Multi-sensor fusion driven ultrasonic energy uniform distribution control method and device

Abstract

The present invention relates to the field of ultrasonic medical technology, and to a method and device for controlling the uniform distribution of driving ultrasonic energy by multi-sensor fusion, the method comprising: collecting sensor data through a total sensor; inputting the sensor data into a pre-constructed sensor data recognition model to output a mode and an abnormal value; inputting the mode, sensor data and abnormal value into a pre-constructed ultrasonic parameter adjustment model to output an ultrasonic adjustment parameter; and adjusting the ultrasonic parameters of an ultrasonic therapeutic apparatus according to the ultrasonic adjustment parameter. The present invention uses digital multi-parameters to coordinate and adjust simultaneously, realizes directional and precise distribution of ultrasonic therapy, improves efficacy and reduces side effects.

Description

Technical Field

[0001] The present invention relates to the field of ultrasonic medical technology, and more specifically, to a method and device for controlling uniform distribution of driving ultrasonic energy by multi-sensor fusion. Background Art

[0002] In the beauty industry, ultrasound therapy devices have been widely used in a variety of skin treatment and beauty procedures, such as firming and lifting, wrinkle elimination and skin reshaping. With the development of technology, the accuracy and effect of therapeutic devices have gradually become the focus of users. In this context, the development of "ultrasound gun", "supersonic gun", "beauty sculpture instrument" and 7D and 8D technology shows the in-depth application and continuous innovation of ultrasound technology in the beauty industry.

[0003] The existing Chinese patent with publication number CN115475338A discloses a control method of an ultrasonic therapeutic apparatus and an ultrasonic therapeutic apparatus, and the control method of the ultrasonic therapeutic apparatus includes the following steps: step S10, detecting the information of the activated therapeutic handle unit; step S20, configuring and displaying the operation area divided based on the surface position of the human skin according to the detected information of the activated therapeutic handle unit for user selection; step S30, configuring and displaying the preset working parameters of the therapeutic handle unit according to the operation area selected by the user, and displaying a confirmation prompt box; step S40, when the confirmation prompt box is triggered, controlling the corresponding therapeutic handle unit to work according to the configured working parameters. The invention can provide guidance information to the operator based on the operation requirements of the device, simplify the user's operation, improve the safety and reliability of the ultrasonic therapeutic apparatus, and enable the ultrasonic therapeutic apparatus to achieve the expected treatment effect.

[0004] Although the patent has made some achievements in the safety and reliability of ultrasonic therapeutic devices, there are still some technical challenges that have not been fully resolved, especially in the uniform distribution of ultrasonic energy. Uneven energy distribution may lead to different treatment effects and even damage the skin in some cases. To make the ultrasonic energy distribution uniform, it is necessary to adjust the ultrasonic parameters of the ultrasonic therapeutic device for energy deployment. If the ultrasonic parameters of the ultrasonic therapeutic device are not adjusted properly, it may cause damage to the skin. In addition, existing equipment usually requires operators to have high professional skills to ensure the safety and effectiveness of treatment, which in disguise increases the complexity of operation and dependence on professional training. Summary of the invention

[0005] In order to solve the above technical difficulties, an embodiment of the present invention provides a multi-sensor fusion driven ultrasonic energy uniform distribution control method, the method comprising:

[0006] Collecting sensor data through total sensors;

[0007] Input sensor data to a pre-built sensor data recognition model to output patterns and outliers;

[0008] Input the pattern, sensor data and abnormal values ​​into a pre-built ultrasonic parameter adjustment model to output ultrasonic adjustment parameters;

[0009] The ultrasonic parameters of the ultrasonic therapeutic apparatus are adjusted according to the ultrasonic adjustment parameters.

[0010] Furthermore, the total sensor includes a temperature sensor, a pressure sensor, a fat clamp and an acoustic wave sensor; the sensing data includes the pressure of the treatment head in contact with the skin, the skin temperature, the ultrasonic wave intensity and the fat content.

[0011] Furthermore, the method for constructing the sensor data recognition model includes:

[0012] Convert each set of sensor data into a first feature vector, wherein the elements of the first feature vector include the contact pressure between the treatment head and the skin, the skin temperature, the ultrasonic intensity and the fat content;

[0013] The set of all first feature vectors is used as input of a sensor data recognition model, wherein the sensor data recognition model uses the predicted pattern and outlier for each set of sensor data as output, uses the actual pattern and outlier corresponding to each set of sensor data as prediction targets, and uses minimizing the sum of the first prediction accuracies of all predicted patterns and outliers as a training target;

[0014] Among them, the calculation formula for the first prediction accuracy is: L z =(Yz-S z ) 2 , where L z is the first prediction accuracy, Yz is the predicted pattern and outlier corresponding to the zth group of sensor data, S z The actual pattern and abnormal value corresponding to the zth group of sensor data; the sensor data recognition model is trained until the sum of the first prediction accuracy reaches convergence and the training is stopped; the sensor data recognition model is a CNN model.

[0015] Furthermore, the ultrasonic parameter adjustment model is a DNN model, which includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer; the input dimension of the input layer is n, and the size of n is the number of patterns output by the sensor data recognition model + 4; the number of nodes in the first hidden layer is set to 2n, the size of the weight matrix of the first hidden layer is (2n, n), the size of the bias vector of the first hidden layer is (2n, 1), and the activation function of the first hidden layer is the ReLU function; the number of nodes in the second hidden layer is set to n, and the size of the weight matrix of the second hidden layer is is (n, 2n), the size of the second hidden layer bias vector is (n, 1), and the second hidden layer activation function is the ReLU function; the number of nodes in the third hidden layer is set to n / 2, the size of the third hidden layer weight matrix is ​​(n / 2, n), the size of the third hidden layer bias vector is (n / 2, 1), and the third hidden layer activation function is the ReLU function; the number of nodes in the output layer is m, m is a preset value, the size of the output layer weight matrix is ​​(m, n / 2), the size of the output layer bias vector is (m, 1), and the output layer activation function is the Softmax function.

[0016] Furthermore, the training method of the ultrasonic parameter adjustment model includes:

[0017] All historical ultrasonic parameter adjustment data and the ultrasonic adjustment parameters corresponding to the historical ultrasonic parameter adjustment data are prepared in the form of numpy arrays. The historical ultrasonic parameter adjustment data in each group of numpy arrays includes patterns, sensor data and outliers. The weight matrices and bias vectors of all layers are initialized and prepared for forward propagation: input historical ultrasonic parameter adjustment data, perform forward calculations by level, and calculate the output of each layer; the calculation expression is: Input layer: a 0 =x, hidden layer: a i =σ(W i ×a i-1 +b i ); Output layer: a J =σ(W J ×a J-1 +b J );

[0018] Among them, σ is the activation function; x is the input historical ultrasonic parameter adjustment data; a i is the output vector of the i-th hidden layer neuron; W i is the weight matrix from the i-th hidden layer to the i+1 hidden layer, b i is the bias vector of the i-th hidden layer; a 0 Represents the input value of the input layer; a J represents the output value of the output layer; WJ is the weight matrix of the output layer, b J is the bias vector of the output layer, i and J are both layer numbers, and i≠J;

[0019] Calculate the loss function Loss of each hidden layer i :

[0020] Loss i =-1 / N∑y i ×log(a i )+(1-y i )×log(1-a i );

[0021] N is the total number of input historical ultrasonic parameter adjustment data; y is the historical ultrasonic parameter adjustment data in the hidden layer; i is the hidden layer index; a i is the probability representation of the output of the i-th hidden layer;

[0022] Calculate the loss function Loss of each hidden layer i For the gradient of the parameters of each layer, the Adam optimizer is used to update the parameter values ​​of each layer according to the gradient descent method; the training configuration is: batch size is 32, learning rate is 0.001, and the number of training rounds is 200-500 rounds; when the loss functions of each hidden layer converge, the training is stopped, that is, the training is completed.

[0023] Furthermore, the multi-sensor fusion driven ultrasonic energy uniform distribution control method also includes:

[0024] The blood vessel data is collected, the blood vessel data and the ultrasound adjustment parameters are input into a pre-built ultrasound final adjustment model, and the ultrasound final adjustment parameters are output.

[0025] Furthermore, the blood vessel data includes the number of blood vessels, blood vessel diameter and blood flow.

[0026] Furthermore, the method for constructing the ultrasonic final adjustment model includes:

[0027] Using the blood vessel data and the ultrasonic adjustment parameters as analysis data, pre-collecting the ultrasonic final adjustment parameters corresponding to the c-group analysis data, and converting the analysis data and the corresponding ultrasonic final adjustment parameters into corresponding second eigenvectors;

[0028] Each group of second eigenvectors is used as the input of the ultrasonic final adjustment model. The ultrasonic final adjustment model uses a group of ultrasonic final adjustment parameters corresponding to each group of analysis data as output, and uses the actual ultrasonic final adjustment parameters corresponding to each group of analysis data as the prediction target. The actual ultrasonic final adjustment parameters are the ultrasonic final adjustment parameters corresponding to the analysis data collected in advance. The training target is to minimize the sum of the prediction errors of all analysis data. The calculation formula of the prediction error is Z K =(α K -μ K ) 2, where Z K is the prediction error, K is the group number of the second eigenvector corresponding to the analysis data, α K is the final ultrasonic adjustment parameter corresponding to the Kth group of analysis data, μ K The actual ultrasonic final adjustment parameters corresponding to the Kth group of analysis data are obtained; the ultrasonic final adjustment model is trained until the sum of the prediction errors reaches convergence and the training is stopped;

[0029] The ultrasonic final adjustment model is specifically one of ANN and SVR.

[0030] A multi-sensor fusion driven ultrasonic energy uniform distribution control device, the device comprising:

[0031] Sensor collection module: collects sensor data through the main sensor;

[0032] Pattern and anomaly recognition module: inputs sensor data into the pre-built sensor data recognition model to output patterns and anomalies;

[0033] The first ultrasonic parameter adjustment module: inputs the mode, sensor data and abnormal value into the pre-built ultrasonic parameter adjustment model to output the ultrasonic adjustment parameter;

[0034] The second ultrasonic parameter adjustment module collects blood vessel data, inputs the blood vessel data and ultrasonic adjustment parameters into a pre-built ultrasonic final adjustment model, and outputs the ultrasonic final adjustment parameters;

[0035] Execution module: adjust the ultrasonic parameters of the ultrasonic therapeutic device according to the final adjustment parameters of the ultrasonic wave.

[0036] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the multi-sensor fusion driven ultrasonic energy uniform distribution control method is implemented.

[0037] The technical effects and advantages of the multi-sensor fusion driven ultrasonic energy uniform distribution control method provided by the present invention are as follows:

[0038] The present invention uses a variety of sensors to collect comprehensive parameter data, considers more influencing factors of the human body, and the model more realistically reflects the complex biological system; establishes an end-to-end deep learning framework, realizes integrated processing from signal acquisition to parameter control, automatically learns micro-control rules, reduces the complexity of operation, and gets rid of the dependence on professional training; the sensor data recognition model uses CNN to learn spatial features and accurately captures changes in tissue state; the ultrasonic parameter adjustment model uses DNN to deeply simulate human imaging relationships to improve adjustment accuracy; the vascular data is integrated to build the final model to achieve personalized deep optimization of the target tissue; the overall method is data-driven, has strong online learning capabilities, and can adapt to different populations and environmental changes; uses digital multi-parameters to coordinate and adjust simultaneously to achieve targeted and precise distribution of ultrasonic treatment, improve efficacy and reduce side effects; solves the problem that traditional single-parameter quantitative methods are difficult to cover the complexity of biological systems; brings a new way of facial ultrasound treatment and promotes the development of this technology from macro-quantification to micro-customization and personalization. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a flow chart of the method for controlling uniform distribution of driving ultrasonic energy by multi-sensor fusion in Example 1;

[0040] Figure 2 This is a flow chart of the multi-sensor fusion driven ultrasonic energy uniform distribution control method in Example 2;

[0041] Figure 3 It is a structural schematic diagram of a multi-sensor fusion driven ultrasonic energy uniform distribution control device;

[0042] Figure 4 It is a schematic diagram of an electronic device of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Embodiment 1:

[0045] like Figure 1 As shown, the multi-sensor fusion driven ultrasonic energy uniform distribution control method described in this embodiment includes:

[0046] Collecting sensor data through total sensors;

[0047] Input sensor data to a pre-built sensor data recognition model to output patterns and outliers;

[0048] Input the pattern, sensor data and abnormal values ​​into a pre-built ultrasonic parameter adjustment model to output ultrasonic adjustment parameters;

[0049] The ultrasonic parameters of the ultrasonic therapeutic apparatus are adjusted according to the ultrasonic adjustment parameters.

[0050] The total sensor includes a temperature sensor, a pressure sensor, a fat clamp and an acoustic wave sensor; the temperature sensor: monitors the temperature changes on the surface and inside of the skin to prevent overheating; the pressure sensor: records the pressure of the treatment head in contact with the skin to ensure the uniformity of the treatment; the fat clamp: detects the fat content of the part where the treatment head is located; the acoustic wave sensor: captures the propagation of ultrasound in each layer of the skin to analyze the uniformity of energy distribution; the sensing data includes the pressure of the treatment head in contact with the skin, the skin temperature, the ultrasonic intensity and the fat content; the correct pressure of the treatment head in contact with the skin is conducive to energy conduction, and too high pressure may cause tissue compression damage; when ultrasound is transmitted between the skin, energy spillage will occur, causing the temperature to rise, and too high temperature may cause damage to the tissue; the ultrasonic intensity refers to the ultrasonic intensity between different skin layers and the ultrasonic intensity between different parts of the same skin layer, that is, the longitudinal ultrasonic intensity and the transverse ultrasonic intensity. The ultrasonic intensity reflects the distribution of energy, and uneven energy distribution may lead to different treatment effects, and even cause damage to the skin in some cases; the fat content in different parts of the human face will affect the propagation speed of ultrasound, and the ultrasonic frequency needs to be adjusted according to the fat content so that the ultrasound is focused on the target tissue for facial treatment;

[0051] The pattern refers to the correlation and regularity between various indicators (such as temperature, pressure, and ultrasonic intensity) during the treatment process. Pattern recognition can help determine whether the treatment effect is ideal and whether there are problems that need to be optimized. It serves as the basis for ultrasonic adjustment. For example, when the output power of the treatment head increases, the skin surface temperature and the deep tissue temperature will increase accordingly; however, the greater the pressure, the better the heat energy is conducted, and the higher the degree of deep temperature increase under the same power will be than under light pressure; the temperature of areas with higher ultrasonic intensity is generally higher than that of other areas with lower ultrasonic intensity, and the deep temperature increase is also greater; abnormal values ​​include 0 or 1. If the abnormal value is 1, the ultrasonic adjustment parameter is directly set to the negative value of the current ultrasonic parameter, that is, the ultrasonic treatment is stopped; if the abnormal value is 0, the output of the ultrasonic parameter adjustment model will not be affected throughout the process.

[0052] Ultrasonic adjustment parameters include ultrasonic parameters such as output power, frequency and waveform;

[0053] The method for constructing the sensor data recognition model includes:

[0054] Convert each set of sensor data into a first feature vector, wherein the elements of the first feature vector include the contact pressure between the treatment head and the skin, the skin temperature, the ultrasonic intensity and the fat content;

[0055] The set of all first feature vectors is used as input of a sensor data recognition model, wherein the sensor data recognition model uses the predicted pattern and outlier for each set of sensor data as output, uses the actual pattern and outlier corresponding to each set of sensor data as prediction targets, and uses minimizing the sum of the first prediction accuracies of all predicted patterns and outliers as a training target;

[0056] Among them, the calculation formula for the first prediction accuracy is: L z =(Y z -S z ) 2 , where L z is the first prediction accuracy, Y z is the predicted pattern and outliers corresponding to the z-th group of sensor data, S z The actual pattern and abnormal value corresponding to the zth group of sensor data; the sensor data recognition model is trained until the sum of the first prediction accuracy reaches convergence and the training is stopped; the sensor data recognition model is a CNN model.

[0057] The ultrasonic parameter adjustment model is a DNN model, which includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer; the input dimension of the input layer is n, and the size of n is the number of patterns output by the sensor data recognition model + 4; the number of nodes in the first hidden layer is set to 2n, the size of the weight matrix of the first hidden layer is (2n, n), the size of the bias vector of the first hidden layer is (2n, 1), and the activation function of the first hidden layer is the ReLU function; the number of nodes in the second hidden layer is set to n, and the size of the weight matrix of the second hidden layer is ( n, 2n), the size of the bias vector of the second hidden layer is (n, 1), and the activation function of the second hidden layer is the ReLU function; the number of nodes in the third hidden layer is set to n / 2, the size of the weight matrix of the third hidden layer is (n / 2, n), the size of the bias vector of the third hidden layer is (n / 2, 1), and the activation function of the third hidden layer is the ReLU function; the number of nodes in the output layer is m, m is a preset value, the size of the weight matrix of the output layer is (m, n / 2), the size of the bias vector of the output layer is (m, 1), and the activation function of the output layer is the Softmax function.

[0058] The training method of the ultrasonic parameter adjustment model includes:

[0059] All historical ultrasonic parameter adjustment data and the ultrasonic adjustment parameters corresponding to the historical ultrasonic parameter adjustment data are prepared in the form of numpy arrays. The historical ultrasonic parameter adjustment data in each group of numpy arrays includes patterns, sensor data and outliers. The weight matrices and bias vectors of all layers are initialized and prepared for forward propagation: input historical ultrasonic parameter adjustment data, perform forward calculations by level, and calculate the output of each layer; the calculation expression is: Input layer: a 0 =x, hidden layer: a i =σ(W i ×a i-1 +b i ); Output layer: a J =σ(W J ×a J-1 +b J );

[0060] Among them, σ is the activation function; x is the input historical ultrasonic parameter adjustment data; a i is the output vector of the i-th hidden layer neuron; W i is the weight matrix from the i-th hidden layer to the i+1 hidden layer, b i is the bias vector of the i-th hidden layer; a 0 Represents the input value of the input layer; a J Represents the output value of the output layer; W J is the weight matrix of the output layer, b J is the bias vector of the output layer, i and J are both layer numbers, and i≠J;

[0061] Calculate the loss function Loss of each hidden layer i :

[0062] Loss i =-1 / N∑y i ×log(a i )+(1-y i )×log(1-α i );

[0063] N is the total number of input historical ultrasonic parameter adjustment data; y is the historical ultrasonic parameter adjustment data in the hidden layer; i is the hidden layer index; a i is the probability representation of the output of the i-th hidden layer;

[0064] Calculate the loss function Loss of each hidden layer i For the gradient of the parameters of each layer, the Adam optimizer is used to update the parameter values ​​of each layer according to the gradient descent method; the training configuration is: batch size is 32, learning rate is 0.001, and the number of training rounds is 200-500 rounds; when the loss functions of each hidden layer converge, the training is stopped, that is, the training is completed.

[0065] Example 2

[0066] like Figure 2 As shown, this embodiment further improves the design on the basis of embodiment 1. The difference is that in the multi-sensor fusion driving ultrasonic energy uniform distribution control method provided in embodiment 1, in actual applications, the number and size of blood vessels in different facial areas (such as eyes and nose) have a certain impact on the safety of ultrasound. Different blood vessel richness has different ultrasonic energy absorption. It is necessary to fine-tune the power parameters according to different parts based on the protection of the facial vascular complex, but because the adjustment range is relatively limited compared with embodiment 1, that is, the original general direction adjustment remains unchanged, fine-tuning based on embodiment 1 should not be directly included in the debugging of embodiment 1. Based on this, the multi-sensor fusion driving ultrasonic energy uniform distribution control method provided in this embodiment also includes:

[0067] collecting blood vessel data, inputting the blood vessel data and ultrasound adjustment parameters into a pre-built ultrasound final adjustment model, and outputting the ultrasound final adjustment parameters;

[0068] The vascular data include the number of blood vessels, blood vessel diameter and blood flow; the vascular data can be obtained by pulse blood volume imaging technology and MRI imaging technology. When the treatment head touches the skin, the vascular data of the contact area between the treatment head and the skin is immediately detected; the blood-rich area has a large number of blood vessels, and the ultrasonic energy will be more easily absorbed and scattered by the blood vessels, so the power in this area needs to be reduced; blood vessels with large diameters are more likely to absorb ultrasound, which can easily lead to adverse reactions such as tissue overheating, so the power in this area needs to be reduced; arteries in high blood flow areas may be blocked by proteins due to ultrasonic oscillations, so the power in this area needs to be reduced.

[0069] The method for constructing the ultrasonic final adjustment model comprises:

[0070] Using the blood vessel data and the ultrasonic adjustment parameters as analysis data, pre-collecting the ultrasonic final adjustment parameters corresponding to the c-group analysis data, and converting the analysis data and the corresponding ultrasonic final adjustment parameters into corresponding second eigenvectors;

[0071] Each group of second eigenvectors is used as the input of the ultrasonic final adjustment model. The ultrasonic final adjustment model uses a group of ultrasonic final adjustment parameters corresponding to each group of analysis data as output, and uses the actual ultrasonic final adjustment parameters corresponding to each group of analysis data as the prediction target. The actual ultrasonic final adjustment parameters are the ultrasonic final adjustment parameters corresponding to the analysis data collected in advance. The training target is to minimize the sum of the prediction errors of all analysis data. The calculation formula of the prediction error is Z K =(α K -μ K ) 2, where Z K is the prediction error, K is the group number of the second eigenvector corresponding to the analysis data, α K is the final ultrasonic adjustment parameter corresponding to the Kth group of analysis data, μ K The actual ultrasonic final adjustment parameters corresponding to the Kth group of analysis data are obtained; the ultrasonic final adjustment model is trained until the sum of the prediction errors reaches convergence and the training is stopped;

[0072] The ultrasonic final adjustment model is specifically one of ANN and SVR.

[0073] Example 3

[0074] See also Figure 3 As shown, a multi-sensor fusion driven ultrasonic energy uniform distribution control device comprises:

[0075] Sensor collection module: collects sensor data through the main sensor;

[0076] Pattern and anomaly recognition module: inputs sensor data into the pre-built sensor data recognition model to output patterns and anomalies;

[0077] The first ultrasonic parameter adjustment module: inputs the mode, sensor data and abnormal value into the pre-built ultrasonic parameter adjustment model to output the ultrasonic adjustment parameter;

[0078] The second ultrasonic parameter adjustment module collects blood vessel data, inputs the blood vessel data and ultrasonic adjustment parameters into a pre-built ultrasonic final adjustment model, and outputs the ultrasonic final adjustment parameters;

[0079] Execution module: adjust the ultrasonic parameters of the ultrasonic therapeutic device according to the final adjustment parameters of the ultrasonic wave.

[0080] Example 4

[0081] See also Figure 4 As shown, according to another aspect of the present application, an electronic device 500 is also provided. The electronic device 500 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are run by one or more processors, the driving ultrasonic energy uniform distribution control method of multi-sensor fusion as described above can be executed.

[0082] The system according to the embodiment of the present application can also be used by Figure 4 The electronic device architecture shown in FIG. Figure 4As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, ROM 503, RAM 504, a communication port 505 connected to a network, an input / output 506, a hard disk 507, etc. The storage device in the electronic device 500, such as ROM 503 or hard disk 507, may store the multi-sensor fusion driven ultrasonic energy uniform distribution control method provided in the present application. Further, the electronic device 500 may also include a user interface 508. Of course, Figure 4 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 4 One or more components of an electronic device are shown.

[0083] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0084] Finally: 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 principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-sensor fusion driven ultrasonic energy uniform distribution control method, characterized in that: Methods include: The sensor data is collected by the total sensor; the total sensor includes a temperature sensor, a pressure sensor, a fat clamp and an acoustic wave sensor; the sensor data includes the pressure of the treatment head in contact with the skin, the skin temperature, the ultrasonic intensity and the fat content; The sensor data is input into the pre-built sensor data recognition model to output the mode and abnormal value. The mode refers to the various indicators in the treatment process, including the correlation and regularity between temperature, pressure and ultrasound intensity; the abnormal value includes 0 or 1. If the abnormal value is 1, the ultrasound adjustment parameter is directly set to the negative value of the current ultrasound parameter, that is, the ultrasound treatment is stopped; if the abnormal value is 0, the output of the ultrasound parameter adjustment model is not affected throughout the process; Input the pattern, sensor data and abnormal values ​​into a pre-built ultrasonic parameter adjustment model to output ultrasonic adjustment parameters; The ultrasonic parameters of the ultrasonic therapeutic apparatus are adjusted according to the ultrasonic adjustment parameters.

2. The multi-sensor fusion driven ultrasonic energy uniform distribution control method according to claim 1, characterized in that: The method for constructing the sensor data recognition model includes: Convert each set of sensor data into a first feature vector, wherein the elements of the first feature vector include the contact pressure between the treatment head and the skin, the skin temperature, the ultrasonic intensity and the fat content; The set of all first feature vectors is used as input of a sensor data recognition model, wherein the sensor data recognition model uses the predicted pattern and outlier for each set of sensor data as output, uses the actual pattern and outlier corresponding to each set of sensor data as prediction targets, and uses minimizing the sum of the first prediction accuracies of all predicted patterns and outliers as a training target; Among them, the calculation formula for the first prediction accuracy is: L z =(Y z -S z ) 2 , where L z is the first prediction accuracy, Y z is the predicted pattern and outliers corresponding to the z-th group of sensor data, S z The actual pattern and abnormal value corresponding to the zth group of sensor data; the sensor data recognition model is trained until the sum of the first prediction accuracy reaches convergence and the training is stopped; the sensor data recognition model is a CNN model.

3. The multi-sensor fusion driven ultrasonic energy uniform distribution control method according to claim 2, characterized in that: The ultrasonic parameter adjustment model is a DNN model, which includes an input layer, a first hidden layer, a second hidden layer, a third hidden layer and an output layer; the input dimension of the input layer is n, and the size of n is the number of patterns output by the sensor data recognition model + 4; the number of nodes in the first hidden layer is set to 2n, the size of the weight matrix of the first hidden layer is (2n, n), the size of the bias vector of the first hidden layer is (2n, 1), and the activation function of the first hidden layer is the ReLU function; the number of nodes in the second hidden layer is set to n, and the size of the weight matrix of the second hidden layer is ( n,2n), the size of the bias vector of the second hidden layer is (n,1), and the activation function of the second hidden layer is the ReLU function; the number of nodes in the third hidden layer is set to n / 2, the size of the weight matrix of the third hidden layer is (n / 2,n), the size of the bias vector of the third hidden layer is (n / 2,1), and the activation function of the third hidden layer is the ReLU function; the number of nodes in the output layer is m, m is the preset value, the size of the weight matrix of the output layer is (m,n / 2), the size of the bias vector of the output layer is (m,1), and the activation function of the output layer is the Softmax function.

4. The multi-sensor fusion driven ultrasonic energy uniform distribution control method according to claim 3 is characterized in that: The training method of the ultrasonic parameter adjustment model includes: All historical ultrasonic parameter adjustment data and the ultrasonic adjustment parameters corresponding to the historical ultrasonic parameter adjustment data are prepared in the form of numpy arrays. The historical ultrasonic parameter adjustment data in each group of numpy arrays includes patterns, sensor data and outliers. The weight matrices and bias vectors of all layers are initialized and prepared for forward propagation: input historical ultrasonic parameter adjustment data, perform forward calculations by level, and calculate the output of each layer; the calculation expression is: input layer: a0=x, hidden layer: a i =σ(W i ×a i-1 +b i ); Output layer: a J =σ(W J ×a J-1 +b J ); Among them, σ is the activation function; x is the input historical ultrasonic parameter adjustment data; a i is the output vector of the i-th hidden layer neuron; W i is the weight matrix from the i-th hidden layer to the i+1 hidden layer, b i is the bias vector of the i-th hidden layer; a0 represents the input value of the input layer; a J Represents the output value of the output layer; W J is the weight matrix of the output layer, b J is the bias vector of the output layer, i and J are both layer numbers, and i≠J; Calculate the loss function Loss of each hidden layer i : Loss i =-1 / N∑y i ×log(a i )+(1-y i )×log(1-a i ); N is the total number of input historical ultrasonic parameter adjustment data; y is the historical ultrasonic parameter adjustment data in the hidden layer; i is the hidden layer index; a i is the probability representation of the output of the i-th hidden layer; Calculate the loss function Loss of each hidden layer i For the gradient of each layer parameter, the Adam optimizer is used to update the parameter value of each layer according to the gradient descent method; the training configuration is: batch size is 32, learning rate is 0.001, and the number of training rounds is 200-500 rounds; when the loss function of each hidden layer converges, the training is stopped, that is, the training is completed.

5. The multi-sensor fusion driven ultrasonic energy uniform distribution control method according to claim 1, characterized in that: Also includes: The blood vessel data is collected, the blood vessel data and the ultrasound adjustment parameters are input into a pre-built ultrasound final adjustment model, and the ultrasound final adjustment parameters are output.

6. The multi-sensor fusion driven ultrasonic energy uniform distribution control method according to claim 5, characterized in that: The blood vessel data includes the number of blood vessels, blood vessel diameter and blood flow.

7. The multi-sensor fusion driven ultrasonic energy uniform distribution control method according to claim 6, characterized in that: The method for constructing the ultrasonic final adjustment model comprises: Using the blood vessel data and the ultrasonic adjustment parameters as analysis data, pre-collecting the ultrasonic final adjustment parameters corresponding to the c-group analysis data, and converting the analysis data and the corresponding ultrasonic final adjustment parameters into corresponding second eigenvectors; Each group of second eigenvectors is used as the input of the ultrasonic final adjustment model. The ultrasonic final adjustment model uses a group of ultrasonic final adjustment parameters corresponding to each group of analysis data as output, and uses the actual ultrasonic final adjustment parameters corresponding to each group of analysis data as the prediction target. The actual ultrasonic final adjustment parameters are the ultrasonic final adjustment parameters corresponding to the analysis data collected in advance. The training target is to minimize the sum of the prediction errors of all analysis data. The calculation formula of the prediction error is Z K =(α K -μ K ) 2 , where Z K is the prediction error, K is the group number of the second eigenvector corresponding to the analysis data, α K is the final ultrasonic adjustment parameter corresponding to the Kth group of analysis data, μ K The actual ultrasonic final adjustment parameters corresponding to the Kth group of analysis data are obtained; the ultrasonic final adjustment model is trained until the sum of the prediction errors reaches convergence and the training is stopped; The ultrasonic final adjustment model is specifically one of ANN and SVR.

8. A multi-sensor fusion driven ultrasonic energy uniform distribution control device, characterized in that: The method for controlling uniform distribution of driving ultrasonic energy by multi-sensor fusion according to any one of claims 1 to 7 is implemented, and the device comprises: Sensing collection module: collects sensing data through the total sensor; the total sensor includes a temperature sensor, a pressure sensor, a fat clamp and an acoustic wave sensor; the sensing data includes the pressure of the treatment head in contact with the skin, the skin temperature, the ultrasonic intensity and the fat content; Mode and abnormality recognition module: inputs sensor data into the pre-built sensor data recognition model to output mode and abnormal value; the mode refers to the correlation and regularity between various indicators in the treatment process, including temperature, pressure and ultrasound intensity; abnormal value includes 0 or 1. If the abnormal value is 1, the ultrasound adjustment parameter is directly set to the negative value of the current ultrasound parameter, that is, the ultrasound treatment is stopped; if the abnormal value is 0, the output of the ultrasound parameter adjustment model is not affected throughout the process; The first ultrasonic parameter adjustment module: inputs the mode, sensor data and abnormal value into the pre-built ultrasonic parameter adjustment model to output the ultrasonic adjustment parameter; The second ultrasonic parameter adjustment module collects blood vessel data, inputs the blood vessel data and ultrasonic adjustment parameters into a pre-built ultrasonic final adjustment model, and outputs the ultrasonic final adjustment parameters; Execution module: adjust the ultrasonic parameters of the ultrasonic therapeutic device according to the final adjustment parameters of the ultrasonic wave.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the multi-sensor fusion driven ultrasonic energy uniform distribution control method described in any one of claims 1 to 7 is implemented.

Images