A method and system for optimizing a liquid medium of an ultrasonic probe based on big data

Abstract

The present invention discloses a method and system for optimizing a liquid medium of an ultrasonic probe based on big data. The method includes constructing a three-dimensional first matrix, which is used to represent the state of the bubble radius corresponding to different probe operating frequencies at adjacent probe operating time points, performing convolution on the three-dimensional matrix using a three-dimensional first convolutional kernel, determining the change situation at adjacent time points to obtain liquid medium bubble characteristics, and using an associated neural network. Not only can the data of the probe operating frequency, the probe operating time point, and the liquid medium category be directly input in subsequent discrimination to obtain the liquid medium bubble characteristics of the corresponding bubble radius change, reducing the detection time, but also the parameters of the associated neural network can be updated in real time. Thus, in subsequent clustering, the liquid medium category that enables the bubble radius to change stably can be found, and the corresponding change cavitation threshold can be found.

Description

Technical Field

[0001] The present invention relates to the field of ultrasound, and in particular, to a method and system for optimizing a liquid medium of an ultrasound probe based on big data. Background Art

[0002] Currently, an ultrasound detector using a variable-frequency ultrasound instrument performs ultrasound detection. Low frequencies are used for biofilm lysis and promoting drug penetration, and high frequencies are used for sterilization, etc. During the operation of the handheld ultrasound probe, cavitation reactions occur when it is inserted into a liquid medium. The cavitation threshold is related to the radius of the bubbles in the medium. The smaller the bubble radius, the higher the cavitation threshold. The bubble size can be optimized and the generation of the cavitation effect can be promoted by adjusting the composition of the liquid medium or adding appropriate microbubbles. And the cavitation threshold is related to the operating frequency. The higher the frequency, the higher the cavitation threshold, and it is more difficult to generate cavitation. Therefore, when improving the liquid medium, the operating frequency of the main unit needs to be considered, and appropriate medium components need to be selected to match this frequency, so as to more easily generate the cavitation effect. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for optimizing a liquid medium of an ultrasound probe based on big data to solve the above problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides a method for optimizing a liquid medium of an ultrasound probe based on big data, including:

[0005] Obtaining a plurality of liquid medium categories, a plurality of probe operating frequencies, a plurality of probe operating time points, and corresponding bubble radii; the liquid medium category represents the category of the liquid medium where the handheld ultrasound probe is located; the probe operating frequency represents the frequency of the handheld ultrasound probe; the probe operating time point represents the duration of the handheld ultrasound probe in the liquid medium; the bubble radius represents the radius of the bubbles caused by the handheld ultrasound probe;

[0006] Based on the plurality of liquid medium categories, the plurality of probe operating frequencies, the plurality of probe operating time points, and the corresponding medium bubble radii, obtaining adjacent time data change characteristics and a plurality of liquid medium bubble characteristics; the adjacent time data change characteristics represent the change situations of the liquid medium category and the probe operating frequency in adjacent probe operating time points; the liquid medium bubble characteristics represent the change situations of the bubble radii corresponding to different probe operating frequencies in adjacent probe operating time points;

[0007] Constructing an association relationship between the adjacent time data change characteristics and the plurality of liquid medium bubble characteristics through an association neural network;

[0008] According to the plurality of liquid medium bubble characteristics, performing clustering to obtain detected liquid medium categories; the detected liquid medium categories are clustering centers;

[0009] Take the bubble radius greater than the other bubble radii in the detected liquid medium category as the first bubble radius;

[0010] Take the probe operating frequency corresponding to the first bubble radius as the optimal probe operating frequency, and the corresponding probe operating time point as the optimal probe operating time point;

[0011] Obtain a variable cavitation threshold according to the first bubble radius.

[0012] Optionally, the obtaining of the liquid medium bubble characteristics through the associated neural network based on the liquid medium category, the probe operating frequency, the probe operating time point, and the corresponding bubble radius includes:

[0013] Based on multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and the corresponding bubble radii, obtain multiple first matrices; the first matrix is a three-dimensional matrix; one liquid medium category corresponds to one first matrix;

[0014] Based on multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating durations, and the first matrix, obtain a second matrix; the second matrix is a two-dimensional matrix;

[0015] Based on the first matrix, obtain the liquid medium bubble characteristics; multiple first matrices correspond to obtaining multiple liquid medium bubble characteristics;

[0016] Based on the second matrix, obtain the adjacent time data change characteristics.

[0017] Optionally, the construction of the association relationship between the adjacent time data change characteristics and the liquid medium bubble characteristics through the associated neural network includes:

[0018] Sort multiple liquid medium bubble characteristics in the order of the liquid medium categories in the adjacent time data change characteristics to obtain the liquid medium fusion characteristics;

[0019] Construct a second associated neural network; the structure of the second associated neural network is the same as that of the associated neural network;

[0020] Based on the adjacent time data change characteristics and the liquid medium fusion characteristics, use the second associated neural network to adjust the parameters in the associated neural network.

[0021] Optionally, the using the second associated neural network to adjust the parameters in the associated neural network based on the adjacent time data change characteristics and the liquid medium fusion characteristics includes:

[0022] Input the adjacent time data change characteristics into the associated neural network, calculate the loss with the liquid medium fusion characteristics, and train the associated neural network;

[0023] Jointly adjust the positions of the same probe working time points corresponding to the adjacent time data change characteristics and the liquid medium fusion characteristics, and train the correlation neural network;

[0024] Input the adjacent time data change characteristics into the second correlation neural network, calculate the loss with the liquid medium fusion characteristics, and train the second correlation neural network;

[0025] Fuse the parameters at the corresponding positions of the second correlation neural network and the correlation neural network, and adjust the correlation neural network.

[0026] Optionally, the jointly adjusting the positions of the same probe working time points corresponding to the adjacent time data change characteristics and the liquid medium fusion characteristics, and training the correlation neural network includes:

[0027] Obtain a plurality of position pairs; each position pair includes a first position and a second position; the first position is a random position in the adjacent time data change characteristics; the second position is a position other than the second position in the adjacent time data change characteristics;

[0028] Successively exchange the values of multiple position pairs in the adjacent time data change characteristics to obtain transformed adjacent time data change characteristics;

[0029] Successively exchange the values of multiple position pairs in the liquid medium fusion characteristics to obtain transformed liquid medium fusion characteristics;

[0030] Input the transformed adjacent time data change characteristics into the correlation neural network, calculate the loss with the transformed liquid medium fusion characteristics, and train the first neural network.

[0031] Optionally, the obtaining the liquid medium bubble characteristics based on the first matrix includes:

[0032] Obtain a first convolution kernel; the first convolution kernel is a three-dimensional convolution kernel of 2*n*m; n is the number of columns of the first matrix; m is the number of pages of the first matrix;

[0033] On the rows of the first matrix, with a step size of 1, perform convolution of the first convolution kernel on the first matrix to obtain the liquid medium.

[0034] Optionally, the performing convolution based on the second matrix to obtain the adjacent time data change characteristics includes:

[0035] Obtain a second convolution kernel of 2*2;

[0036] On the rows of the second matrix, with a step size of 1, perform convolution of the second convolution kernel on the first matrix to obtain the adjacent time data change characteristics.

[0037] Optionally, obtaining multiple first matrices based on multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and corresponding bubble radii includes:

[0038] Sequentially associate the probe operating time points from smallest to largest with the rows in the first matrix with subscripts from smallest to largest, the probe operating frequencies from smallest to largest with the pages in the first matrix with subscripts from smallest to largest, and the bubble radii from smallest to largest with the columns in the first matrix with subscripts from smallest to largest;

[0039] Multiple liquid medium categories respectively obtain multiple first matrices;

[0040] In the first matrix corresponding to the liquid medium category, mark the positions of the probe operating frequency, probe operating time point, and corresponding bubble radius as 1.

[0041] Optionally, obtaining multiple second matrices based on multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating durations, and the first matrix includes:

[0042] Sequentially associate the probe operating time points from smallest to largest with the rows in the second matrix with subscripts from smallest to largest, the liquid medium category with the first column of the second matrix, and the probe operating frequency with the second column of the second matrix;

[0043] Multiple liquid medium categories respectively obtain multiple second matrices;

[0044] Use the liquid medium category corresponding to the first matrix as the first liquid medium category;

[0045] Fill the first liquid medium category into the first column of the second matrix;

[0046] Fill the probe operating frequency in the first matrix into the second column of the same row in the second matrix.

[0047] In a second aspect, an embodiment of the present invention provides a system for optimizing a liquid medium by an ultrasonic probe based on big data, including:

[0048] An acquisition module, configured to acquire multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and corresponding bubble radii; the liquid medium category represents the category of the liquid medium where the handheld ultrasonic probe is located; the probe operating frequency represents the frequency of the handheld ultrasonic probe; the probe operating time point represents the duration of the handheld ultrasonic probe in the liquid medium; the bubble radius represents the radius of the bubble caused by the handheld ultrasonic probe;

[0049] A feature extraction module, configured to obtain adjacent time data change features and multiple liquid medium bubble features based on the multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and the corresponding medium bubble radii; the adjacent time data change features represent the change conditions of the liquid medium category and the probe operating frequency in adjacent probe operating time points; the liquid medium bubble features represent the change conditions of the bubble radii corresponding to different probe operating frequencies in adjacent probe operating time points;

[0050] A feature correlation module, configured to construct a correlation relationship between the adjacent time data change features and the multiple liquid medium bubble features through a correlation neural network;

[0051] A clustering module, configured to perform clustering based on the multiple liquid medium bubble features to obtain the detected liquid medium categories; the detected liquid medium categories are the clustering centers;

[0052] A detection module, configured to use the bubble radius greater than other bubble radii in the detected liquid medium categories as the first bubble radius; use the probe operating frequency corresponding to the first bubble radius as the optimal probe operating frequency, and the corresponding probe operating time point as the optimal probe operating time point;

[0053] A cavitation detection module, configured to obtain a variable cavitation threshold according to the first bubble radius.

[0054] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0055] The embodiments of the present invention further provide a method and system for optimizing a liquid medium by an ultrasonic probe based on big data.

[0056] In the present invention, a three-dimensional first matrix is constructed according to multiple probe operating frequencies, multiple probe operating time points, and corresponding medium bubble radii; multiple first matrices are obtained corresponding to multiple liquid medium categories. The first matrix is used to represent the state of the bubble radius corresponding to different probe operating frequencies in adjacent probe operating time points. A three-dimensional first convolutional kernel is used to perform convolution on the three-dimensional matrix to determine the change situation between adjacent time points, and the liquid medium bubble characteristics are obtained. By constructing a correlation neural network, a correlation relationship is constructed between the liquid medium bubble characteristics and the changes in data such as the probe operating frequency, the probe operating time point, and the liquid medium category. By using the correlation neural network, not only can the liquid medium bubble characteristics with corresponding bubble radius changes be obtained directly by inputting the data of the probe operating frequency, the probe operating time point, and the liquid medium category during subsequent discrimination, reducing the detection time, but also the parameters of the correlation neural network can be updated in real time, making the detection more accurate. Additionally, adjusted data and a second correlation neural network are used as a network for adjusting the correlation neural network to more accurately construct the correlation neural network. Feature matching is performed using multiple liquid medium bubble characteristics and adjacent time data change characteristics. The change situation between two adjacent time points can be used for matching, making the matching more accurate and more in line with the change requirements. Thus, during subsequent clustering, the liquid medium category that causes the bubble radius to change stably can be found. The technical effect of more accurately finding the change cavitation threshold is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 FIG. is a flowchart of a method for optimizing a liquid medium by an ultrasonic probe based on big data provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] The present invention will be described in detail below with reference to the accompanying drawings.

[0059] Embodiment 1

[0060] As Figure 1 shown, an embodiment of the present invention provides a method for optimizing a liquid medium by an ultrasonic probe based on big data. The method includes:

[0061] S101: Obtain multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and corresponding bubble radii; the liquid medium category represents the category of the liquid medium where the handheld ultrasonic probe is located; the probe operating frequency represents the frequency of the handheld ultrasonic probe; the probe operating time point represents the duration of the handheld ultrasonic probe in the liquid medium; the bubble radius represents the radius of the bubble caused by the handheld ultrasonic probe.

[0062] S102: Based on the multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and the corresponding medium bubble radii, obtain the adjacent time data change characteristics and multiple liquid medium bubble characteristics; the adjacent time data change characteristics represent the change conditions of the liquid medium category and the probe operating frequency in adjacent probe operating time points; the liquid medium bubble characteristics represent the change conditions of the bubble radii corresponding to different probe operating frequencies in adjacent probe operating time points.

[0063] S103: Through an association neural network, construct the association relationship between the adjacent time data change characteristics and the multiple liquid medium bubble characteristics.

[0064] Among them, the association neural network can be updated in real time. The input values of the association neural network are multiple probe operating frequencies, multiple probe operating time points, and multiple liquid medium categories, and the output value of the association neural network is the characteristic representing the change of the adjacent time points of the bubble radius corresponding to the data of the input values.

[0065] S104: According to the multiple liquid medium bubble characteristics, perform clustering to obtain the detected liquid medium categories; the detected liquid medium categories are the clustering centers.

[0066] Among them, in this embodiment, the k-means method is used to calculate the Euclidean distance for each value corresponding to the position in the liquid medium bubble characteristics for clustering.

[0067] S105: Take the bubble radius greater than other bubble radii in the detected liquid medium categories as the first bubble radius.

[0068] Among them, the first bubble radius represents the largest bubble radius among the multiple bubble radii corresponding to the detected liquid medium categories.

[0069] S106: Take the probe operating frequency corresponding to the first bubble radius as the optimal probe operating frequency, and the corresponding probe operating time point as the optimal probe operating time point.

[0070] Among them, output the optimal probe operating frequency, the optimal probe operating time point, and the detected liquid medium categories.

[0071] S107: Obtain the variable cavitation threshold according to the first bubble radius.

[0072] Among them, in this embodiment, the variable cavitation threshold is obtained through the ultrasonic cavitation frequency threshold calculation formula as follows: f = (2σ / ρ)^(1 / 2) / (πR). f represents the variable cavitation threshold, σ represents the surface tension of the liquid, ρ represents the density of the liquid, and R represents the first bubble radius.

[0073] Among them, the variable cavitation threshold represents the value at which a cavitation reaction can occur.

[0074] Optionally, obtaining the adjacent time data change characteristics and the liquid medium bubble characteristics based on the multiple liquid medium categories, the multiple probe operating frequencies, the multiple probe operating time points, and the corresponding bubble radii includes:

[0075] Based on the multiple liquid medium categories, the multiple probe operating frequencies, the multiple probe operating time points, and the corresponding bubble radii, obtain multiple first matrices; the first matrix is a three-dimensional matrix; one liquid medium category corresponds to one first matrix;

[0076] Based on the multiple liquid medium categories, the multiple probe operating frequencies, the multiple probe operating durations, and the first matrix, obtain a second matrix; the second matrix is a two-dimensional matrix;

[0077] Based on the first matrix, obtain the liquid medium bubble characteristics; multiple first matrices correspond to obtaining multiple liquid medium bubble characteristics;

[0078] Based on the second matrix, obtain the adjacent time data change characteristics.

[0079] Optionally, constructing the correlation relationship between the adjacent time data change characteristics and the liquid medium bubble characteristics through the correlation neural network includes:

[0080] Sort the multiple liquid medium bubble characteristics according to the order of the liquid medium categories in the adjacent time data change characteristics to obtain the liquid medium fusion characteristics.

[0081] Construct a second correlation neural network; the structure of the second correlation neural network is the same as that of the correlation neural network.

[0082] Among them, in this embodiment, the correlation neural network and the second correlation neural network are long short-term memory neural networks (LSTM).

[0083] Based on the adjacent time data change characteristics and the liquid medium fusion characteristics, use the second correlation neural network to adjust the parameters in the correlation neural network.

[0084] Optionally, based on the adjacent time data change characteristics and the liquid medium fusion characteristics, using the second correlation neural network to adjust the parameters in the correlation neural network includes:

[0085] Input the adjacent time data change characteristics into the correlation neural network, calculate the loss with the liquid medium fusion characteristics, and train the correlation neural network.

[0086] Among them, the adjacent time data change characteristics are input into an associated neural network to obtain first prediction data. The loss is calculated by taking the first prediction data and the liquid medium fusion characteristics, and the associated neural network is trained.

[0087] Among them, the cross-entropy loss function is used to calculate the loss.

[0088] The positions of the same probe working time points corresponding to the adjacent time data change characteristics and the liquid medium fusion characteristics are jointly adjusted to train the associated neural network.

[0089] The adjacent time data change characteristics are input into a second associated neural network, and the loss is calculated by taking the liquid medium fusion characteristics, and the second associated neural network is trained.

[0090] Among them, the adjacent time data change characteristics are input into a second associated neural network to obtain third prediction data; the loss is calculated by taking the third prediction data and the liquid medium fusion characteristics, and the second associated neural network is trained.

[0091] Among them, the cross-entropy loss function is used to calculate the loss.

[0092] The parameters at the corresponding positions of the second neural network and the first neural network are fused to adjust the first neural network to obtain an associated neural network.

[0093] Among them, in this embodiment, the averaging method is used for fusion.

[0094] Optionally, the jointly adjusting the positions of the same probe working time points corresponding to the adjacent time data change characteristics and the liquid medium fusion characteristics to train the associated neural network includes:

[0095] Obtain a plurality of position pairs; the position pair includes a first position and a second position; the first position is a random position in the adjacent time data change characteristics; the second position is a position other than the second position in the adjacent time data change characteristics.

[0096] Among them, the position pairs in this embodiment are [2,7], [4,9], [5,8], [3,6].

[0097] The values of the plurality of position pairs in the adjacent time data change characteristics are sequentially exchanged to obtain transformed adjacent time data change characteristics.

[0098] Among them, as in this embodiment, the value with the subscript 2 and the value with the subscript 7 in the adjacent time data change feature are exchanged to obtain the first transformed adjacent time data change feature; the value with the subscript 4 and the value with the subscript 9 in the first transformed adjacent time data change feature are exchanged to obtain the second transformed adjacent time data change feature; the value with the subscript 5 and the value with the subscript 8 in the second transformed adjacent time data change feature are exchanged to obtain the third transformed adjacent time data change feature; the value with the subscript 3 and the value with the subscript 8 in the third transformed adjacent time data change feature are exchanged to obtain the transformed adjacent time data change feature.

[0099] Successively exchange the values of multiple position pairs in the liquid medium fusion feature to obtain the transformed liquid medium fusion feature.

[0100] Among them, as in this embodiment, the value with the subscript 2 and the value with the subscript 7 in the liquid medium fusion feature are exchanged to obtain the first transformed liquid medium fusion feature; the value with the subscript 4 and the value with the subscript 9 in the first transformed liquid medium fusion feature are exchanged to obtain the second transformed liquid medium fusion feature; the value with the subscript 5 and the value with the subscript 8 in the second transformed liquid medium fusion feature are exchanged to obtain the third transformed liquid medium fusion feature; the value with the subscript 3 and the value with the subscript 8 in the third transformed liquid medium fusion feature are exchanged to obtain the transformed liquid medium fusion feature.

[0101] Input the transformed adjacent time data change feature into the correlation neural network, calculate the loss with the transformed liquid medium fusion feature, and train the first neural network.

[0102] Among them, input the transformed adjacent time data change feature into the correlation neural network to obtain the second predicted data; calculate the loss between the second predicted data and the transformed liquid medium fusion feature, and train the first neural network.

[0103] Among them, the cross-entropy loss function is used to calculate the loss.

[0104] Optionally, obtaining the liquid medium bubble feature based on the first matrix includes:

[0105] Obtain the first convolution kernel; the first convolution kernel is a three-dimensional convolution kernel of 2*n*m; n is the number of columns of the first matrix; m is the number of pages of the first matrix.

[0106] Among them, n and m are positive integers.

[0107] On the rows of the first matrix, perform convolution of the first convolution kernel on the first matrix with a step size of 1 to obtain the liquid medium bubble feature.

[0108] Among them, convolution is performed through the convolution method in a three-dimensional convolutional neural network (3D Convolutional Neural Networks, 3D CNN).

[0109] Optionally, the performing convolution based on the second matrix to obtain the adjacent time data change feature includes:

[0110] Obtain a 2×2 second convolution kernel;

[0111] On the rows of the second matrix, with a stride of 1, perform convolution of the second convolution kernel on the first matrix to obtain the adjacent time data change feature.

[0112] Among them, convolution is performed through the convolution method in a convolutional neural network (Convolutional Neural Networks, CNN).

[0113] Optionally, the obtaining multiple first matrices based on multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and corresponding bubble radii includes:

[0114] Sequentially associate the probe operating time points from small to large with the rows in the first matrix with subscripts from small to large, sequentially associate the probe operating frequencies from small to large with the pages in the first matrix with subscripts from small to large, and sequentially associate the bubble radii from small to large with the columns in the first matrix with subscripts from small to large;

[0115] Multiple liquid medium categories respectively obtain multiple first matrices;

[0116] In the first matrix corresponding to the liquid medium category, mark the positions of the probe operating frequency, the probe operating time point, and the corresponding bubble radius as 1.

[0117] Among them, in the two-dimensional matrix formed by the column and the page of the first matrix corresponding to one probe operating time point in the first matrix corresponding to the liquid medium category, the position where 1 is located represents the bubble radius that changes with the change of the probe operating frequency.

[0118] Optionally, the obtaining multiple second matrices based on multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating durations, and the first matrix includes:

[0119] Sequentially associate the probe operating time points from small to large with the rows in the second matrix with subscripts from small to large, associate the liquid medium category with the first column of the second matrix, and associate the probe operating frequency with the second column of the second matrix.

[0120] Multiple liquid medium categories respectively obtain multiple second matrices;

[0121] Take the liquid medium category corresponding to the first matrix as the first liquid medium category.

[0122] Fill the first column of the second matrix with the first liquid medium category.

[0123] Fill the probe operating frequency in the first matrix into the second column of the same row in the second matrix.

[0124] Example 2

[0125] The acoustic wave action time can also be considered: the cavitation threshold is related to the length of the acoustic wave action time. When designing the debridement process, the duration of the acoustic wave needs to be considered to ensure sufficient time to generate the cavitation effect.

[0126] Enhance the cavitation effect of the liquid medium: The cavitation effect in the working fluid can be enhanced by increasing the power output of the ultrasonic vibration device. This can be achieved by adjusting the composition of the liquid medium or adding specific additives to make it easier to generate cavitation bubbles at a given power.

[0127] Optimize the liquid circulation: Improving the electrolyte circulation can enhance the cavitation effect. This can be achieved by designing a more efficient liquid circulation system or adding components that promote liquid flow.

[0128] Utilize micro-nucleus cavitation bubbles: The ultrasonic cavitation effect is based on the micro-nucleus cavitation bubbles existing in the liquid. The generation of the cavitation effect can be promoted by adding an appropriate amount of micro-bubbles to the liquid medium or selecting a liquid containing appropriate micro-nuclei.

[0129] Adjust the sound pressure: When the sound pressure reaches a certain value, the cavitation bubbles will undergo the process of growth - acceleration - closure - rupture. The cavitation effect can be enhanced by adjusting the acoustic characteristics of the liquid medium to make it easier to reach the required sound pressure value at a given host frequency.

[0130] Example 3

[0131] Based on the above method for optimizing the liquid medium of an ultrasonic probe based on big data, an embodiment of the present invention further provides a system for optimizing the liquid medium of an ultrasonic probe based on big data, the system includes an acquisition module, a feature extraction module, a feature association module, a clustering module, a detection module, and a cavitation detection module.

[0132] The acquisition module is used to acquire multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and corresponding bubble radii; the liquid medium category represents the category of the liquid medium where the handheld ultrasonic probe is located; the probe operating frequency represents the frequency of the handheld ultrasonic probe; the probe operating time point represents the duration of the handheld ultrasonic probe in the liquid medium; the bubble radius represents the radius of the bubble caused by the handheld ultrasonic probe.

[0133] A feature extraction module, configured to obtain adjacent time data change features and multiple liquid medium bubble features based on the multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points, and the corresponding medium bubble radii; the adjacent time data change features represent the change situations of the liquid medium categories and probe operating frequencies in adjacent probe operating time points; the liquid medium bubble features represent the change situations of the bubble radii corresponding to different probe operating frequencies in adjacent probe operating time points;

[0134] A feature association module, configured to construct an association relationship between the adjacent time data change features and the multiple liquid medium bubble features through an association neural network;

[0135] A clustering module, configured to perform clustering based on the multiple liquid medium bubble features to obtain detected liquid medium categories; the detected liquid medium categories are the clustering centers;

[0136] A detection module, configured to use the bubble radius that is greater than other bubble radii in the detected liquid medium categories as the first bubble radius; use the probe operating frequency corresponding to the first bubble radius as the optimal probe operating frequency, and the corresponding probe operating time point as the optimal probe operating time point;

[0137] A cavitation detection module, configured to obtain a variable cavitation threshold according to the first bubble radius.

[0138] Regarding the system in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0139] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The structure required to construct such a system will be apparent from the above description. In addition, the present invention is not directed to any specific programming language. It should be understood that the content of the present invention described herein can be implemented using various programming languages, and the description of the specific language above is for disclosing the best mode of the present invention.

[0140] In the specification provided herein, a large number of specific details are set forth. However, it can be understood that the embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and technologies have not been shown in detail so as not to obscure the understanding of this specification.

[0141] Each component embodiment of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the device according to the embodiments of the present invention. The present invention can also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program for implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

Claims

1. A method for optimizing liquid medium using an ultrasonic probe based on big data, characterized in that: include: Obtain multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points and corresponding bubble radii; The liquid medium category indicates the category of the liquid medium in which the handheld ultrasonic probe is located; the probe operating frequency indicates the frequency of the handheld ultrasonic probe; the probe operating time point indicates the time length of the handheld ultrasonic probe in the liquid medium; the bubble radius indicates the radius of the bubble caused by the handheld ultrasonic probe; Based on the multiple liquid medium categories, multiple probe working frequencies, multiple probe working time points and corresponding medium bubble radii, adjacent time data change characteristics and multiple liquid medium bubble characteristics are obtained; The adjacent time data change characteristics represent the changes in the liquid medium type and the probe working frequency at adjacent probe working time points; the liquid medium bubble characteristics represent the changes in the bubble radius corresponding to different probe working frequencies at adjacent probe working time points; By associating neural networks, the correlation between the changing characteristics of adjacent time data and the characteristics of multiple liquid medium bubbles is constructed; Clustering is performed according to the plurality of liquid medium bubble characteristics to obtain a detected liquid medium category; The detected liquid medium category is the cluster center; The bubble radius that is larger than the other bubble radius in the detected liquid medium category is used as the first bubble radius; The probe operating frequency corresponding to the first bubble radius is used as the optimal probe operating frequency, and the corresponding probe operating time point is used as the optimal probe operating time point; A variable cavitation threshold is obtained according to the first bubble radius.

2. The method for optimizing liquid medium using an ultrasonic probe based on big data according to claim 1, characterized in that: The liquid medium bubble characteristics are obtained by associating a neural network based on the liquid medium type, the probe working frequency, the probe working time point and the corresponding bubble radius, including: Based on multiple liquid medium categories, multiple probe working frequencies, multiple probe working time points and corresponding bubble radii, multiple first matrices are obtained; the first matrix is ​​a three-dimensional matrix; one liquid medium category corresponds to one first matrix; Based on multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating durations and the first matrix, a second matrix is ​​obtained; the second matrix is ​​a two-dimensional matrix; Based on the first matrix, a liquid medium bubble characteristic is obtained; a plurality of first matrices correspond to a plurality of liquid medium bubble characteristics; Based on the second matrix, the adjacent time data variation characteristics are obtained.

3. The method for optimizing liquid medium using ultrasonic probe based on big data according to claim 1, characterized in that: The method of constructing the correlation relationship between the adjacent time data change characteristics and the liquid medium bubble characteristics through the correlation neural network includes: Sort multiple liquid medium bubble features according to the order of liquid medium categories in the adjacent time data change features to obtain liquid medium fusion features; Constructing a second associative neural network; wherein the structure of the second associative neural network is the same as the structure of the associative neural network; Based on the changing characteristics of adjacent time data and the fusion characteristics of the liquid medium, a second associative neural network is used to adjust the parameters in the associative neural network.

4. The method for optimizing liquid medium using an ultrasonic probe based on big data according to claim 3, characterized in that: The method of using a second associative neural network to adjust parameters in the associative neural network based on the adjacent time data change characteristics and the liquid medium fusion characteristics includes: Inputting the adjacent time data change characteristics into the associated neural network, fusing the characteristics with the liquid medium to obtain the loss, and training the associated neural network; The positions of the same probe working time points corresponding to the adjacent time data variation characteristics and the liquid medium fusion characteristics are jointly adjusted to train the associated neural network; Inputting the adjacent time data change characteristics into a second association neural network, fusing the characteristics with the liquid medium to obtain the loss, and training the second association neural network; The parameters of the second associative neural network and the corresponding positions of the associative neural network are merged to adjust the associative neural network.

5. The method for optimizing liquid medium using ultrasonic probe based on big data according to claim 4, characterized in that: The method of jointly adjusting the positions of the same probe working time points corresponding to the adjacent time data variation characteristics and the liquid medium fusion characteristics, and training the associated neural network, comprises: A plurality of position pairs are obtained; the position pairs include a first position and a second position; the first position is a random position in the adjacent time data change feature; the second position is a position in the adjacent time data change feature other than the second position; sequentially exchanging the values ​​of multiple position pairs in the adjacent time data change feature to obtain a transformed adjacent time data change feature; sequentially exchanging the values ​​of multiple position pairs in the liquid medium fusion feature to obtain a transformed liquid medium fusion feature; The transformed adjacent time data change characteristics are input into the associated neural network, and the loss is calculated by fusing the transformed liquid medium characteristics to train the first neural network.

6. The method for optimizing liquid medium using ultrasonic probe based on big data according to claim 2, characterized in that: The step of obtaining the bubble characteristics of the liquid medium based on the first matrix includes: Obtain a first convolution kernel; the first convolution kernel is a 2*n*m three-dimensional convolution kernel; n is the number of columns of the first matrix; m is the number of pages of the first matrix; On the rows of the first matrix, the first convolution kernel is convolved on the first matrix with a step size of 1 to obtain the liquid medium.

7. The method for optimizing liquid medium using ultrasonic probe based on big data according to claim 2, characterized in that: The convolution is performed based on the second matrix to obtain the adjacent time data change characteristics, including: Get the second convolution kernel of 2*2; On the rows of the second matrix, the second convolution kernel is convolved on the first matrix with a step size of 1 to obtain the adjacent time data change characteristics.

8. The method for optimizing liquid medium using ultrasonic probe based on big data according to claim 2, characterized in that: Based on the multiple liquid medium categories, the multiple probe working frequencies, the multiple probe working time points and the corresponding bubble radii, multiple first matrices are obtained, including: The probe working time points from small to large are sequentially associated with the rows with small to large subscripts in the first matrix, the probe working frequencies from small to large are sequentially associated with the pages with small to large subscripts in the first matrix, and the bubble radii from small to large are sequentially associated with the columns with small to large subscripts in the first matrix; A plurality of first matrices are obtained corresponding to a plurality of liquid medium categories; In the first matrix corresponding to the liquid medium category, the positions of the probe working frequency, the probe working time point and the corresponding bubble radius are marked as 1.

9. The method for optimizing liquid medium using ultrasonic probe based on big data according to claim 2, characterized in that: The method of obtaining a plurality of second matrices based on a plurality of liquid medium categories, a plurality of probe operating frequencies, a plurality of probe operating durations and a first matrix includes: The probe working time points from small to large are sequentially associated with the rows with subscripts from small to large in the second matrix, the liquid medium category is associated with the first column of the second matrix, and the probe working frequency is associated with the second column of the second matrix; A plurality of second matrices are obtained corresponding to a plurality of liquid medium categories; Using the liquid medium category corresponding to the first matrix as the first liquid medium category; Filling the first column of the second matrix with the first liquid medium category; Fill the probe operating frequency in the first matrix into the second column of the same row as in the second matrix.

10. A system for optimizing liquid medium using ultrasonic probe based on big data, characterized in that: include: An acquisition module, used to acquire multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points and corresponding bubble radii; The liquid medium category indicates the category of the liquid medium in which the handheld ultrasonic probe is located; the probe operating frequency indicates the frequency of the handheld ultrasonic probe; the probe operating time point indicates the time length of the handheld ultrasonic probe in the liquid medium; the bubble radius indicates the radius of the bubble caused by the handheld ultrasonic probe; A feature extraction module, for obtaining adjacent time data variation features and multiple liquid medium bubble features based on the multiple liquid medium categories, multiple probe operating frequencies, multiple probe operating time points and corresponding medium bubble radii; The adjacent time data change characteristics represent the changes in the liquid medium type and the probe working frequency at adjacent probe working time points; the liquid medium bubble characteristics represent the changes in the bubble radius corresponding to different probe working frequencies at adjacent probe working time points; A feature association module is used to construct an association relationship between the change characteristics of adjacent time data and the characteristics of multiple liquid medium bubbles through an association neural network; A clustering module, used for clustering the plurality of liquid medium bubble features to obtain a detected liquid medium category; The detected liquid medium category is the cluster center; A detection module, used to use a bubble radius greater than other bubble radii in the detected liquid medium category as a first bubble radius; use a probe operating frequency corresponding to the first bubble radius as an optimal probe operating frequency, and use a corresponding probe operating time point as an optimal probe operating time point; The cavitation detection module is used to obtain a variable cavitation threshold according to the first bubble radius.

Images