))＝ln(R _G(f _i))+n _b(f _i-f _i-1)-4az _k(f _i-f _i-1) (12)。

ln (Rs (f) may then be performed _i,z _k) Linear fit against (versus) depth, as shown in fig. 1B, to find the intercept with the vertical axis: y is _{Intercept of a beam}. The frequency dependent backscatter coefficients can then be estimated as:

wherein n is _b,iIs at f _iThe estimated backscatter coefficients of. Thereafter, all n estimated on the power spectrum may be summed _b,iThe values are combined for final estimation of the backscatter coefficients. One example of a combination is a weighted or unweighted average. In another example, n may be _b,iThe values are plotted as a function of frequency and, as a first order approximation, n _b,iRelative (v)ersus) the slope of the frequency may be used as a parameter for characterizing tissue.

Since tissue may typically have small structures that are more complex and non-uniform, resulting in larger oscillations in the spectral ratio decay curve, in some cases, the frequency slope fitting method may have lower estimation accuracy. In these cases, least squares ("LSM") may be applied to improve the estimation accuracy. In LSM, the natural log-backward spectral ratio (i.e., ln (Rs (f)) _i,z _k) ) may be adapted to a single-parameter model or a multi-parameter model. The model may automatically search for the best solution to minimize the estimation error. Two parametric models and a single parametric model may be provided as two non-limiting examples.

In a two-parameter model, the transducer response at each frequency may be measured or calibrated in advance. Then, from equation (12), the following expression can be derived:

(f _i-f _i-1)·(n _b-4az _k)＝ln(Rs(f _i,z _k))-ln(R _G(f _i)) (14)。

equation (14) can be simplified based on the following expression,

A(f _i,z _k)＝[(f _i-f _i-1)-4(f _i-f _i-1)z _k](15)；

V(f _i,z _k)＝[ln(Rs(f _i,z _k))-ln(R _G(f _i))](17)；

wherein A is a size N _i,kX 2 matrix, for a total number of N used in the estimation _i,kFrequency component f of _iAnd depth z _k(ii) a U is a column vector with n for the parameter to be estimated _bAnd two elements of a; and V is of size N _i,kA column vector of x 1.

Using the expressions in equations (15) - (17), equation (14) can be written as:

AU＝V (18)。

to solve for two unknowns n in equation (18) _bAnd a, V (f) _i,z _k) Two or more measurements of (a) may be sufficient. As one example, this may be accomplished by setting the value at two different values (e.g., f) _i＝f ₁,f ₂) By changing the frequency component, or by changing the depth of measurement (e.g. z) _k＝z ₁,z ₂) Or both, to obtain two measurements. When using a broadband transducer with a large frequency range, or when the ROI has a relatively large depth range (i.e., N) _i,k> 2), n can be estimated _bAnd a plurality of values of a. These n _bAnd the a values may be combined to generate a final estimate. For example, the combination may include a weighted average and an unweighted average or other suitable combination.

Another way to narrow down the estimate to a single solution is to apply a least squares method, such as,

where K and I are the total number of depth and frequency components used in the LSM fitting, respectively.

At two measurements (i.e. N) _i,k2), the solution of equation (19) will correspond to the direct inverse result in equation (18), as follows:

U＝A ^-1V (20)。

for more than two measurements (i.e. N) _i,k(> 2), equation (19) will correspond to the pseudo-inverse result in equation (18) as follows:

U＝[A ^*A] ^-1A ^*V (21)；

wherein A is ^*Is the transpose or conjugate transpose of matrix a.

In this two-parameter model, the backscattering coefficient n can be estimated simultaneously _bAnd the frequency dependence of the attenuation coefficient a.

If the transducer frequency response is still unknown, this term may need to be cancelled in order to estimate the attenuation coefficient. In this case, a single parameter may be usedAnd (4) modeling. In a single parameter model, the model can be first modeled by a reference depth z _rThe value obtained (i.e., ln (Rs (f)) _i,z _r) Where r is a reference depth index) to a spectral ratio Rs (f) _i,z _k) Normalization is performed. Then, since both the transducer frequency response and the backscatter coefficients in equation (5) are independent of depth, they can be cancelled out. Multiple reference depths z _rMay be used to average out errors caused by noise, such as electrical noise. The normalization step can be described as:

taking the natural logarithm of both sides of equation (22) results,

ln(Rs _nor(f _i,z _k,z _r))＝-4af _iz _k+4af _iz _r-4af _i-1z _r+4af _i-1z _k＝-4a(f _i-f _i-1)(z _k-z _r) (23)；

when z is _k≠z _rThen (c) is performed.

To simplify equation (23), in this single parameter model, the following expression may be used:

V(f _i,z _k,z _r)＝[ln(Rs _nor(f _i,z _k,z _r))](24)；

A(f _i,z _kz _r)＝[-4(f _i-f _i-1)(z _k-z _r)](25)；

U＝a

in this single parameter model, both V and A are of size N _i,k,rTwo column vectors of x 1, where N _i,k,rIs f used in the estimation model _i、z _kAnd z _rAnd U is a scalar with a single element equal to the attenuation coefficient a. Equation (24) can then be written as:

V＝AU (26)。

in this single-parameter tissue model, there is only one unknown a to beAnd (6) solving. V (f) _i,z _k,z _r) One of the values should generate an estimate of the attenuation coefficient a. If multiple measurements are available, multiple attenuation coefficient values may be estimated and combined to generate a final estimate of the attenuation coefficient, such as by using weighted or unweighted combining or averaging.

In some implementations, the LSM method can be applied to narrow the estimate down to a single solution, such as,

when z is _k≠z _rWhen the current is over; and wherein R is the total number of reference depths used in the least squares fit. At one measured value (i.e., N) _i,k,r1), the solution of equation (27) will correspond to the direct inverse result in equation (26), as follows:

U＝A ^-1V (28)。

for more than one measurement (i.e. N) _i,k,r(> 1), equation (27) will correspond to the pseudo-inverse result in equation (26) as follows:

U＝[A ^*A] ^-1A ^*V (29)。

to stabilize the estimates in LSM, the model may be subject to certain constraints, as follows:

a _min≤a≤a _max(30)。

n _b,min≤n _b≤n _b,max

may be based on the physical properties of different physical media, such as soft tissue, as a and n _bA real boundary is selected. For example, the relative safety range constrained by the liver attenuation coefficient may be selected as:

0≤a _liver(dB/cm/MHz)≤2 (31)。

maximum and minimum constraints may also be adaptively determined based on the acquired data. For example, when using a single parameter model, since V (f) _i,z _k,z _r) Can generate an attenuation coefficient value a, so that all measured values can be used firstAll possible attenuation coefficient values are calculated as follows:

or the like, or, alternatively,

U＝V./A (33)；

where "/" represents division by element. As a non-limiting example, the distribution of the attenuation coefficient values may be obtained by a histogram of all attenuation coefficient values.

Figure 2 shows an example of a histogram obtained from a 0.5dB/cm/MHz phantom. In this example, the bin width (bin width) of the histogram is set to 0.1 dB/cm/MHz. The cut-off boundaries on both sides were selected at 75% of the maximum value (indicated by the red dashed line). In FIG. 2, a _min0.3dB/cm/MHz and a _max0.8 dB/cm/MHz. These values may then be used as constraints for the search range in the single parameter model LSM.

In practice, bin width and cutoff boundaries can be flexibly adjusted to accommodate different conditions. Different cut-off determination methods may also be applied, such as applying curve fitting to the histogram. As a non-limiting example, a gaussian function may be fitted to the histogram in fig. 2. The peak position of the fitted gaussian function may be considered the best estimate of the attenuation coefficient (thus providing another method for estimating the attenuation coefficient), and the lower and upper limits of the attenuation coefficient for the LSM search may be determined from the fitted gaussian function (e.g., when the gaussian function falls to 75% of its peak).

The two-parameter model constraints may be determined in a similar manner.

As one non-limiting example, for quality control, the residual error of a single or two-parameter least squares model may be used as a metric to indicate the reliability of the attenuation estimate.

Clinically, during an ultrasound scan, the body wall often causes reverberation artifacts to the underlying soft tissue. In the presence of these reverberation artifacts, accurate attenuation coefficient estimation can be challenging. In these cases, the methods described in this disclosure may be combined with harmonic imaging to reduce reverberation artifacts. For example, B-mode images may be acquired using a harmonic imaging mode with specially designed pulses (e.g., pulse inversion, amplitude modulation, pulse inversion amplitude modulation, or other suitable methods) or using filter-based harmonic imaging. After pulse recombination, the methods described in this disclosure may be applied to estimate the attenuation coefficient, the backscatter coefficient, or both.

Clutter filters may also be used with multiple frame acquisitions to suppress reverberation artifacts. As a non-limiting example, the unwanted signals may be rejected using a clutter filter based on singular value decomposition as described in co-pending patent application (PCT/US 2017/016190, U.S. patent application No.15/887,029, which is incorporated herein by reference in its entirety). Reverberation from the body wall will typically have a different motion pattern (e.g., less motion due to respiration compared to actual tissue motion). When imaging the target, the sonographer may also indicate small movements, such as by slightly translating the transducer. Reverberation from the body wall will have a more fixed motion pattern compared to the underlying tissue due to a more fixed position relative to the transducer surface. In both examples, reverberation may be identified and rejected by the clutter filter.

For a more general model, tissue attenuation can be written as:

wherein n is _aIs the decay frequency dependence. The above example assumes n _a1 is ═ 1; however, the method is also readily applicable to n _aNon-linear case of not equal to 1. For example, in a frequency slope fitting method, equation (8) may be modified to:

there is no closed form solution for solving equation (35). As one non-limiting example, a variable projection ("varpor") method may be used to estimate the attenuation coefficient. Taking the single parameter LSM model as an example, equation (34) may be replaced in equation (22) to yield:

taking the natural logarithm of both sides of equation (36) results,

in the case of single parameter non-linearity, this results in:

AU＝V (38)；

wherein the content of the first and second substances,

V(f _i,z _k,z _r)＝ln(Rs _nor(f _i,z _k,z _r)) (39)；

U＝a

due to U and n in equation (38) _aBoth are unknown, so their joint estimates correspond to non-linear least squares estimates. However, due to the separable structure of equation (38), these unknown parameters may be sequentially determined using the varro technique.

In such an implementation, for a given n _aFirst, a U analysis is determined as n using a common (e.g., linear) least squares method _aFunction of (c):

or

U(n _a)＝(A ^*(n _a)A(n _a)) ^-1A ^*(n _a)V (41)。

As one non-limiting example, equation (41) corresponds to the pseudo-inverse result (i.e., more than one measurement) in equation (40).

Then, the obtained expression U (n) _a) Embedding into a non-linear least squares residual J (n) _a) In order to obtain the compound of (1),

then, n _aCan be implemented in small steps deltan _aMove to obtain the updated value,

n′ _a＝n _a+Δn _a#。

update value n' _aWill be repeatedly updated as described above until it covers the whole predefined search range n _a,min≤n _a≤n _a,maxUntil now. May be based on a known property of soft tissue as n _aSelecting an actual search range, such as 1 ≦ n _aLess than or equal to 2. For a non-limiting example, n _aIt may be moved from a minimum limit to a maximum limit or in the opposite direction.

Linear search n within a predefined search range _aThen, when the residual J (n) _a) When the minimum value is reached, n can be identified _aIs measured. Once n has been identified _aThe optimum value of (c) is selected to use the n _aThe corresponding attenuation coefficient a of the value estimate is taken as the final estimate. For a two-parameter LSM model, n may be similarly determined via VARPRO _aAnd then n is _aFor n _bAnd in the derivation of the best estimate of a,

referring now to FIG. 3, a flowchart is shown illustrating steps of an example method for estimating a value of an acoustic property of a physical medium using a reference frequency method. The physical medium may be tissue, such as soft tissue in a subject, and the acoustic property value may be an attenuation coefficient value, a backscatter coefficient value, or both.

The method includes providing ultrasound data to a computer system for processing, as indicated at step 302. Providing ultrasound data may include retrieving previously acquired data from a memory or other data store, or may include acquiring data with the ultrasound system and transferring the data to a computer system, which may form part of the ultrasound system.

Generally, as described above, the ultrasound data includes first ultrasound data acquired at a first frequency and second ultrasound data acquired at a second frequency spectrally adjacent to the first frequency. Both the first ultrasound data and the second ultrasound data are acquired at a plurality of different depth positions in a region of interest containing at least one physical medium whose acoustic property values are to be estimated. Ultrasound data may be acquired over a range of different frequency values, such as a plurality of different first and second frequency values.

The spectral ratio is calculated using the first and second ultrasound data, as indicated at step 304. For example, the spectral ratio may be calculated as a ratio between the first ultrasound data and the second ultrasound data acquired from the same depth position. As indicated at step 306, a logarithm of each spectral ratio value is calculated to generate a logarithm value, and as indicated at step 308, an acoustic property value is calculated based at least in part on the logarithm value. As described above, the log value may be a natural log value; it should be noted, however, that logarithms other than natural logarithms may be used in the methods described in this disclosure. For example, in the above method, a base-10 logarithm may be used instead of a natural logarithm for estimating the attenuation coefficient and the backscatter coefficient in a similar manner.

In some implementations, the acoustic property values include attenuation coefficient values, and estimating the attenuation coefficient values includes forming a spectral ratio decay curve by plotting natural log values as a function of depth. The slope of this curve may be calculated and the attenuation coefficient value estimated based on the slope. The slope may be calculated using a linear fit or a non-linear fit. When non-linear fitting is used, a method such as the varpor method may be performed.

A plurality of such curves may be calculated for spectral ratio values calculated based on ultrasound data acquired at different frequencies, and the attenuation coefficient values calculated for these plurality of different curves may be combined to generate a final estimate of the attenuation coefficient. The attenuation coefficient values may be combined, for example, using an average, which may be a weighted average.

The y-intercept of such a curve may be calculated and the backscatter coefficients may be estimated based on the y-intercept values. In these implementations, the transducer responses may also be provided to a computer system and used to weight the spectral ratios.

In some other implementations, estimating the acoustic property values includes fitting natural log values to a parametric model of the spectral ratios. The parametric model may be a single parametric model, in which case the acoustic property values may comprise attenuation coefficients. The parametric model may also be a two-parameter model, in which case the acoustic property values may comprise an attenuation coefficient and a backscatter coefficient. Other parametric models may also be implemented. Fitting the natural logarithm to the parametric model may include using a least squares estimation. The least squares estimation may be constrained using one or more constraints, such as constraints on expected acoustic property values of the physical medium, or constraints on adaptive data generation.

The description provided above assumes that the transmitted ultrasound beam is perpendicular to the transducer surface. However, the methods described in this disclosure are applicable to steered ultrasound beams. In the case of steered beams, the above "normalization along depth" may be understood as a more general normalization along the ultrasound path. That is, depth and depth position may be understood to mean a position along the ultrasound path.

As another example, ultrasonic beams steered at different angles may be used to estimate acoustic properties of the same physical region. These estimates may be combined (e.g., using a weighted or unweighted average) to provide a final estimate of the acoustic properties.

After the acoustic property values have been estimated, a report may be generated based on the acoustic property values, as indicated at step 310. As one example, the report may comprise a digital image depicting a spatial distribution of acoustic property values in the region of interest. As another example, the report may include a data graph, textual information, or other form of visually depicting or representing the acoustic property values.

Fig. 4 illustrates an example of an ultrasound system 400 in which the methods described in this disclosure may be implemented. The ultrasound system 400 includes a transducer array 402, the transducer array 402 including a plurality of individually driven transducer elements 404. The transducer array 402 may include any suitable ultrasound transducer array, including a linear array, a curved array, a phased array, and so forth. Similarly, the transducer array 402 may include 1D transducers, 1.5D transducers, 1.75D transducers, 2D transducers, 3D transducers, and so forth.

A given transducer element 404 produces a burst of ultrasonic energy when excited by a transmitter 406. Ultrasound energy (e.g., echoes) reflected back to the transducer array 402 from the subject or subject under study are converted to electrical signals (e.g., echo signals) by the individual transducer elements 404 and may be applied individually to the receivers 408 through a set of switches 410. The transmitter 406, receiver 408, and switch 410 operate under the control of a controller 412, which controller 412 may include one or more processors. As one example, the controller 412 may comprise a computer system.

The transmitter 406 may be programmed to transmit unfocused or focused ultrasound waves. In some configurations, the transmitter 406 may also be programmed to transmit a diverging wave, a spherical wave, a cylindrical wave, a plane wave, or a combination thereof. In addition, transmitter 406 may be programmed to transmit spatially or temporally encoded pulses.

Receiver 408 may be programmed to perform an appropriate detection sequence for the imaging task of the hand. In some embodiments, the detection sequence may include one or more of progressive scan, complex plane wave imaging, synthetic aperture imaging, and complex divergent beam imaging.

In some configurations, the transmitter 406 and receiver 408 may be programmed to achieve a high frame rate (high frame rate). For example, a frame rate associated with an acquisition pulse repetition frequency ("PRF") of at least 100Hz may be achieved. In some configurations, the ultrasound system 400 may sample and store at least one hundred sets of echo signals in the time direction.

As is known in the art, the controller 412 may be programmed to design an imaging sequence. In some embodiments, the controller 412 receives user input defining various factors used in the design of the imaging sequence.

Scanning may be performed by setting switches 410 to their transmit positions, thereby instructing transmitters 406 to be temporarily turned on to energize transducer elements 404 during a single transmit event according to an imaging sequence. The switches 410 may then be set to their receive positions and subsequent echo signals produced by the transducer elements 404 in response to one or more detected echoes are measured and applied to the receiver 408. The separate echo signals from the transducer elements 404 may be combined in the receiver 408 to produce a single echo signal.

The echo signals are communicated to a processing unit 414 to process the echo signals or images generated from the echo signals, the processing unit 414 may be implemented by a hardware processor and memory. As an example, the processing unit 414 may estimate acoustic properties of a physical medium, such as soft tissue, using the methods described in this disclosure. The image produced by the processing unit 414 from the echo signals may be displayed on a display system 416.

Referring now to fig. 5, there is shown a block diagram of an example of a computer system 500 that may perform the methods described in this disclosure. The computer system 500 generally includes an input 502, at least one hardware processor 504, a memory 506, and an output 508. Thus, the computer system 500 is typically implemented with a hardware processor 504 and a memory 506. In some examples, computer system 500 may also be implemented by a workstation, a notebook computer, a tablet device, a mobile device, a multimedia device, a web server, a mainframe, one or more controllers, one or more microcontrollers, or any other general purpose or special purpose computing device.

The computer system 500 may operate autonomously or semi-autonomously, or may read executable software instructions from memory 506 or a computer-readable medium (e.g., hard drive, CD-ROM, flash memory), or may receive instructions from a user or any other source logically connected to a computer or device, such as another networked computer or server, via input 502. Thus, in some embodiments, computer system 500 may also include any suitable device for reading computer-readable storage media.

Generally, the computer system 500 is programmed or otherwise configured to implement the methods and algorithms described in this disclosure. For example, computer system 500 may be programmed to estimate acoustic properties of a physical medium, such as soft tissue, using the methods described in this disclosure.

Input 502 can take any suitable shape or form as desired for operation of computer system 500, including the ability to select, input, or otherwise specify parameters consistent with performing tasks, processing data, or operating computer system 500. In some aspects, the input 502 may be configured to receive data, such as data acquired with an ultrasound system. Such data may be processed as described above to estimate acoustic properties, such as attenuation coefficients, backscatter coefficients, or both, or physical media, such as soft tissue. Further, the input 502 may also be configured to receive any other data or information deemed useful for estimating acoustic properties of a physical medium, such as soft tissue, using the methods described above. For example, as described above, the input 502 may also be configured to receive transducer response data or reference depth data.

In processing tasks for operating the computer system 500, the one or more hardware processors 504 may also be configured to perform any number of post-processing steps on data received through the input 502.

The memory 506 may contain software 510 and data 512, such as data acquired with an ultrasound system, and may be configured to store and retrieve processed information, instructions, and data to be processed by the one or more hardware processors 504. In some aspects, software 510 may include instructions that direct the estimation of acoustic properties of a physical medium, such as soft tissue, from ultrasound data.

Further, the output 508 may take any shape or form as desired, and may be configured to display images, maps, data maps, textual information, or other visual depictions or representations of acoustic characteristics in addition to other desired information.

The present disclosure has described one or more preferred embodiments, and it should be understood that many equivalents, alternatives, variations, and modifications, other than those expressly stated, are possible and are within the scope of the invention.