RF data-based lung injury rapid imaging and quantification method and device

Through the two-dimensional MAD and transverse gradient MAD algorithm based on RF data, and quantitative analysis is carried out in combination with radio frequency autocorrelation algorithm, the problem of lung injury area identification and quantitative evaluation in the prior art is solved, and the problem of lung injury area identification and quantitative evaluation is achieved Rapid and accurate identification and quantitative assessment of lung injury.

CN118967853APending Publication Date: 2024-11-15XI AN JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410992191.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing technology is difficult to quickly and effectively identify and quantify lung injury areas, traditional ultrasound imaging technology cannot intuitively display the location of lung injury, and deep learning algorithms are large in computing and not flexible enough.

Method used

Using RF data-based method, the B-line artifact areas in lung tissues were quickly identified and imaged through two-dimensional MAD algorithm and transverse gradient MAD algorithm, and quantitative analysis of lung injury depth was performed with radio frequency autocorrelation algorithm.

Benefits of technology

Achieve rapid identification and quantitative evaluation of lung injury areas, which can remove background tissue, retain B-line areas related to lung injury while taking into account respiratory displacement and hand shaking, and directly calculate the depth of lung tissue damage, improve the The accuracy and efficiency of identification are obtained.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967853A_ABST
    Figure CN118967853A_ABST
Patent Text Reader

Abstract

The invention discloses a lung injury rapid imaging and quantification method and device based on RF data, and belongs to the technical field of ultrasonic imaging and quantification. For ultrasonic radio frequency data, a two-dimensional MAD algorithm and a transverse gradient MAD algorithm are used to carry out rapid imaging on lung B-line artifacts, and a radio frequency echo autocorrelation algorithm is used to quantify the degree of B-line lung injury. The method comprises an imaging part and a quantification part, and aims at firstly carrying out rapid identification imaging on a B-line artifact to realize positioning of a lung injury area, and then carrying out lung injury depth quantification based on data of a B-line area to realize quantitative judgment of the lung injury degree. According to the method, richer information can be obtained, ultrasonic physical propagation characteristics can be better conformed, and lung tissue information can be represented more truly.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of ultrasonic imaging and quantification, and relates to a method and device for rapid imaging and quantification of lung damage based on RF data. Background Art

[0002] The generation of B-line artifacts is believed to be due to the reverberation artifact caused by partial absorption and re-emission of ultrasound energy due to the complex spatial structure of subpleural lesions in the lungs. The absorption, scattering and interference between the reflected waves received by the ultrasound transducer, the amount of air contained in the alveoli, and the spatial connectivity of the subpleural tissues will also jointly affect the generation of lung B-line artifacts. However, traditional ultrasound imaging technology can only characterize common artifacts in lung tissue, but cannot intuitively locate and display the lung injury area. It requires physicians with many years of clinical experience to determine the location of lung injury. At present, the research on the recognition of ultrasound artifacts includes some simple data processing methods such as Hilbert transform, filtering, absolute value, and the use of algorithms such as deep learning to identify lung B lines. However, the recognition ability of the former is affected by the parameter settings of the ultrasound imaging system, and the latter requires a large amount of training data and a large amount of calculation, which is not flexible enough. At present, there is still a lack of an effective and fast B-line area recognition method. Summary of the invention

[0003] The purpose of the present invention is to overcome the above-mentioned shortcomings of the prior art and provide a method and device for rapid imaging and quantification of lung injuries based on RF data to solve the problem of lack of effective and rapid B-line area identification method in the prior art.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] A rapid imaging and quantification method for lung injury based on RF data comprises the following steps:

[0006] S1, acquiring two adjacent frames of ultrasound radio frequency data both containing B-line images, and using Hilbert transform to obtain the envelope of each scanning line of the ultrasound radio frequency data to obtain two-dimensional envelope energy data;

[0007] S2, using the two-dimensional MAD algorithm to perform B-line recognition imaging on the two-dimensional envelope energy data to obtain a process image 1 for preliminary elimination of background tissue;

[0008] S3, performing secondary processing on the process image 1 obtained in S2 by using the transverse gradient MAD algorithm to obtain the process image 2 that further eliminates the background tissue;

[0009] S4, calculating the total energy of each scanning line in the second process image, selecting the scanning line whose total energy is greater than the set threshold, and obtaining the B line area;

[0010] S5, calculate each scanning line in the B-line area through the radio frequency autocorrelation algorithm, obtain the sampling point time difference between two adjacent strong reflections in the ultrasonic echo signal, and obtain the actual depth of each scanning line based on the sampling point time difference. The actual depth is the injury depth, completing the quantitative process of lung injury.

[0011] A further improvement of the present invention is:

[0012] Preferably, the process of obtaining two-dimensional envelope energy data in S1 is: obtaining two adjacent frames of ultrasonic radio frequency signals, and performing Hilbert transform on the ultrasonic radio frequency signal of each frame of ultrasonic radio frequency data; combining the Hilbert transform result and the ultrasonic radio frequency signal of each frame of ultrasonic radio frequency data to obtain an analytical signal, and based on the analytical signal, obtaining two-dimensional envelope energy data for each frame of ultrasonic radio frequency data.

[0013] Preferably, the formula for performing B-line recognition imaging on the two-dimensional envelope energy data by the two-dimensional MAD algorithm in S2 is:

[0014]

[0015] Where D(p,q) is the two-dimensional MAD value of any point in the image, H(p,q) is an M×N two-dimensional Hanning window, M and N are the vertical and horizontal window widths, m and n represent the row and column sizes of the two-dimensional image, 1≤p≤m-M+1, 1≤q≤n-N+1, For non-integer Take the value rounded down, E S (p, q) is the previous moment image of the data envelope of two adjacent frames in the two-dimensional MAD algorithm, E T (p, q) is the last moment image of the data envelope of two adjacent frames in the two-dimensional MAD algorithm, s, t are the sth and tth horizontal points in the window respectively, 1≤s≤M, 1≤t≤N, and the compensation gain is ε.

[0016] Preferably, in S3, the specific process of performing secondary processing on the process image obtained in S2 by the transverse gradient MAD algorithm is: setting a one-dimensional Hanning window, performing transverse subtraction on the two-dimensional MAD data corresponding to two adjacent columns of images, summing up all the differences, and obtaining the transverse gradient MAD value of the window center point.

[0017] Preferably, the calculation process of the transverse gradient MAD value is:

[0018]

[0019] In the formula, G(p,q) represents the MAD value of the horizontal gradient at any point in the image, and H(q) represents the one-dimensional Hanning window. M is the longitudinal window width, For non-integer The rounded-down value, S(p,q) represents the value of any point in the subtracted image, 1≤s≤M, and the compensation gain is ε.

[0020] Preferably, in S4, the total energy calculation formula for each scan line is:

[0021]

[0022] Wherein, E(q) represents the total energy intensity of a scan line in a certain B-line region, p is the row index, q is the column index, and m represents the data length of a scan line.

[0023] Preferably, the specific process of S5 is:

[0024] (1) obtaining a radio frequency signal of a certain scanning line in the B-line area, setting a step size, a window length, and a delay, and obtaining a sliding window autocorrelation value;

[0025] (2) Based on the sliding window autocorrelation value, a normalized sliding autocorrelation value is obtained;

[0026] (3) based on the set window length gradient sequence, step length gradient sequence and delayed gradient sequence, combined with the normalized sliding autocorrelation value obtained in step (2), obtain the normalized sliding autocorrelation maximum value;

[0027] (4) Based on the window length corresponding to the maximum value of the normalized sliding autocorrelation, the ultrasonic propagation velocity and the sampling frequency of the data, the actual depth of the scan line is obtained.

[0028] Preferably, after S5, the average damage depth of the continuous B-line area can be obtained, and the damage depth distribution map of the B-line area of ​​the entire image can be obtained; the calculation formula of the average damage depth is:

[0029]

[0030] Among them, w(q) represents the weight, q represents the scan line position, that is, the column index, q0 represents the scan line position at the center of a certain B-line area, and N represents the number of scan lines in the area. Indicates that If it is not divisible, round down.

[0031] A rapid imaging and quantification device for lung injury based on RF data, comprising:

[0032] A two-dimensional envelope energy acquisition module is used to acquire two adjacent frames of ultrasound radio frequency data both containing B-line images, and to obtain the envelope of each scanning line of the ultrasound radio frequency data by using Hilbert transform to obtain two-dimensional envelope energy data;

[0033] A preliminary elimination module, used for performing B-line recognition imaging on the two-dimensional envelope energy data through a two-dimensional MAD algorithm to obtain a process image 1 for preliminary elimination of background tissue;

[0034] A depth elimination module, used for performing secondary processing on the process image 1 obtained by S2 through a transverse gradient MAD algorithm to obtain a process image 2 that further eliminates background tissue;

[0035] The B-line region module is used to calculate the total energy of each scan line in the second process image, select the scan line whose total energy is greater than the set threshold, and obtain the B-line region;

[0036] The quantitative module is used to calculate each scanning line in the B-line area through the radio frequency autocorrelation algorithm, obtain the sampling point time difference between two adjacent strong reflections in the ultrasonic echo signal, and obtain the actual depth of each scanning line based on the sampling point time difference. The actual depth is the damage depth, thereby completing the quantitative process of lung damage.

[0037] Preferably, it also includes:

[0038] Weighted average module, used to calculate the average lesion depth of continuous B-line areas;

[0039] The quantitative module is used to draw the damage depth distribution map of the B-line area of ​​the entire image.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The present invention discloses a method for rapid imaging and quantification of lung injury based on RF data. The method first uses a two-dimensional mean of absolute differences (MAD) algorithm and a transverse gradient MAD algorithm to rapidly identify and image the B-line artifact area in the lung tissue, and then uses a radio frequency autocorrelation algorithm based on the identified B-line area data to quantify the depth of lung injury. The advantages of the present invention are:

[0042] (1) RF data contains the original information of the ultrasonic signal. Using RF data for imaging and quantitative analysis can obtain richer information, better fit the physical propagation characteristics of ultrasound, and provide a more realistic representation of lung tissue information.

[0043] (2) The absolute difference average algorithm is used for imaging. It can remove the regions of no interest (non-ROI regions) in the image, including background tissue, A line, pleural line, etc., while taking into account the respiratory displacement of the rabbit's lungs and the slight shaking of the ultrasound operator's hands, and retain the B line area (ROI region) that is highly correlated with lung damage.

[0044] (3) The transverse gradient MAD algorithm is used to perform secondary processing on the image data after being processed by the two-dimensional MAD algorithm, thereby enhancing the robustness of the algorithm.

[0045] (4) Since the MAD algorithm is simple and easy to use and is not limited by the parameter settings of the ultrasound imaging equipment, it can achieve fast and effective B-line area imaging.

[0046] (5) Compared with the traditional semi-quantitative method of counting and scoring the number of B-lines, the radio frequency autocorrelation algorithm can directly calculate the depth of lung tissue damage quantitatively. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of the general steps of the invention;

[0048] Figure 2 This is a diagram of the B-line artifact imaging steps based on the 2D MAD and transverse gradient MAD algorithms;

[0049] Figure 3 is an example of comparison of the results of each imaging step;

[0050] Among them, (a) is the reference image; (b) is the two-dimensional image processed by the two-dimensional MAD algorithm; (c) is the two-dimensional image processed by the transverse gradient MAD algorithm; (d) is the B-line area.

[0051] Figure 4 This is a schematic diagram of the damage depth quantification method based on the radio frequency autocorrelation algorithm;

[0052] Figure 5 It is a step diagram of the damage depth quantification method based on the radio frequency autocorrelation algorithm;

[0053] Figure 6 This is a schematic diagram of the depth distribution of lung damage. DETAILED DESCRIPTION

[0054] In the following, the terms "first", "second", "third", and "fourth" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, a feature defined as "first", "second", "third", and "fourth" may explicitly or implicitly include one or more of the features.

[0055] The co-shooting method provided in the embodiment of the present application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle-mounted devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra-mobile personal computers (UMPC), netbooks, personal digital assistants (PDA), etc. The embodiment of the present application does not impose any restrictions on the specific type of the terminal device.

[0056] It should be noted that the terms "first", "second", etc. in the specification and drawings of the present invention are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0057] The present invention proposes a method based on ultrasonic raw radio frequency (RF) data, which first uses a two-dimensional mean of absolute differences (MAD) algorithm and a transverse gradient MAD algorithm to quickly identify and image the B-line artifact area in the lung tissue, and then uses a radio frequency autocorrelation algorithm based on the identified B-line area data to quantify the depth of lung injury.

[0058] The overall steps of the method are as follows: (1) acquiring continuous ultrasound radio frequency data, selecting two adjacent frames of data, and using a two-dimensional MAD algorithm to perform B-line recognition imaging on the data; (2) using a transverse gradient MAD algorithm on the image processed in step (1) to further obtain a high-contrast B-line area recognition image; (3) based on the recognized B-line area data, the depth of lung injury is quantified using a radio frequency autocorrelation algorithm.

[0059] See also Figure 1 , Figure 2 and Figure 5 , the steps of the method are:

[0060] Step 1: Use Hilbert transform to obtain the envelope of each scan line for two adjacent frames of RF data to obtain two-dimensional envelope energy data;

[0061] First, the ultrasonic radio frequency signals of two adjacent frames are obtained. Both frames contain B lines. The data is set as a two-dimensional radio frequency signal x(p,q), and its Hilbert transform is expressed as H[x(p,q)]. The analytical signal A(p,q) can be expressed as:

[0062] A(p,q)=x(p,q)+jH[x(q)] (1)

[0063] Here, j is an imaginary unit.

[0064] The modulus of the analytical signal is the envelope E(p,q) of the original signal; for each column of the analytical signal A(t), the envelope E(t) can be obtained by calculating the modulus of the analytical signal:

[0065]

[0066] That is, by repeatedly calculating the analytical signal modulus of the one-dimensional data for each column of RF data in the image, the envelope energy matrix of the entire two-dimensional signal can be obtained.

[0067] Step 2: Use the two-dimensional MAD algorithm to perform B-line recognition imaging on two adjacent frames of two-dimensional envelope energy data of ultrasound, preliminarily reduce the signal energy of background tissue, pleural line, A-line artifacts, etc. in the lung tissue, and obtain process image 1;

[0068] The envelope energy data of two adjacent frames are calculated using the following two-dimensional MAD algorithm:

[0069]

[0070] Where D(p,q) is the two-dimensional MAD value of any point in the image, H(p,q) is an M×N two-dimensional Hanning window, M and N are the vertical and horizontal window widths, m and n represent the row and column sizes of the two-dimensional image, 1≤p≤m-M+1, 1≤q≤n-N+1, For non-integer Take the integer value downward. The data of the two adjacent frames selected in the two-dimensional MAD algorithm are recorded as the image envelope at the previous moment as E S (p,q), the image envelope at the next moment is E T (p, q), s, t are the sth and tth horizontal points in the window respectively, 1≤s≤M, 1≤t≤N. The compensation gain is ε.

[0071] Step 3: Use the transverse gradient MAD algorithm to perform secondary processing on the image data processed in step 2, further remove the A line and pleural line perpendicular to the ultrasound propagation direction, retain the B line area parallel to the ultrasound propagation direction, increase the contrast of the identified image, and obtain process image 2.

[0072] Due to the differences in patients' physiological conditions, the uncertainty of physician operations, and the differences in ultrasound equipment, the imaging effect of the image obtained by using the two-dimensional MAD algorithm for different radio frequency data sets may not be stable. In order to enhance the robustness and adaptability of the algorithm, the image data can be further processed on the basis of the two-dimensional MAD algorithm to enhance the contrast of the image. The core method adopted by the present invention is the transverse gradient MAD algorithm. Since the background tissues in non-interested areas such as the A line and the pleural line are mostly in a direction perpendicular to the ultrasound propagation direction, that is, the transverse direction in the image, and the B-line artifacts in the area of ​​interest are mostly in a direction parallel to the ultrasound propagation direction, that is, the longitudinal direction in the image. Therefore, similar to the two-dimensional MAD algorithm, this method sets a one-dimensional Hanning window, and the data after obtaining the two-dimensional MAD corresponding to the two adjacent columns of images are horizontally subtracted, and the difference is summed as the transverse gradient MAD value of the center point of the window, which further eliminates the background tissues in non-interested areas, while retaining the B-line artifacts in the area of ​​interest.

[0073] The image processed in step 2 is processed according to the following transverse gradient MAD algorithm:

[0074]

[0075] In the formula, G(p,q) represents the MAD value of the horizontal gradient at any point in the image, and H(q) represents the one-dimensional Hanning window. M is the longitudinal window width, For non-integer Taking the resin rounded down, S(p,q) represents the value of any point in the subtraction image, 1≤s≤M, and the compensation gain is ε.

[0076] Step 4: Set an appropriate threshold (the threshold is adjustable), calculate the total energy of each scan line in the image obtained in step 3, and select the scan lines with total energy values ​​greater than the threshold as B-line areas;

[0077] After the two-dimensional MAD algorithm and the horizontal gradient MAD algorithm, the total energy of the B-line area is obviously greater than that of the non-B-line area. At this time, the B-line area in the image is marked for subsequent use. The core of the method is to calculate the total energy value of each scan line in the B-line area. First, the total energy intensity of each scan line of the image after steps 1-3 is calculated:

[0078]

[0079] Where E(q) represents the total energy intensity of a scan line in a B-line region, p is the row index, q is the column index, and m represents the data length of a scan line, where a scan line refers to each column of data along the ultrasonic propagation direction. Calculate the total energy value E(q) of the scan lines in all B-line regions.

[0080] Then, set the threshold k (the k value is user-adjustable). If the total energy value of the scan line is less than k, the scan line is marked as a non-B-line area, otherwise it is marked as B-line data. Users can change the threshold to find the threshold that maximizes the retention of B-line areas and removes non-B-line areas, mark the image, and overlay the marking result on the original horizontal gradient image. Users can continuously change the threshold k to keep the B-line area as much as possible and the non-B-line area as much as possible.

[0081] Step 5: For the scan lines in the B-line area selected in step 4, the depth of the subpleural lung injury is calculated based on the original radio frequency data using the radio frequency autocorrelation algorithm to achieve quantification of the lung injury in the B-line area. The present invention uses the radio frequency autocorrelation algorithm to vertically quantify the depth of the injury, that is, to obtain the sampling point time difference W between two adjacent strong reflections on the ultrasonic echo signal. Each scan line in the B-line area is subjected to the radio frequency autocorrelation algorithm described in the following steps to obtain the radio frequency periodicity formed by multiple reflections of the B-line, and then obtain the actual depth and position.

[0082] After the B-line area is marked, the scan lines in all the B-line areas in the image are processed as follows:

[0083] (1) For a certain RF signal X of a scanning line in the B-line area, the step size S, window length W, and delay τ are set to obtain the calculation formula of the sliding window autocorrelation value R(X, S, W, τ):

[0084]

[0085] Wherein, N represents the length of the RF signal X; The maximum number of sliding window slides after rounding down.

[0086] (2) Based on the sliding window autocorrelation value, the normalized sliding autocorrelation value R is obtained Normal (X,S,W,τ):

[0087]

[0088] (3) Set the window length gradient sequence {W1, W2, …, W n}, the step gradient sequence is {S1,S2,…,S n}, the delayed gradient sequence is {τ1,τ2,…,τ n}, substitute into formula (10) to obtain the normalized sliding autocorrelation R Normal (X,S,W,τ) takes its maximum value R MaxNor The window length W of (X,S,W,τ) m :

[0089] R MaxNor (X,S m ,Wm ,τ m )=max[R Normal (X,S p ,W q ,τ λ )] (10)

[0090] 1≤p≤k,1≤q≤n,1≤λ≤q

[0091] Among them, S m ,W m ,τ m Indicates the step size, window length and delay value corresponding to the maximum normalized sliding autocorrelation value.

[0092] (4) Obtain the obtained W m It reflects the most reliable periodicity in the RF data, and then finds the corresponding depth d:

[0093]

[0094] Among them, c represents the ultrasonic propagation velocity, which is generally 1540m / s in human tissue, and Fs represents the sampling frequency of the data.

[0095] Step 6: Perform a linear weighted average calculation on the estimated lung injury depth of each scan line in the B-line area retained after using the two-dimensional transverse gradient MAD or MSD algorithm to obtain the average injury depth of the area, and then draw a lesion depth distribution map for different B-line areas.

[0096] After obtaining the damage depth d(j) of all B-line scan lines of the image in step 5, first take the linear weighted average of the depth d in the B-line area as the average damage depth of this B-line area, and then draw a damage depth distribution map of the B-line area of ​​the entire image.

[0097] When performing linear weighting, the maximum value of the center of the window needs to be aligned with the center of the B-line damage area, the center is assigned the largest weight, and the damage depths on both sides of the B-line area are assigned smaller weights. Linear weighting is implemented using a triangular window, and the weighted average formula is as follows:

[0098]

[0099]

[0100] Among them, w(q) represents the weight, q represents the scan line position, that is, the column index, q0 represents the scan line position at the center of a certain B-line area. N represents the number of scan lines in the area, Indicates that If it is not divisible, round down.

[0101] The second aspect of the present invention further discloses a rapid imaging and quantification device for lung injury based on RF data, comprising:

[0102] A two-dimensional envelope energy acquisition module is used to acquire two adjacent frames of ultrasound radio frequency data both containing B-line images, and to obtain the envelope of each scanning line of the ultrasound radio frequency data by using Hilbert transform to obtain two-dimensional envelope energy data;

[0103] A preliminary elimination module, used for performing B-line recognition imaging on the two-dimensional envelope energy data through a two-dimensional MAD algorithm to obtain a process image 1 for preliminary elimination of background tissue;

[0104] A depth elimination module, used for performing secondary processing on the process image 1 obtained by S2 through a transverse gradient MAD algorithm to obtain a process image 2 that further eliminates background tissue;

[0105] The B-line region module is used to calculate the total energy of each scan line in the second process image, select the scan line whose total energy is greater than the set threshold, and obtain the B-line region;

[0106] The quantitative module is used to calculate each scanning line in the B-line area through the radio frequency autocorrelation algorithm, obtain the sampling point time difference between two adjacent strong reflections in the ultrasonic echo signal, and obtain the actual depth of each scanning line based on the sampling point time difference. The actual depth is the damage depth, thereby completing the quantitative process of lung damage.

[0107] The following is further described in conjunction with specific embodiments.

[0108] Example 1

[0109] In one embodiment of the present invention, a method for rapid imaging and quantification of lung damage based on RF data is disclosed. The method images B-line artifacts of lung tissue. The specific contents are as follows:

[0110] (1) Use B-mode to image tissue as a reference image, such as Figure 3 As shown in Figure (a) in the figure, and the RF data is exported.

[0111] (2) Select two adjacent frames of RF data and calculate the column-wise two-dimensional envelope data according to formula (2).

[0112] (3) According to formula (3), the two-dimensional MAD algorithm is used to subtract the two frames of data. The vertical and horizontal window widths M and N can be appropriately adjusted to obtain a higher contrast and clarity of the image. Figure 3 As can be seen in Figure (b), the processed B-line image may still have non-ROIs such as A-line, pleural line, and tissue background. Although the number is small, these residual artifacts often have a large intensity, which can easily cause the B-line artifact to become darker or distorted after normalization.

[0113] (4) Calculate the horizontal gradient MAD value of the entire image according to formula (5) after the image is processed by the two-dimensional MAD algorithm and form an image. The vertical window width M can be appropriately adjusted to obtain a higher contrast and clarity of the image. Figure 3 As can be seen from Figure (c), after two adjacent longitudinal windows are subtracted point by point and then summed up as the gradient value of the window center, and the image is normalized, the residual tissue background, pleural line, and A-line artifacts in the image disappear or their grayscale values ​​are relatively reduced, while the grayscale value of the B-line area is retained to a greater extent.

[0114] (4) According to formula (7), the total energy intensity of each scan line on the image is calculated, and an appropriate threshold is set. The scan lines with a total energy greater than the threshold in the whole image are marked as B-line areas, and those with a total energy greater than the threshold are not marked. Finally, the horizontal distribution range of lung damage is determined. This threshold can be adjusted by setting a slider. The user can continuously adjust the threshold and select an energy threshold that can retain the B-line area as much as possible and eliminate the non-B-line area. The result is as follows: Figure 3 As shown in Figure (d).

[0115] (5) A water-air interface is formed at the junction of the pleural line and the lung injury area on the alveolar structure. The ultrasound will produce a strong echo here due to impedance mismatch, followed by another strong echo at the bottom of the injury area. After completing the imaging and marking of the B-line area according to the above steps (1)-(4), follow Figure 5 The steps shown and referring to formulas (8) to (11) calculate the depths of all scan lines in the B-line area.

[0116] (6) Calculate the damage depth d of all scan lines in each independent and continuous B-line area in turn, perform linear weighted average according to formula (12), and use it as the average depth of the damaged area here. If there are multiple B-line areas in the image, the following can be generated: Figure 6 The damage depth distribution diagram is shown, where the horizontal axis represents the number of scan lines and the vertical axis represents the damage depth.

[0117] The present invention uses the two-dimensional MAD algorithm and the transverse gradient MAD algorithm for rapid imaging of the B-line artifacts in the lungs based on the ultrasound radio frequency data, and uses the radio frequency echo autocorrelation algorithm to quantify the degree of lung damage in the B-line. The invention includes two parts, imaging and quantification, and aims to firstly rapidly identify and image the B-line artifacts to locate the lung damage area, and then quantify the depth of lung damage based on the data of the B-line area to achieve quantitative judgment of the degree of lung damage.

[0118] When imaging the artifact area of ​​pulmonary tissue, the present invention considers using a two-dimensional MAD algorithm and a transverse gradient MAD algorithm to achieve rapid B-line area recognition, which is simple and effective. The two-dimensional MAD algorithm subtracts two adjacent frames of data by adding windows, thereby eliminating signals such as the pleural line, A-line artifacts, and tissue background in the lung tissue. In the respiratory action of the lung tissue, artifacts such as the tissue background and the pleural line change less with time, and the B-line artifact is caused by displacement caused by the sliding of the lung breathing, and is accompanied by complex scattering, which changes greatly with time. Therefore, the A-line, pleural line, tissue background and other areas in the two frames of data with a small time gap do not change much, while the damaged area that produces the B-line has a large change. The damaged area that produces the B-line can be identified and imaged using two frames of data that are relatively adjacent in time through a two-dimensional MAD algorithm. Then, the method uses a transverse gradient MAD algorithm for processing, aiming to further remove the A-line and pleural line perpendicular to the direction of ultrasound propagation, retain the B-line area parallel to the direction of ultrasound propagation, and increase the contrast of the recognized image.

[0119] After visually imaging the lung injury, it is necessary to quantify the severity of the lung injury. In addition to using traditional image signal processing methods, a scoring system is often used in clinical practice to semi-quantitatively study the ultrasound B-line artifacts of lung tissue. This mechanism is to calculate the B-line score in different areas of the examination area according to several prescribed scoring methods, and then judge the severity of the disease according to the total score of the entire area. The current scoring standards include B-line fusion score, modified B-line fusion score, etc. However, the lung ultrasound scoring system has certain limitations. When the ultrasound probe scans the pleura between the ribs longitudinally, the scanning results are affected by the width of the intercostal space. Due to different congenital and acquired conditions, the spacing between ribs varies greatly in different patients, resulting in different scoring results for lung injury based on the number of artifacts, which greatly reduces the reliability of the scoring. In addition, at present, B-line detection and quantification are mostly based on image data for positioning and analysis. Since image and video data will lose some original information during signal processing, they cannot reflect the information of real lung tissue. In contrast, radio frequency data can have a more realistic representation of lung tissue signals.

[0120] Therefore, the present invention proposes to use a radio frequency autocorrelation algorithm to perform a quantitative analysis of real lung damage. As the severity of lung damage increases, the damaged area will be deeper and the area will be larger. Therefore, after determining the horizontal width of the B-line area, this quantification method can estimate the damage depth of each scan line based on the radio frequency data of the B-line area, so as to approximately restore the position and size of the lung damage and achieve a quantitative assessment of the degree of lung damage. Due to the impedance mismatch between the contact surface of the pleural line and the lung damage area, and the boundary between the lung damage area and the healthy tissue area, the ultrasonic wave will be reflected multiple times at these interfaces, so that the lung ultrasound data presents a modulated waveform with an ultrasonic frequency as the carrier and an envelope peak with a significant correlation. The periodicity of the longitudinal propagation of the ultrasonic wave is calculated, and the depth of the damage is then obtained.

[0121] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A rapid imaging and quantification method for lung injury based on RF data, characterized in that: The following steps are involved: S1, acquiring two adjacent frames of ultrasound radio frequency data both containing B-line images, and using Hilbert transform to obtain the envelope of each scanning line of the ultrasound radio frequency data to obtain two-dimensional envelope energy data; S2, using the two-dimensional MAD algorithm to perform B-line recognition imaging on the two-dimensional envelope energy data to obtain a process image 1 for preliminary elimination of background tissue; S3, performing secondary processing on the process image 1 obtained in S2 by using the transverse gradient MAD algorithm to obtain the process image 2 that further eliminates the background tissue; S4, calculating the total energy of each scanning line in the second process image, selecting the scanning line whose total energy is greater than the set threshold, and obtaining the B line area; S5, calculate each scanning line in the B-line area through the radio frequency autocorrelation algorithm, obtain the sampling point time difference between two adjacent strong reflections in the ultrasonic echo signal, and obtain the actual depth of each scanning line based on the sampling point time difference. The actual depth is the injury depth, completing the quantitative process of lung injury.

2. The method for rapid imaging and quantification of lung injury based on RF data according to claim 1, characterized in that: The process of obtaining the two-dimensional envelope energy data in S1 is: obtaining two adjacent frames of ultrasonic radio frequency signals, performing Hilbert transform on the ultrasonic radio frequency signal of each frame of ultrasonic radio frequency data; combining the Hilbert transform result and the ultrasonic radio frequency signal of each frame of ultrasonic radio frequency data to obtain an analytical signal, and based on the analytical signal, obtaining the two-dimensional envelope energy data of each frame of ultrasonic radio frequency data.

3. The method for rapid imaging and quantification of lung injury based on RF data according to claim 1, characterized in that: The formula for B-line recognition imaging of two-dimensional envelope energy data using the two-dimensional MAD algorithm in S2 is: Where D(p,q) is the two-dimensional MAD value of any point in the image, H(p,q) is an M×N two-dimensional Hanning window, M and N are the vertical and horizontal window widths, m and n represent the row and column sizes of the two-dimensional image, 1≤p≤m-M+1, 1≤q≤n-N+1, For non-integer Take the value rounded down, E S (p, q) is the previous moment image of the data envelope of two adjacent frames in the two-dimensional MAD algorithm, E T (p, q) is the last moment image of the data envelope of two adjacent frames in the two-dimensional MAD algorithm, s, t are the sth and tth horizontal points in the window respectively, 1≤s≤M, 1≤t≤N, and the compensation gain is ε.

4. The method for rapid imaging and quantification of lung injury based on RF data according to claim 1, characterized in that: In S3, the specific process of performing secondary processing on the process image 1 obtained in S2 by the transverse gradient MAD algorithm is as follows: setting a one-dimensional Hanning window, performing transverse subtraction on the two-dimensional MAD data corresponding to two adjacent columns of images, summing up all the differences, and obtaining the transverse gradient MAD value of the window center point.

5. The method for rapid imaging and quantification of lung injury based on RF data according to claim 4, characterized in that: The calculation process of the transverse gradient MAD value is: In the formula, G(p,q) represents the MAD value of the horizontal gradient at any point in the image, and H(q) represents the one-dimensional Hanning window. 1≤q≤n-1, M is the vertical window width, For non-integer The rounded-down value, S(p,q) represents the value of any point in the subtracted image, 1≤s≤M, and the compensation gain is ε.

6. The method for rapid imaging and quantification of lung injury based on RF data according to claim 1, characterized in that: In S4, the total energy calculation formula for each scan line is: Wherein, E(q) represents the total energy intensity of a scan line in a certain B-line region, p is the row index, q is the column index, and m represents the data length of a scan line.

7. The method for rapid imaging and quantification of lung injury based on RF data according to claim 1, characterized in that: The specific process of S5 is as follows: (1) obtaining a radio frequency signal of a certain scanning line in the B-line area, setting a step size, a window length, and a delay, and obtaining a sliding window autocorrelation value; (2) Based on the sliding window autocorrelation value, a normalized sliding autocorrelation value is obtained; (3) based on the set window length gradient sequence, step length gradient sequence and delayed gradient sequence, combined with the normalized sliding autocorrelation value obtained in step (2), obtain the normalized sliding autocorrelation maximum value; (4) Based on the window length corresponding to the maximum value of the normalized sliding autocorrelation, the ultrasonic propagation velocity and the sampling frequency of the data, the actual depth of the scan line is obtained.

8. The method for rapid imaging and quantification of lung injury based on RF data according to claim 1, characterized in that: After S5, the average damage depth of the continuous B-line area can be obtained, and the damage depth distribution map of the B-line area of ​​the entire image can be obtained; the calculation formula of the average damage depth is: Among them, w(q) represents the weight, q represents the scan line position, that is, the column index, q0 represents the scan line position at the center of a certain B-line area, and N represents the number of scan lines in the area. Indicates that If it is not divisible, round down.

9. A rapid imaging and quantification device for lung injury based on RF data, characterized in that: include: A two-dimensional envelope energy acquisition module is used to acquire two adjacent frames of ultrasound radio frequency data both containing B-line images, and to obtain the envelope of each scanning line of the ultrasound radio frequency data by using Hilbert transform to obtain two-dimensional envelope energy data; A preliminary elimination module, used for performing B-line recognition imaging on the two-dimensional envelope energy data through a two-dimensional MAD algorithm to obtain a process image 1 for preliminary elimination of background tissue; A depth elimination module, used for performing secondary processing on the process image 1 obtained by S2 through a transverse gradient MAD algorithm to obtain a process image 2 that further eliminates background tissue; The B-line region module is used to calculate the total energy of each scan line in the second process image, select the scan line whose total energy is greater than the set threshold, and obtain the B-line region; The quantitative module is used to calculate each scanning line in the B-line area through the radio frequency autocorrelation algorithm, obtain the sampling point time difference between two adjacent strong reflections in the ultrasonic echo signal, and obtain the actual depth of each scanning line based on the sampling point time difference. The actual depth is the damage depth, thereby completing the quantitative process of lung damage.

10. The device for rapid imaging and quantification of lung injury based on RF data according to claim 9, characterized in that: Also includes: Weighted average module, used to calculate the average lesion depth of continuous B-line areas; The quantitative module is used to draw the damage depth distribution map of the B-line area of ​​the entire image.