Method and system for estimating ultrasound attenuation parameters

Abstract

The present invention relates to a method for estimating ultrasonic attenuation parameters of a region in a medium, wherein the method is implemented by a processing unit (8) associated with at least one ultrasonic transducer (2), the method comprising the following steps: (a) a transmission step (101), wherein at least one pulse is transmitted in the medium by the transducer; (b) a receiving step (102), wherein data is acquired by the transducer in response to the pulse; (c) a processing step (103), wherein the data is processed by the processing unit to provide backscatter acquisition data of the region; (d) a function determination step (104), wherein an autocorrelation function of the backscatter acquisition data is determined, the autocorrelation function being a function of depth in the spatiotemporal domain, the autocorrelation function being determined when the lag is zero; (e) an attenuation estimation step (105), wherein an ultrasonic attenuation parameter is estimated based on the autocorrelation function. The present invention also relates to a corresponding system.

Description

Technical Field

[0001] The present invention relates to an imaging method and an apparatus for implementing the method, in particular a method for medical imaging.

[0002] The present disclosure relates in particular to a method for estimating ultrasonic attenuation parameters of a region in a medium, and more specifically to a method for estimating and imaging local ultrasonic attenuation parameters of a region inside a medium. The method may be implemented by a processing unit associated with a plurality of transducers (e.g., transducers in a row or array) associated with the medium. Background Art

[0003] Conventional ultrasound imaging consists of insonification of the medium with a cylindrical wave focused on a given point. By using the backscattered echoes of this single insonification, a complete line of the image can be calculated using a dynamic receive beamforming process. To build a complete image, this process is repeated by sending a set of focused waves that scan along the side lines at a given depth (called the focal plane). For each focused wave, dynamic beamforming is performed, building up line by line to obtain a complete image. Dynamic beamforming guarantees uniform focusing in receive mode, while in transmit mode the focus is fixed at a given depth. The final image is optimal in the focal plane and in a limited area of ​​the medium corresponding to the axial focal length. However, outside this area where the diffraction laws act, the image quality at other depths (in the near and far fields of the focused beam) degrades rapidly.

[0004] To overcome this limitation, the classic solution is to perform multifocal imaging: different transmission focal depth arcs are used to obtain uniform quality across the image. Transmission at each given focal depth enables local imaging to be performed in the region defined by the axial focal length. The final image is obtained using a recombination of these local images corresponding to the various depths. Optimal multifocal images typically require dozens of focal planes. This results in a frame rate limitation, typically less than 10 frames / second, which is unacceptable for ultrasound imaging. A good compromise between image quality and frame rate is images of approximately 4 focal depths.

[0005] It is envisioned that image quality could be improved by performing synthetic dynamic transmit focusing. This approach involves resynthesizing dynamic transmit focusing (ie, as many focal depths as pixels in the image) through beamforming and then combining a set of different insonifications.

[0006] In addition, an improved synthetic ultrasound imaging method is known from EP2101191 (A2) filed by the applicant, which can improve the traditional plane wave synthetic ultrasound imaging method, such as the method disclosed in US patent application US-6551246.

[0007] European patent application EP2101191 (A2) proposes a method for ultrasound imaging, the method comprising at least the following steps:

[0008] a) a transmitting step, wherein a plurality of ultrasonic waves are transmitted into an imaging region, and a corresponding group of raw data is collected by a transducer array in response to each ultrasonic wave, wherein for each of a plurality of imaging positions in the region, the ultrasonic wave has a different spatial spectrum, and each group of raw data represents a time domain signal received by the transducer in response to the corresponding ultrasonic wave;

[0009] b) a coherence enhancement step, wherein, for each of a plurality of virtual transmission focal regions in the imaging region, at least one set of coherent data is synthesized from the original data set;

[0010] c) a beamforming step, wherein, for each of a plurality of positions included in each virtual transmission focal region, an image pixel is calculated by beamforming using the set of coherent data.

[0011] Due to these configurations, the spatial coherence of the raw data is restored in step b) before beamforming, thereby enabling accurate combination of the data received from the transmission of the various ultrasound waves. The need to restore the spatial coherence is due to the fact that, when the imaging region is illuminated with a wavefield that is widely spread in space, the echoes returning from the medium can be considered as wavefields that originate from incoherent sources (scatterers) and are randomly distributed in the imaging region: therefore, the spatial coherence of the wavefield is lost (or very poor) in the raw data. Beamforming can then be performed on the coherent data resulting from the coherence restoration step, resulting in a more accurate image.

[0012] Based on this technology, which can be called synthetic beamforming, WO2017093778 (A1) discloses another imaging method for generating an image of the inner area of ​​a medium with less speckle noise. To this end, the method comprises the following steps:

[0013] (a) a transmitting step, wherein a first plurality of waves is transmitted within a medium via a transducer;

[0014] (b) a receiving step, wherein a set of data is collected by the transducer in response to the wave;

[0015] (c) a beamforming step, wherein the set of data is processed by a second plurality of beamforming processes to provide beamformed pixel values ​​for at least a portion of the image, each beamforming process using a set of data corresponding to a wave generated using a transmission weighting vector or using the transmission weighting vector in a beamforming pixel value operation; and

[0016] (d) a combining step, wherein the second plurality of beamformed pixel values ​​are combined to provide a pixel value for a pixel within an image, and wherein the transmit weighting vectors are different and orthogonal to each other.

[0017] Due to these characteristics, each transmission weighting vector produces uncorrelated speckle noise, and the combination of weighted data allows the calculation of images of areas with less speckle noise.

[0018] However, another phenomenon that needs to be considered in ultrasound imaging is the ultrasound attenuation in the examined medium. Ultrasound attenuation directly affects the response to the transmitted ultrasound waves. Attenuation thus constitutes a delicate frequency and depth-dependent phenomenon. It is therefore desirable to compensate for any effects of attenuation on the resulting computed image, typically by accounting for tissue attenuation, for example, through time gain compensation.

[0019] On the other hand, reliable estimates of ultrasound attenuation can be used for other ultrasound diagnostic purposes. For example, global attenuation parameters estimated for the examined medium (e.g., human liver or muscle) can be used to determine its fat content. Furthermore, local attenuation distributions (e.g., after processing to calculate an image) can be used to detect cancer (e.g., human breast cancer).

[0020] However, conventional methods for estimating ultrasound attenuation usually employ frequency (or spectrum) analysis. However, such analysis implies several disadvantages.

[0021] Since attenuation is a subtle frequency and depth dependent phenomenon, accurate evaluation requires sufficiently large spatial and frequency observation ranges. This means that one would like to use large frequency and spatial ranges. However, due to attenuation, the larger the spatial range, the more variable the frequency content and the larger the deviation. On the other hand, the smaller the spatial range, the worse the spectral analysis. Moreover, the larger the spatial range, the greater the effect of diffraction. In addition, speckle noise increases the variance of attenuation estimates.

[0022] Therefore, taking into account this necessary trade-off, estimating ultrasound attenuation by frequency (or spectrum) analysis leads to imprecise results while requiring a significant computational cost (particularly due to the need for Fourier transformation). As a result, there is the problem of increased variance of possible results and therefore reduced reliability of the estimate, i.e. reduced reproducibility. Therefore, a trade-off must be made between acquisition time and level of accuracy.

[0023] A. Lymberis et al. proposed a method for estimating frequency-dependent attenuation. In particular, in the proposed method 2, an average frequency estimate is obtained from all available autocorrelation lags (ACn) of the demodulated signal, see LYMBERIS et al.: "Estimation of frequency-dependent attenuation based on parametric spectral analysis and correlation lags of the demodulated echo signal", ULTRASONIC IMAGING, DYNAMEDIA INC., SILVER SPRING, MD, US, Vol. 13, No. 1, January 1, 1991, pp. 1-26, XP026409816, ISSN: 0161-7346.

[0024] In addition, Hyungsuk et al. disclosed an attenuation estimation method using spectral cross-correlation, see HYUNGSUKKIM et al.: “Attenuation estimation using spectral cross-correlation”, IEEE TRANSACTIONS ON ULTRASONICS, FERROELECTRICS AND FREQUENCY CONTROL, IEEE, US, Vol. 54, No. 3, March 1, 2007, pp. 510-519, XP011175817, ISSN: 0885-3010, DOI: 10.1109 / TUFFC.2007.274. Summary of the invention

[0025] There is still a desire to overcome the above-mentioned problems, in particular to provide a method and system for reliably estimating ultrasonic attenuation parameters of a region in a medium, which may advantageously be faster, less computationally expensive and thus require less processing power.

[0026] Therefore, according to an embodiment of the present disclosure, a method for estimating ultrasonic attenuation parameters of a region in a medium is provided. The method is implemented by a processing unit associated with at least one ultrasonic transducer (the ultrasonic transducer may be associated with the medium). The method comprises the following steps:

[0027] (a) a transmitting step, wherein at least one pulse is transmitted in a medium by means of a transducer;

[0028] (b) a receiving step, wherein data is collected by the transducer in response to the pulse;

[0029] (c) a processing step, wherein the data is processed by the processing unit to provide backscatter acquisition data for the area;

[0030] (d) a function determination step, wherein an autocorrelation function of the backscatter acquisition data is determined, the autocorrelation function being a function of depth in the spatiotemporal domain, the autocorrelation function being determined when a lag is zero (0);

[0031] (e) An attenuation estimating step, wherein an ultrasonic attenuation parameter is estimated based on the autocorrelation function.

[0032] The method may also be referred to as a global attenuation estimation method, eg a global attenuation estimation method for one attenuation parameter for the complete region under examination, eg for the complete ultrasound image.

[0033] By providing such a method, it becomes possible to estimate ultrasound attenuation parameters by performing the analysis in the spatiotemporal domain rather than in the frequency (or spectral) domain. Thus, advantageously, the estimate is more accurate and implies lower computational costs, for example, which improves real-time (or at least quasi-real-time) computational modes. Furthermore, due to the increased accuracy, the variance is reduced, thereby enabling improved reproducibility.

[0034] The processing step (c) may include a beamforming step, in which the data is processed by a beamforming process to provide beamformed acquisition data of the region. For example, the beamformed acquisition data or backscattered acquisition data may be presented in the form of pixel values ​​or voxel values.

[0035] Through the process of beamforming, the diffraction pattern in the acquired data can be reduced.

[0036] For example, the beamforming process may be a synthetic beamforming process. This allows the diffraction pattern to be further reduced.

[0037] In addition, the beamforming process of step c) may be configured to further reduce speckle noise. For example, step (c) may include:

[0038] (c1) a beamforming step, wherein the data is processed by a plurality of beamforming processes to provide beamformed acquisition data for at least the region, each beamforming process using data corresponding to pulses generated using a transmission weighting vector or using the transmission weighting vector in calculation of the beamformed acquisition data, and

[0039] (c2) A combining step, wherein the plurality of beamformed acquisition data are combined to provide pixel values ​​for pixels within a generated ultrasound image, and wherein the transmission weighting vectors are different and orthogonal to each other.

[0040] However, alternatives to the above-described beamforming process may be used to provide backscatter acquisition data. For example, the transmit pulses (and optionally the receive signals) may be acoustically beam shaped by an acoustic lens or mechanical sweeping / scanning. In such cases, the beamforming process may be omitted in processing step (c).

[0041] In the attenuation estimation step (e), the ultrasound attenuation parameter may be estimated over an interval Δz in the depth direction. For example, the interval Δz may be predefined between a first specified depth and a second specified depth in the medium. It should be noted that the depth in the medium may be defined as the distance between the probe surface and a point of interest (spatial position).

[0042] In the attenuation determination step, for example, for the interval Δz, the ultrasound attenuation parameter may be determined by solving an autocorrelation function.

[0043] The autocorrelation function of the backscatter acquisition data can be a function of depth and time.

[0044] In the function determination step (d), residual diffraction patterns in the backscatter acquisition data may be compensated based on predefined diffraction reference data.

[0045] Residual diffraction patterns in backscatter acquisition data are compensated by applying the method comprising steps (a') to (e') on a reference medium having predetermined reference ultrasound attenuation parameters and comparing autocorrelation functions.

[0046] In other words, the estimation method can be applied to a reference medium, such as a phantom, having predetermined (i.e. known) ultrasound attenuation parameters. Thus, the determined autocorrelation function (which can be referred to as the reference autocorrelation function) can be solved to determine the residual diffraction pattern in the backscattered acquisition data and / or compensate for it in the autocorrelation function of the medium including the region of interest to be examined.

[0047] Comparing the autocorrelation functions may include calculating a delta function between the two functions and solving the delta function to determine the ultrasonic attenuation parameter of the medium.

[0048] For example, the reference ultrasonic attenuation parameter may be a parameter of an interval Δz′ in a reference medium that is similar to the interval Δz in the medium to be examined.

[0049] The predetermined diffraction reference data may include predetermined parameters for different depths in the reference medium.

[0050] The pulse may be a broadband pulse and / or a plurality of pulses. For example, the use of one or more broadband pulses may improve the estimation accuracy, as it allows for spatial confinement, ie a more precise interval determination.

[0051] The present disclosure also relates to a method for estimating local ultrasonic attenuation parameters of a region in a medium. The method may include the above-mentioned global attenuation estimation method (i.e., steps (a) to (e)). For each of the multiple axes in the depth direction z in the medium (e.g., the lines scanned in steps (a) to (c)), the local ultrasonic attenuation B in the equation F(Ey(z), B)=0 is estimated by at least one of the following methods, where Ey is the autocorrelation function and z is the depth:

[0052] - modeling the autocorrelation function as a noisy state-space model with a driving input,

[0053] - regularizing the autocorrelation function by a regularization method, the regularization method comprising: a Tikhonov regularization method, a ridge regression method and / or a Bayesian regression method,

[0054] - filtering of the autocorrelation function using Savitzky-Golay filters, Monte Carlo methods, singular spectrum analysis, and

[0055] Compressed sensing and / or sparse sampling methods.

[0056] The method may also be referred to as a local attenuation estimation method, e.g. a local attenuation estimation method for a plurality of attenuation parameters for the entire region under examination (e.g. for one ultrasound image). The parameters may be assigned to pixels or voxels in the ultrasound image. They may also form segments or superpixels or clusters in the ultrasound image of an additional attenuation map of the region under examination.

[0057] By this method it is made possible to estimate the local ultrasound attenuation parameters of the region of interest, so that a map and / or image(s) of the attenuation characteristics of the entire region can be created.

[0058] Compared to conventional inversion methods that may be used to solve the equation F(Ey(z),B), the proposed methods imply less computational effort since they converge faster. At the same time, they make the results more precise, for example by making the spatial resolution of the local estimates obtained higher.

[0059] The autocorrelation function can be discretized in the depth direction (z) to obtain the local attenuation function.

[0060] The drive input may include a predetermined attenuation parameter of the reference medium.

[0061] The modeled state-space model may use a Kalman filter or a predefined Bayesian method to determine the local ultrasound attenuation parameters.

[0062] Using a Kalman filter may have the advantage of optimizing the minimum mean square error criterion.

[0063] The reflectivity pattern of the medium can be compensated by iterating the EM algorithm on the modeled state space model until a predefined convergence threshold is obtained. For example, the EM algorithm can use a first Kalman filter that outputs a reflectivity parameter and can use an ultrasonic attenuation parameter as a driving input. A second Kalman filter can output an ultrasonic attenuation parameter and can use the reflectivity parameter as a driving input. The two filters can be connected to use the output of the other filter as a driving input, for example, until a predefined convergence threshold is obtained.

[0064] The present disclosure also relates to a method for imaging local ultrasound attenuation in a medium region. The method may include the local attenuation estimation method as described above. An attenuation image may be constructed based on local attenuation parameters (eg, for each axis).

[0065] The method for imaging local ultrasonic attenuation within a medium region may further comprise the following steps: feeding the attenuation image into a machine learning model, which is trained to segment the image into one or several regions of interest, each of which is associated with at least one predefined category.

[0066] The machine learning model may include, for example, an artificial neural network, such as a convolutional neural network (the model may be, for example, implemented by a computer). The model may be trained by a supervised training method using, for example, separately labeled images as training data, or by an unsupervised training method.

[0067] Finally, the present disclosure relates to a system for estimating ultrasonic attenuation parameters of a region in a medium, comprising a processing unit associated or connected to at least one ultrasonic transducer (e.g., associated with the medium). The system or processing unit is configured to:

[0068] (a) transmitting at least one pulse in the medium via a transducer;

[0069] (b) acquiring data in response to the pulse via a transducer;

[0070] (c) processing the data by the processing unit to provide backscatter acquisition data for the area;

[0071] (d) determining an autocorrelation function of the scatter acquisition data, the autocorrelation function being a function of depth in the spatiotemporal domain, the autocorrelation function being determined at a lag of zero (0);

[0072] (e) Estimate the ultrasonic attenuation parameter based on the autocorrelation function.

[0073] The system may also optionally include at least one transducer.

[0074] The at least one transducer may be a single transducer configured to transmit pulses and receive tissue responses. For example a focused transducer having, for example, a concave form or a corresponding lens. It may also be a scanning single transducer.

[0075] Multiple transducers and / or transducer arrays may also be used. For example, a linear array may be provided, which typically includes hundreds of transducers (e.g., 100 to 300) juxtaposed along an axis X (horizontal or array direction X). The present disclosure may also be implemented with a 3D probe.

[0076] The same transducer(s) may be used to send the pulse and receive the response, or different transducers may be used for sending and receiving.

[0077] The present disclosure may also relate to a computer program, which includes instructions for executing the steps of at least one of the above methods when the program is executed by a computer.

[0078] Finally, the present disclosure may also relate to a computer-readable recording medium having a computer program recorded thereon, wherein the computer program includes instructions for executing the steps of at least one of the above methods when the program is executed by a computer.

[0079] Unless otherwise contradictory, the above elements may be combined with elements within the specification.

[0080] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are provided for the purpose of illustration only and are not restrictive of the present disclosure.

[0081] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present disclosure and, together with the description, serve to support and explain the principles thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 A schematic diagram of an ultrasonic device according to an embodiment of the present disclosure is shown;

[0083] Figure 2 Shows Figure 1 a block diagram of a portion of a device;

[0084] Figure 3 It is shown that according to the present disclosure and Figure 1 A diagram of a method for estimating ultrasound attenuation parameters implemented in a device of FIG.

[0085] Figure 4 A first exemplary embodiment of a method according to the present disclosure (global attenuation estimation) is shown;

[0086] Figure 5 Shown by Figure 3 or Figure 4 Examples of ultrasound images obtained by the method;

[0087] Figure 6 A second exemplary embodiment of a method according to the present disclosure is shown (local attenuation estimation). DETAILED DESCRIPTION

[0088] Reference will now be made in detail to exemplary embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings. Wherever possible, the same reference numerals will be used throughout the drawings to refer to the same or like parts.

[0089] Figure 1 The device shown is suitable for imaging a region 1 of a medium, such as living tissue, in particular human tissue of a patient. For example, the device may include:

[0090] · At least one transducer, for example a single transducer configured to send pulses and receive tissue responses. In addition, multiple transducers and / or transducer arrays 2 may be used. For example, a linear array may be provided, typically comprising hundreds of transducers (e.g. 100 to 300), which are juxtaposed along an axis X (horizontal or array direction X), as is known in conventional probes. In this case, the array 2 is adapted to perform two-dimensional (2D) imaging of the region 1, but the array 2 may also be a two-dimensional array adapted to perform 3D imaging of the region 1. The transducer array 2 may also be a convex array comprising multiple transducers arranged along a curve. The same (one or more) transducers may be used to send pulses and receive responses, or different transducers may be used for sending and receiving;

[0091] ● Electronic cabinet 3, controlling the transducer array and collecting signals therefrom;

[0092] A microcomputer 4 for controlling the electronic cabinet 3 and, for example, viewing images obtained from the electronic cabinet (in a variant, a single electronic device can fulfil all the functions of the electronic cabinet 3 and the microcomputer 4).

[0093] Figure 1 The axis Z in is the axis perpendicular to the axis X, and it is usually the direction of the ultrasound beam generated by the transducer of the array, for example in the depth direction of the inspected medium. This direction is designated as the vertical direction or axial direction in this article.

[0094] For example, Figure 2 As shown, the electronic cabinet 3 may include:

[0095] L analog-to-digital converters 5 (A / Di-A / DL), respectively connected to the L transducers (T1-TL) of the transducer array 2;

[0096] L buffer memories 6 (Bi-Bn), connected to n analog-to-digital converters 5 respectively;

[0097] A central processing unit 8 (CPU), communicating with the buffer memory 6 and the microcomputer 4;

[0098] A memory 9 (MEM) connected to the central processing unit 8;

[0099] A digital signal processor 10 (DSP) connected to the central processing unit 8.

[0100] The device disclosed in this article is a device for ultrasonic imaging, the transducer is an ultrasonic transducer, and the implemented method estimates the ultrasonic attenuation parameter of the area 1, and can optionally generate an ultrasonic image of the area 1.

[0101] However, the device may be any imaging device using other waves (waves having a wavelength different from that of ultrasonic waves) than ultrasonic waves, so that the transducer and electronic cabinet components are adapted to said waves.

[0102] Figure 3 It is shown that according to the present disclosure and Figure 1 A method for estimating ultrasonic attenuation parameters implemented in a device.

[0103] The method steps are mainly controlled by the central processing unit 8, which is ultimately controlled by means of a digital signal processor 10 or any other device. The method comprises the following main steps:

[0104] (a) a transmitting step (101), wherein at least one pulse is transmitted in the medium by means of a transducer;

[0105] (b) a receiving step (102), wherein data is collected by the transducer in response to the pulse;

[0106] (c) a processing step (103), wherein the data is processed by the processing unit to provide backscatter acquisition data for the region;

[0107] (d) a function determination step (104), wherein an autocorrelation function of the backscattered acquisition data is determined, the autocorrelation function being a function of depth in the spatiotemporal domain;

[0108] (e) An attenuation estimation step (105), wherein an ultrasonic attenuation parameter is estimated based on the autocorrelation function.

[0109] Optionally, an ultrasound image may also be generated based on the backscatter acquisition data of step 103 .

[0110] For example, the estimated ultrasound attenuation parameters can be used to:

[0111] Assessment of hepatic steatosis when a global estimator is provided.

[0112] ● Testing for Duchenne muscular dystrophy carriers when a global estimate is available, and / or

[0113] • When an attenuation map is provided, characterize the breast mass(es).

[0114] These examples are provided as illustrative applications of the methods described in this disclosure, but other applications, such as applications related to the animal body and / or material analysis may also be of interest.

[0115] Steps 101 to 105 may optionally form a loop, for example by returning from step 105 to 101 to form loop 107. This may allow for the estimation of multiple attenuation parameters, wherein, as an option in step 103, in each cycle an ultrasound image may additionally be generated. By returning from step 105 to step 104, loop 108 may be added to loop steps 104 and 105 in addition or alternatively. As in Figure 6 As described in more detail in the context of , this may allow estimating multiple local estimates (optionally in different parts of the region of interest).

[0116] In some embodiments, the method can be used to estimate a global ultrasound attenuation parameter (e.g., Figure 4 and 5 ), or for estimating local ultrasound attenuation parameters (as described in more detail in the context of Figure 6 However, before describing the detailed embodiments of the present disclosure in detail, some basic knowledge is provided below.

[0117] The following description uses the following glossary of terms:

[0118] ACF: Autocorrelation function

[0119] β: Ultrasonic attenuation (dB / cm / MHz or Np / cm / MHz)

[0120] c: speed of sound in the medium

[0121] f0: center frequency of the pulse

[0122] λ: Pulse wavelength (λ=c / f0)

[0123] MMSE :Minimum Mean Square Error

[0124] PSD: Power Spectral Density

[0125] ROI: Region of Interest

[0126] RF Data: A two-dimensional array of digitized signals collected on the probe (transducer vs. time)

[0127] σ f : Standard deviation of the Gaussian spectrum

[0128] The following assumes that the medium is homogeneous and contains weak scatterers. Assume that the analytical signal of the transmitted pulse has a Gaussian envelope and a center frequency f0:

[0129]

[0130] In the following, a general signal modeling that may be used in the present disclosure is described. Thus, the signal modeling is progressively enhanced by progressively adding the following to the model: ultrasound echo (ie PSD), ultrasound attenuation, diffraction, reflectivity.

[0131] First, it is described how to model the power signal density (PSD) measured at a pixel. In this description, for simplicity, ultrasound attenuation is not considered. For a focused aperture, the signal backscattered by an isochronous volume at time t can be modeled as:

[0132]

[0133] Among them, n(t) is a zero-mean Gaussian random process with a microscopic autocorrelation function, that is:

[0134] E{n(t)n * (t+τ)}=γδ(τ) (3)

[0135] Therefore, the expression of the autocorrelation function of the backscattered signal can be expressed as follows:

[0136]

[0137]

[0138]

[0139]

[0140]

[0141] Using the Wiener-Kintchine theorem, the PSD S(f) of the backscattered signal can be expressed as:

[0142] S(f)=TF[E{s(t)s * (t+τ)}]

[0143]

[0144]

[0145] The beamforming process consists of integrating the signal acquired at the transducer with a delay corresponding to the propagation time from the pixel of interest to the transducer. The PSD measured at the pixel is therefore given by equation (5).

[0146] Next it is described how ultrasound attenuation can be modeled. In particular, ultrasound attenuation can be modeled as a filter with the following transfer function.

[0147] |H(f,z)|=e -2βfz (6)

[0148] The factor 2 accounts for the round trip propagation. Assuming axial transmission and reception beams, combining (5) and (6) yields an expression for the PSD of the beamformed pixel at depth z:

[0149]

[0150]

[0151] in,

[0152] Equation (7) gives: Applying the Wiener-Kintchine theorem from the spectral domain to the spatiotemporal domain on the PSD yields

[0153] R(τ,z)=E{s(t,z)s * (t+τ,z)}=TF -1 {S(f,z)}

[0154]

[0155]

[0156] The ACF gives the average power of the pixel value at lag τ = 0. Using equation (8) we get:

[0157] E(z)=R(0,z)

[0158]

[0159]

[0160] The transformation from (9) to (10) is reasonable because the right-hand side of (9) is the integral of the Gaussian function from -∞ to +∞, so its integral is 1.

[0161] Next, we describe how the complete pixel signal can be modeled. Considering the transducer sensitivity, beamforming antenna gain, diffraction, and combining equation (10) yields the following expression:

[0162]

[0163]

[0164] Equation (11) represents the core of the method described in this paper. In this framework, constant (z) must be estimated.

[0165] The inventors of the present disclosure have found that the ultrasonic attenuation estimation method based on equation (11) has the following advantages:

[0166] Less sensitive to diffraction effects because:

[0167] o It operates on the synthetic beamforming data so that transmission and reception focus is ensured at any spatial location;

[0168] oIt works in the time and space domain, and the diffraction effects are averaged over the frequency range of the pulse;

[0169] oThe transmission pulse is broadband;

[0170] ●It does not require time (or space) frequency trade-offs like frequency-based methods.

[0171] The local energy estimate of the beamformed data is simple and reliable, so the attenuation estimate is more reliable.

[0172] refer to Figure 4 , a method for estimating global ultrasound attenuation parameters is described. Thus, Figure 4 A first exemplary embodiment of a method according to the present disclosure (global attenuation estimation) is shown. In particular, Figure 4 The method can be Figure 3 A specific embodiment of the method.

[0173] Figure 4 The method begins with a receiving step b, here referenced as 102 and 102', wherein data is collected by the transducer in response to the pulses. This data is referred to herein as a "RF data set" for the inspected medium, and a "RF training data set" for a reference medium (e.g., a phantom). Steps 102 to 104 relate to data processing for the inspected medium, while similar steps 102' to 104' relate to data processing for the reference medium.

[0174] The following steps 103 and 103' "synthesized beamforming" correspond to steps (c) and (d) of the present disclosure, respectively. Figure 3(ie, a processing step) in which the data is processed by a processing unit to provide backscatter acquisition data for the region.

[0175] The following steps 104 and 104' "average power estimation" correspond to steps (d) and (c) of the present disclosure respectively. Figure 3 (ie, function determination step), in which the autocorrelation function of the backscatter acquisition data (particularly the beamforming acquisition data of step 103) is determined, the autocorrelation function being a function of depth in the spatiotemporal domain.

[0176] In a next step 104a, the two autocorrelation functions are compared to each other to compensate for diffraction patterns in the backscatter acquisition data and / or unknown transducer sensitivity. In particular, a delta function can be determined between the two functions. An exemplary embodiment of the comparison step is described below.

[0177] In step 105, the ultrasound attenuation parameter is estimated by solving a comparison of the two autocorrelation functions (ie, in particular their delta functions). In more detail:

[0178] In the following description, the following basic assumptions are made:

[0179] ●The attenuation in the medium is constant, and,

[0180] The constant(z) in equation (11) does not depend on z, ie it is constant over the region of interest ROI.

[0181] Figure 3 and Figure 4 The goal of the method is to estimate β in a specific ROI. The ROI can be a homogeneous part of the liver, such as Figure 5 shown.

[0182] Let z0 and z1 denote the depth of the top and bottom of the ROI (respectively). Depth is the distance between the probe surface and the point of interest (spatial location).

[0183] Evaluating the difference in the logarithm of the mean power at depths z0 and z1 yields

[0184]

[0185] Equation (12) (which can be understood as the result of step 104) includes two unknowns: constant (z1, z2) and β.

[0186] To estimate β, we need to estimate constant(z1, z2). Although constant(z1, z2) depends on several factors, it does not depend on the medium, so it can be learned from a training dataset (i.e., step 102' reference data) where β is known. This can be achieved by acquiring ultrasound data on an attenuation-calibrated phantom (i.e., the reference medium). Let δE ref (z0,z1) represents the difference in logarithm of the mean power measured at depths z0 and z1 of the calibrated phantom, β ref represents its calibrated (known) attenuation, while ΔE(z0,z1) represents δE(z0,z1) and δE ref Then, we can determine

[0187]

[0188] Equation (13) may be an exemplary result of comparison step 104a. Inspection of equation (13) (ie, step 105) reveals that solving for ΔE(z0, z1) = 0 gives an estimate of β.

[0189] Note that equation (13) can be approximated as:

[0190]

[0191] This is because the quadratic term in equation (13) is relatively small and can therefore be ignored.

[0192] Note that an important step of the method is to estimate the average power at depths z0 and z1. The first stage of this estimation process can be to perform synthetic beamforming on the ultrasound radio frequency data acquired in step 102 in step 103, with the goal of minimizing diffraction effects. The second stage can consist of estimating the average power of these beamformed data at a specific depth. However, beamformed data have a well-known problem in ultrasound imaging: they are corrupted by speckle noise. Speckle noise can be smoothed by spatial averaging, but at the expense of spatial resolution. Alternatively, beamforming can be used to obtain beamformed acquisition data. In this case, it is possible to use orthogonal apodization at the aperture in the synthetic beamforming stage to reduce the variance of the average power estimate.

[0193] The method of the present disclosure, in particular the exemplary method described above, has the following advantages:

[0194] • An equation solver may be based on equations (13) and (14) (eg, because the quadratic terms are relatively small).

[0195] ●No spectral analysis is performed, so there is no need to trade off time / space vs. frequency resolution.

[0196] ●Broadband pulses can be used to minimize diffraction.

[0197] ● Synthetic beamforming can be performed to ensure transmission and reception are focused, reducing diffraction.

[0198] ● Orthogonal apodization can be used to smooth speckle to reduce the variance estimate of the mean power and thus reduce the variance of the ultrasound attenuation estimate.

[0199] ●Learning medium-independent parameters such as diffraction or probe sensitivity on a calibrated phantom.

[0200] Figure 6 A second exemplary embodiment of a method according to the present disclosure is shown (local attenuation estimation).

[0201] Figure 6 The method corresponds to Figure 4 method, namely, step 102 to step 104 and step 102' to step 104'. However, (instead of Figure 4 In step 104a and step 105) in the subsequent step 106, the two autocorrelation functions are compared by using a Kalman filter in order to estimate the local attenuation parameter.

[0202] Figure 6 The goal of the method is to obtain one or several local attenuation parameters, for example in the form of local estimation maps. This means that Figure 4 The basic assumptions are slightly different compared to the method of . In particular, it must now be taken into account that the ultrasound attenuation parameters vary with spatial position.

[0203] Here, it is proposed to estimate the local ultrasound attenuation along the axis of the depth direction z and to scan along the loss line, ie in a similar manner as is usually performed in B-mode imaging.

[0204] Therefore, the main goal of the process is to estimate the ultrasound attenuation parameter on a particular scan line as a function of depth, i.e., β(z). To this end, it is proposed to invert equation (11). Directly inverting this equation is not a straight forward process. Therefore, it is proposed to model the logarithm of the average power as a function of depth by observing (or measuring) the noisy state space system.

[0205] The logarithm of the average power as a function of depth can be discretized and modeled as a measure of a discrete-time controlled process governed by the following linear stochastic difference equation (15):

[0206]

[0207] in,

[0208] ●x k is the (hidden) state vector

[0209] A is the transfer matrix

[0210] ●w k and v k is the process and measurement noise

[0211] H is the measurement matrix

[0212] ●LE k is a vector containing the difference between the logarithm of the average power of the measured medium and the reference medium with a known attenuation depth kΔz at index k and the logarithm of the average power of the reference medium at depth kΔz

[0213] ●Δz is the depth sampling step size

[0214] ●B is the control matrix

[0215] ●u k Is the drive input

[0216] Let E[k] denote the logarithm of the average power estimate at depth kΔz. Taylor expansion of equation (11) yields:

[0217]

[0218] E[k]=E[k-1]-4f0β[k]Δz+constant(k) (16)

[0219] The equation can be approximated as:

[0220]

[0221] Let ΔE[k] be the difference in logarithm of the mean power at index k between the measured medium and a reference medium (e.g., a calibrated phantom) with known attenuation β ref .Then,

[0222] ΔE[k]=ΔE[k-1]-4β[k-1]f0Δz+4β ref [k-1]f0Δz

[0223] If (constant(k)-constant(k-1)) is contained in v k In, we get:

[0224] ΔE[k]=ΔE[k-1]-4β[k-1]f0Δz+4β ref [k-1]f0Δz+n1[k]

[0225]

[0226] in,

[0227] ●E ref is the logarithm of the average power estimated at index k on the reference medium

[0228] n1[k], n2[k], n3[k] are state noises: zero-mean Gaussian white noises.

[0229] if,

[0230]

[0231]

[0232] Then equation (17) can be transformed into the form of equation (15).

[0233] It has been found that the Kalman filter provides the state vector x in equation (15) in the MMSE sense k Therefore, it provides the best estimator of β[k] (in the MMSE sense).

[0234] Throughout the specification, including the claims, unless otherwise stated, the term "comprising a" should be understood as being synonymous with "comprising at least one". In addition, unless otherwise stated, any ranges set forth in the specification, including the claims, should be understood to include its end value(s). Specific values ​​of the described elements should be understood to be within acceptable manufacturing or industrial tolerances known to those skilled in the art, and any use of the terms "substantially" and / or "approximately" and / or "generally" should be understood to fall within such acceptable tolerances.

[0235] Although the present disclosure has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the present disclosure.

[0236] It is intended that the specification and examples be considered exemplary only, with the true scope of the disclosure being indicated by the following claims.

[0237] In summary, the method according to the present disclosure as described above can achieve more accurate attenuation estimation and means lower computational costs, which improves the real-time computation mode in particular. In addition, due to the increased accuracy, the variance is reduced, thereby enabling improved reproducibility.

Claims

1. A method for estimating an ultrasonic attenuation parameter of a region in a medium, wherein: The method is implemented by a processing unit (8) associated with at least one ultrasonic transducer (2), and comprises the following steps: (a) a transmitting step (101), wherein at least one pulse is transmitted in the medium by means of a transducer; (b) a receiving step (102), wherein data is collected by a transducer in response to the pulse; (c) a processing step (103), wherein the data is processed by the processing unit to provide backscatter acquisition data of the region; (d) a function determination step (104), wherein an autocorrelation function of the backscattered acquisition data is determined, the autocorrelation function being a function of depth in the spatiotemporal domain; (e) An attenuation estimation step (105), wherein the ultrasonic attenuation parameter is estimated based on the autocorrelation function obtained at a lag τ=0.

2. The method according to claim 1, wherein: The processing step (c) comprises a beamforming step (103) in which the data is processed by a beamforming process to provide beamformed acquisition data of the region.

3. The method according to claim 1 or 2, wherein: In the attenuation estimation step (e), the ultrasound attenuation parameter is estimated over a depth-direction interval (Δz).

4. The method according to claim 3, wherein: In the attenuation estimation step (e), the ultrasound attenuation parameter is determined by solving the autocorrelation function for the interval (Δz).

5. The method according to any one of claims 1 and 2, wherein: The autocorrelation function of the backscatter acquisition data is a function of depth and time.

6. The method according to any one of claims 1 and 2, wherein: In the function determination step (d), residual diffraction patterns in the backscatter acquisition data are compensated based on predetermined diffraction reference data.

7. The method according to claim 6, wherein: Residual diffraction patterns in the backscatter acquisition data are compensated by applying the method comprising steps (a) to (e) on a reference medium having predetermined reference ultrasound attenuation parameters and comparing the autocorrelation functions.

8. The method according to claim 7, wherein: Comparing the autocorrelation functions includes calculating a delta function ΔE(z0, z1) between the autocorrelation functions and solving the delta function ΔE(z0, z1)=0(0102) to determine the ultrasonic attenuation parameter of the medium, wherein z0 and z1 are two different depths in the medium and the reference medium, respectively.

9. The method according to claim 8, wherein: The predetermined diffraction reference data include predetermined reference ultrasound attenuation parameters at at least two different depths (z0, z1) in the reference medium.

10. The method according to any one of claims 1 and 2, wherein: The pulses are broadband pulses and / or multiple pulses.

11. A method for estimating a local ultrasonic attenuation parameter of a region in a medium, The method according to any one of claims 1 to 10, and Also included is: for each of a plurality of axes in the depth direction (z) within the medium, The local ultrasound attenuation β in the equation F(Ey(z), β)=0 is estimated by performing at least one of the following methods, where Ey is the autocorrelation function and z is the depth: - modeling the autocorrelation function as a noise state space model with a driving input; - regularizing the autocorrelation function by a regularization method, wherein the regularization method comprises a Tikhonov regularization method, a ridge regression method and / or a Bayesian regression method; - filtering the autocorrelation function using a Savitzky-Golay filter, a Monte Carlo method, a singular spectrum analysis, and - Compressed sensing and / or sparse sampling methods.

12. The method according to claim 11, wherein: In case the autocorrelation function is modeled as a noise state space model, the drive input comprises a predetermined reference ultrasound attenuation parameter of a reference medium.

13. The method according to any one of claims 11 to 12, wherein: The noise state space model determines the local ultrasound attenuation parameter using a Kalman filter or a predefined Bayesian method.

14. The method according to any one of claims 11 to 12, wherein: by iterating the EM algorithm on the noisy state-space model to compensate for the reflectivity pattern of the medium until a predefined convergence threshold is obtained, The EM algorithm uses a first Kalman filter and a second Kalman filter, wherein the first Kalman filter outputs a reflectivity parameter and uses an ultrasonic attenuation parameter as a driving input, and the second Kalman filter outputs an ultrasonic attenuation parameter and uses a reflectivity parameter as a driving input. Therein, two filters are connected to use the output of the other filter as a driving input respectively.

15. A method for imaging local ultrasound attenuation within a medium region, The method according to any one of claims 11 to 14, in, An attenuation image is constructed based on the local ultrasound attenuation parameters.

16. The method for imaging local ultrasound attenuation in a medium region according to claim 15, further comprising the steps of: The attenuation image is fed into a machine learning model, which is trained to segment the image into one or more regions of interest, each of which is associated with at least one predefined category.

17. A system for estimating an ultrasonic attenuation parameter of a region in a medium, comprising a processing unit (8) associated with at least one ultrasonic transducer (2), the system being configured to: (a) transmitting at least one pulse in the medium via a transducer; (b) acquiring data via a transducer in response to the pulse; (c) processing the data by the processing unit to provide backscatter acquisition data of the area; (d) determining an autocorrelation function of the scatter acquisition data, the autocorrelation function being a function of depth in the spatiotemporal domain; (e) estimating the ultrasonic attenuation parameter based on the autocorrelation function acquired at a lag τ=0.

Images