Ultrasonic scalpel rod temperature control method and system based on temperature distribution function model

1 . A temperature control method for a blade shaft of an ultrasonic scalpel based on a temperature distribution function model, comprising the following steps:

saving the temperature distribution function model and at least one threshold value;

inputting a corresponding input feature to the temperature distribution function model, and outputting corresponding temperature data information;

comparing at least one value in the temperature data information with the threshold value; and

adjusting, based on the comparison result, a power level applied to a transducer of the ultrasonic scalpel to modulate a temperature of the blade shaft of the ultrasonic scalpel,

wherein the temperature distribution function model is a neural network algorithm model, comprising one or a combination of more than one algorithm models of a feedforward neural network, a memory neural network, and an attention neural network; and a training method for the model is one or a combination of more than one of a supervised learning, a semi-supervised learning, an unsupervised learning, and a reinforcement learning, and

wherein the training method for the model specifically comprises: extracting an input feature from a training set and inputting the same into the neural network algorithm model to calculate an intermediate value and a gradient value for each neuron, wherein a loss function of the model is a mean square error MSE or an average absolute error MAE; updating a weight by a gradient descent method; repeating the foregoing process until the model meets a predetermined stop condition; and stopping training and saving the model after the stop condition is met.

2 . The method according to claim 1 , wherein the input feature of the temperature distribution function model comprises one or a combination of more than one of a working feedback parameter, a physical structure feature parameter, and an environmental parameter.

3 . The method according to claim 2 , wherein the working feedback parameter comprises, but is not limited to a real-time voltage U, a real-time current I, a power P, an impedance R, and a real-time response frequency f; the physical structure feature parameter comprises, but is not limited to a material of the blade shaft of the ultrasonic scalpel, and a length of the blade shaft; and the environmental parameter comprises, but is not limited to an environmental temperature, and an environmental humidity.

4 . The method according to claim 1 , wherein the temperature data information comprises a real-time temperature value at any point on the blade shaft of the ultrasonic scalpel, and/or a maximum temperature value, a minimum temperature value, and an average temperature of a certain area on the blade shaft of the ultrasonic scalpel.

5 . The method according to claim 1 , wherein the temperature distribution function model consists of layers, corresponding neurons, and weights; wherein weight parameters and application programs are saved in a memory of a generator; the memory comprises a flash, an EEPROM, or another non-volatile storage device; the application program runs in a processor; and the processor comprises an ARM, a DSP, a FPGA, a CPU, a GPU, or an ASIC chip existing in the generator, or is a remote server connected through a network.

6 . The method according to claim 1 , wherein said outputting corresponding temperature data information forms a one-dimensional spatial temperature distribution T ( 1 ) along the blade shaft of the ultrasonic scalpel, which is a solution of an equation.

7 . A temperature control system for a blade shaft of an ultrasonic scalpel based on a temperature distribution function model, comprising:

a storage unit, configured to save the temperature distribution function model and at least one threshold value;

a processing unit, configured to input a corresponding input feature to the temperature distribution function model, and output corresponding temperature data information;

a comparison unit, configured to compare at least one value in the temperature data information with the threshold value; and

an adjusting unit, configured to adjust, based on the comparison result, a power level applied to a transducer of the ultrasonic scalpel to modulate the temperature of the blade shaft of the ultrasonic scalpel,

wherein the temperature distribution function model is a neural network algorithm model, comprising one or a combination of more than one algorithm models of a feedforward neural network, a memory neural network, and an attention neural network; and a training method for the model is one or a combination of more than one of a supervised learning, a semi-supervised learning, an unsupervised learning, and a reinforcement learning, and

wherein the training method for the model specifically comprises: extracting an input feature from a training set and inputting the same into the neural network algorithm model to calculate an intermediate value and a gradient value for each neuron, wherein a loss function of the model is a mean square error MSE or an average absolute error MAE; updating a weight by a gradient descent method; repeating the foregoing process until the model meets a predetermined stop condition; and stopping training and saving the model after the stop condition is met.

8 . A generator for temperature control based on a temperature distribution function model, comprising:

a control circuit coupled to a memory, wherein the control circuit is configured to be able to:

save the temperature distribution function model and at least one threshold value;

input a corresponding input feature to the temperature distribution function model, and output corresponding temperature data information;

compare at least one value in the temperature data information with the threshold value; and

adjust, based on the comparison result, a power level applied to a transducer of the ultrasonic scalpel to modulate the temperature of the blade shaft of the ultrasonic scalpel,

wherein the temperature distribution function model is a neural network algorithm model, comprising one or a combination of more than one algorithm models of a feedforward neural network, a memory neural network, and an attention neural network; and a training method for the model is one or a combination of more than one of a supervised learning, a semi-supervised learning, an unsupervised learning, and a reinforcement learning, and

wherein the training method for the model specifically comprises: extracting an input feature from a training set and inputting the same into the neural network algorithm model to calculate an intermediate value and a gradient value for each neuron, wherein a loss function of the model is a mean square error MSE or an average absolute error MAE; updating a weight by a gradient descent method; repeating the foregoing process until the model meets a predetermined stop condition; and stopping training and saving the model after the stop condition is met.

9 . The generator according to claim 8 , wherein the control circuit is configured that the input feature input to the temperature distribution function model comprises one or a combination of more than one of a working feedback parameter, a physical structure feature parameter, and an environmental parameter.

10 . The generator according to claim 9 , wherein the working feedback parameter comprises, but is not limited to a real-time voltage U, a real-time current I, a power P, an impedance R, and a real-time response frequency f; the physical structure feature parameter comprises, but is not limited to a material of the blade shaft of the ultrasonic scalpel, and a length of the blade shaft; and

the environmental parameter comprises, but is not limited to an environmental temperature, and an environmental humidity.

11 . The generator according to claim 8 , wherein the temperature data information comprises a real-time temperature value at any point on the blade shaft of the ultrasonic scalpel, and/or a maximum temperature value, a minimum temperature value, and an average temperature of a certain area on the blade shaft of the ultrasonic scalpel.

12 . An ultrasonic scalpel surgical instrument based on a temperature distribution function model, comprising:

an ultrasonic electromechanical system, comprising an ultrasonic transducer connected to an ultrasonic scalpel through an ultrasonic waveguide; and

a generator, configured to supply power to the ultrasonic transducer, wherein the generator comprises a control circuit configured to be able to:

save the temperature distribution function model and at least one threshold value;

input a corresponding input feature to the temperature distribution function model, and output corresponding temperature data information;

compare at least one value in the temperature data information with the threshold value; and

adjust, based on the comparison result, a power level applied to the transducer of the ultrasonic scalpel to modulate the temperature of a blade shaft of the ultrasonic scalpel,

wherein the temperature distribution function model is a neural network algorithm model, comprising one or a combination of more than one algorithm models of a feedforward neural network, a memory neural network, and an attention neural network; and a training method for the model is one or a combination of more than one of a supervised learning, a semi-supervised learning, an unsupervised learning, and a reinforcement learning, and

wherein the training method for the model specifically comprises: extracting an input feature from a training set and inputting the same into the neural network algorithm model to calculate an intermediate value and a gradient value for each neuron, wherein a loss function of the model is a mean square error MSE or an average absolute error MAE; updating a weight by a gradient descent method; repeating the foregoing process until the model meets a predetermined stop condition; and stopping training and saving the model after the stop condition is met.

13 . The ultrasonic scalpel surgical instrument according to claim 12 , wherein the control circuit is configured that the input feature input to the temperature distribution function model comprises one or a combination of more than one of a working feedback parameter, a physical structure feature parameter, and an environmental parameter; wherein the working feedback parameter comprises, but is not limited to a real-time voltage U, a real-time current I, a power P, an impedance R, and a real-time response frequency f; the physical structure feature parameter comprises, but is not limited to a material of the blade shaft of the ultrasonic scalpel blade, and a length of the blade shaft; and the environmental parameter comprises, but is not limited to an environmental temperature, and an environmental humidity.

14 . The ultrasonic scalpel surgical instrument according to claim 12 , wherein the temperature data information comprises a real-time temperature value at any point on the blade shaft of the ultrasonic scalpel, and/or a maximum temperature value, a minimum temperature value, and an average temperature of a certain area on the blade shaft of the ultrasonic scalpel.

15 . An ultrasonic scalpel system based on a temperature distribution function model, comprising a processor and a non-volatile storage device comprising an application program, wherein the application program, when executed by the processor, enables the processor to:

save the temperature distribution function model and at least one threshold value;

input a corresponding input feature to the temperature distribution function model, and output corresponding temperature data information;

compare at least one value in the temperature data information with the threshold value; and

adjust, based on the comparison result, a power level applied to a transducer of the ultrasonic scalpel to modulate the temperature of a blade shaft of the ultrasonic scalpel,

wherein the temperature distribution function model is a neural network algorithm model, comprising one or a combination of more than one algorithm models of a feedforward neural network, a memory neural network, and an attention neural network; and a training method for the model is one or a combination of more than one of a supervised learning, a semi-supervised learning, an unsupervised learning, and a reinforcement learning, and

wherein the training method for the model specifically comprises: extracting an input feature from a training set and inputting the same into the neural network algorithm model to calculate an intermediate value and a gradient value for each neuron, wherein a loss function of the model is a mean square error MSE or an average absolute error MAE; updating a weight by a gradient descent method; repeating the foregoing process until the model meets a predetermined stop condition; and stopping training and saving the model after the stop condition is met.

16 . The ultrasonic scalpel system according to claim 15 , wherein the temperature distribution function model consists of layers, corresponding neurons, and weights; wherein weight parameters and the application program are saved in a memory of the generator; the memory comprises a flash, an EEPROM, or another non-volatile storage device; the application program runs in a processor; and the processor comprises an ARM, a DSP, a FPGA, a CPU, a GPU, or an ASIC chip existing in the generator, or is a remote server connected through a network.