An AI radio frequency detection, verification and identification system and method

Abstract

The present invention provides an AI radio frequency detection verification and identification system and method, which relates to the field of radio frequency treatment, an AI radio frequency detection verification and identification system, including electrodes and radio frequency power supplies, wherein a terminal AI processing module is also provided, and the terminal AI processing module is used to process and analyze the data collected by the electrodes, and dynamically adjust specific radio frequency output parameters according to the feedback data. Through the cooperation between the above structures, the following beneficial effects can be achieved: first, by establishing an impedance database based on individuals, automated data analysis and clustering can be achieved, and personalized impedance detection can be performed using a large amount of data; second, according to the different impedances of each person, different frequency bands and frequency points can be selected through the system for treatment to ensure that energy is effectively absorbed in tissues with different impedances; third, through the AI ​​identification system, the patient's impedance changes can be monitored and analyzed in real time, and the radio frequency energy and frequency can be dynamically adjusted.

Description

Technical Field

[0001] The present invention relates to the field of radio frequency therapy, and in particular to an AI radio frequency detection, verification and identification system and method. Background Art

[0002] Radiofrequency therapy is achieved by delivering radiofrequency current to the target tissue to achieve the treatment purpose. It has been proven that radiofrequency therapy has significant therapeutic effects in the treatment of skin pigmentation, wrinkle removal, skin tightening, local tissue ablation, etc. However, due to differences in subjects, such as differences in subject impedance and capacitance, it is difficult to achieve the best treatment effect for the subject. In some subjects, the expected treatment effect was not achieved, while in other subjects, damage beyond the treatment expectations was shown.

[0003] Chinese patent document CN118384420A discloses a protection control method and device for a high-power short-wave therapeutic device, which records a scheme for setting a temperature sensor for feedback to protect the treated tissue. However, in this scheme, a single-band radio frequency signal is used for treatment, which makes it difficult to provide personalized treatment based on individual differences. For example, during radio frequency treatment, even if the same type of treatment is required in different patients, the response of the tissue to radio frequency may vary due to individual differences and real-time tissue changes, which may result in the treatment effect between some individuals not achieving the expected effect and posing certain safety risks. Summary of the invention

[0004] In view of the deficiencies in the above-mentioned prior art, the technical problem to be solved by the present invention is to provide an AI radio frequency detection, verification and identification system and method, which can perform AI intelligent processing and feedback on the collected data, and select different radio frequency bands for personalized treatment based on personal impedance feedback.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: an AI radio frequency detection, verification and identification system, including electrodes and a radio frequency power supply, wherein a terminal AI processing module is also provided, the terminal AI processing module is used to process and analyze the data collected by the electrodes, and dynamically adjust specific radio frequency output parameters according to the feedback data, and send the optimal radio frequency output parameters to the radio frequency power supply through instructions, and the radio frequency power supply outputs specific radio frequency output parameters according to the instructions.

[0006] In the preferred solution, a cloud database connected to the terminal AI processing module is also provided, which is responsible for storing the individual impedance value data collected by the terminal AI processing into the cloud database, and classifying and labeling the uploaded data; Among them, each piece of classified and labeled data corresponds to the patient's information, tissue type, and ablation site; The terminal AI processing unit and the cloud database work together to exchange data to match different tissue states and determine the corresponding impedance patterns.

[0007] In a preferred embodiment, the electrode is electrically connected to the terminal AI processing module to collect and feed back the RF signal source from the electrode to the terminal AI processing module, and the electrode is electrically connected to the RF power supply to contact the target area of ​​the tissue and send the RF signal; The RF signal is output in a preset manner with a continuously changing RF frequency to obtain measurement data at different frequencies, including current, voltage, electrical impedance, electric power, capacitive reactance, inductive reactance, temperature and reflected signal. The terminal AI processing module analyzes the measurement data to obtain the special response frequency of the corresponding RF signal, and selects the best frequency based on the special response frequency. The preset mode is one of linear scanning or logarithmic scanning; The electrodes include radio frequency electrodes and collection electrodes; The radiofrequency electrode transmits radiofrequency energy to tissues of different depths and ranges by changing the current path and distribution in different working modes; The acquisition electrode is responsible for real-time acquisition of measurement data generated during the interaction between the radiofrequency electrode and the tissue; Among them, different working modes of radiofrequency electrodes include monopolar mode and bipolar mode; In the monopolar mode, the RF electrode acts as the energy transmitting end and cooperates with a specific return electrode to construct a long current path, so that the RF energy penetrates deep into the tissue for therapeutic intervention in deep tissues. In bipolar mode, two adjacent radio frequency electrodes serve as positive and negative electrodes to form a local current loop, focusing energy on the area between the electrodes for superficial or local lesions to control the depth and range of treatment.

[0008] In a preferred solution, the terminal AI processing unit includes: The data acquisition module is responsible for receiving the measurement data transmitted by the acquisition electrodes; A preprocessing unit, which preprocesses the collected measurement data; The feature extraction unit uses a deep learning algorithm to extract features from the preprocessed data, identify and extract the impedance characteristics of different tissues; A trend analysis unit, used to calculate the trend line of the impedance characteristics of different tissues after changing over time using a linear regression model; The feature recognition unit compares and analyzes the calculated impedance characteristics with the annotated historical data in the cloud database, identifies the characteristics of the current tissue, obtains the comparison and analysis results, generates the optimal RF output parameters based on the comparison and analysis results, and sends the optimal RF output parameters to the RF power supply through instructions.

[0009] Dynamic adjustment module, which dynamically adjusts the RF output parameters between different frequencies based on the comparative analysis results and through a specific algorithm to match the characteristics of different tissues; The radio frequency output parameter is a radio frequency point or a frequency band or a specific wavelength; When the RF output parameter is a specific wavelength, the selection of the specific wavelength is automatically determined based on the optimal frequency and through a specific algorithm; Among them, long-wavelength radiofrequency waves are used for deep tissue heating or ablation, and short-wavelength radiofrequency waves are used for shallow tissue treatment.

[0010] In a preferred solution, the cloud database is also provided with a model training module, which uses a clustering algorithm to classify similar patient data to form different impedance patterns, and uses the classified and labeled data as a data set, which is divided into a training set, a validation set, and a test set; The model is trained multiple times based on the training set data, the training results are monitored on the validation set, and tested on the test set to obtain the optimized model, which is then sent to the terminal AI processing unit to improve the model's recognition and prediction capabilities.

[0011] In the preferred solution, the cloud database is also provided with an evaluation feedback module, which is used to evaluate and compare the current data and historical data of the target area of ​​the same organization to obtain evaluation results, and generate a verification and identification report based on the evaluation results, and upload the generated verification and identification report to the cloud database to optimize the model training module.

[0012] In a preferred embodiment, the present invention further provides a method for applying the above-mentioned AI radio frequency detection verification and identification system, which comprises the following steps: S1. Preliminary detection: If the electrode contacts the target tissue area and the contact state lasts for a preset time, it is determined to be effective contact, and the data acquisition module starts to collect data, including current, voltage, and electrical impedance; S2, data analysis: the data acquisition module transmits the collected data to the preprocessing unit; The preprocessing unit preprocesses the data, including removing noise and interference data; A specific deep learning algorithm is used to extract features from the preprocessed data to obtain impedance characteristics of different tissues, wherein the specific deep learning algorithm is a convolutional neural network (CNN); Compare and analyze the extracted impedance characteristics of different tissues with the classified and labeled data stored in the cloud database to identify the characteristics of the current tissue, including fat distribution, muscle density, blood concentration, and epidermal thickness; Calculate the trend line of impedance value changing over time according to the characteristics of the current tissue to obtain the analysis result; S3. Differentiation setting: Calculate and determine the patient's current impedance value change rate based on the analysis results , and obtain multi-factor information of patients, including age ,gender , physical condition , according to the rate of change of the patient's impedance value The relative importance weights of the RF output parameters are determined by using multi-factor information, and the relative importance weights of the RF output parameters are calculated comprehensively to obtain the indicators of the RF output parameters. , according to the indicator Match and determine the optimal RF output parameter combination in the pre-set parameter database to ensure that energy is effectively absorbed in tissues with different impedances, reducing energy loss and the risk of burns; Send instructions to the RF power supply according to the optimal RF output parameter combination.

[0013] S4, output execution: the RF power supply receives the instruction and identifies the instruction as an execution event, outputs the set RF signal through the electrode according to the execution event, and detects the RF signal before executing the output; Among them, the frequency fluctuation range of the RF signal is within ±3%; S5. Data monitoring: Determine the appropriate data acquisition frequency and use the data acquisition module to obtain various parameters during the treatment process in real time to obtain monitoring data, including temperature, current, and voltage; Input real-time monitoring data into the terminal AI processing module to dynamically adjust indicators , and based on the adjusted indicators The values ​​are matched in the parameter database and the RF output parameters are updated.

[0014] S6. Data evaluation: Establish a historical data screening and verification mechanism, use the terminal AI processing module to evaluate and compare the current data of the target area of ​​the same organization with the screened and verified historical data to obtain the evaluation results, and input the evaluation results into the evaluation feedback module for analysis to generate a verification and identification report, and upload the generated verification and identification report to the cloud database; S7, classification and model optimization: adopt a specific clustering algorithm and set appropriate parameters, use the clustering algorithm to identify unlabeled data with similar impedance value change patterns to trigger labeling events, classify the unlabeled data according to the labeling events to obtain correctly labeled data, and organize the correctly labeled data into a data set, adopt the back propagation algorithm in deep learning and optimize the model training module according to the data set; Among them, the specific clustering algorithm is the K-Means clustering algorithm. The initial clustering center of the K-Means clustering algorithm is determined by combining random selection with a density-based method, and the number of clusters is dynamically determined according to the distribution characteristics of the impedance value change pattern.

[0015] In the preferred solution, in step S2, the preprocessing unit preprocesses the data, including using a specific filtering algorithm to remove noise data, using a data smoothing algorithm to process interference data, and according to the normalization formula Normalize the data; in is the original data, and are the minimum and maximum values ​​of the feature in the data set; The convolutional neural network CNN is used to extract features from the preprocessed data, and the output is processed by the softmax function. Convert to probability vector ; in, is the output vector of the fully connected layer in the convolutional neural network CNN. elements, probability vector Indicates the probability that the input data belongs to each tissue type to identify and extract the impedance characteristics of different tissues.

[0016] Compare and analyze the extracted impedance characteristics of different tissues with the classified and labeled data stored in the cloud database to identify the characteristics of the current tissue, including fat distribution, muscle density, blood concentration, and epidermal thickness; Let time point be the independent variable , the impedance value is the dependent variable , the slope is , the intercept is , is a random error term, representing the influence of other unconsidered factors on the impedance value except the time factor, and finally the analysis result is obtained; Slope It is expressed as the change in impedance value per unit time in the impedance characteristic, and its calculation formula is: ; in, is the sample size, and They are Sample and value, and Represented as time points and impedance value The average value of The formula can quantify the rate of change of impedance value over time; In radiofrequency treatment, if the slope , it means the impedance value increases, and the RF energy is increased; If the slope , it means the impedance value is reduced, and the RF energy is reduced; intercept It represents the impedance value when the time point is 0, and its calculation formula is: ; Among them, the intercept Used to determine the impedance level in the initial state; The linear regression model is used to calculate the trend line of the impedance value over time, and the formula is as follows: ; Among them, the impedance value The target variable used to predict or analyze its changing pattern, the time point is , used to convert the impedance value With time point Establish a relationship to observe the changing trend of impedance value.

[0017] In the preferred solution, in step S3, first, a judgment matrix is ​​constructed by using the hierarchical analysis method. : ; in, Indication factors Relative to factors The importance of the can be determined through expert verification; And calculate the weight vector according to the following steps : First, calculate the judgment matrix The product of each row element : ; Among them, the element product Used to combine multiple relative importance evaluation values ​​into an intermediate calculation quantity; Next, calculate of Second root: ; in, For element-wise multiplication The initial operation of normalizing the product of , so that the importance values ​​of different factors are at the same level; Finally, Normalize to get the weight vector ; in, ; According to the final weight vector Determining the relative importance weights of multiple factor information in determining RF output parameters; Comprehensive calculation of RF output parameter indicators The formula is as follows: ; Among them, the indicator It is used to adaptively determine the optimal radiofrequency treatment parameters based on multiple individual factors, reducing the risk of energy loss and burns caused by inadequate consideration of a single factor.

[0018] In the preferred solution, in step S7, the clustering effectiveness evaluation formula is the sum of squared errors It is expressed as follows: ; in, is the number of clusters, Indicates clusters, Is a cluster The data points in the equation are each specific impedance value data sample. It is The center of the cluster is All data points in The mean, sum of squared errors Used to evaluate the effectiveness of clustering and calculate each data point With the cluster center The smaller the value, the smaller the data points in the cluster. The more closely clustered they are in the center of the cluster, The better the clustering effect is, that is, similar impedance value change pattern data are more accurately classified into one category to achieve automated data analysis and clustering based on individual impedance databases; Next, the model is trained in multiple rounds according to the training set data, the training results are monitored on the validation set, and tested on the test set to obtain the optimized model, which is then sent to the terminal AI processing unit to improve the recognition and prediction capabilities of the model. The cross entropy loss function is used in the training process. : ; in, is the number of samples, the size of the dataset used to train the convolutional neural network, is the number of categories, indicating the total number of different tissue types, and the output of the convolutional neural network CNN corresponds to the category , in order to make classification predictions on the input data, Indicates The samples belong to The true label of the class, Indicates The sample is predicted to be Class probabilities; cross entropy loss function Used in the training process to predict the results With the true label Perform a comparison to measure the difference between two probability distributions.

[0019] The present invention provides an AI radio frequency detection verification and identification system and method. Through the cooperation between the above structures, the following beneficial effects can be achieved: First, by establishing an individual-based impedance database, automated data analysis and clustering can be achieved, and personalized impedance testing can be performed using large amounts of data; Second, the system can select different frequency bands and frequency points for treatment according to the different impedances of each person, ensuring that energy is effectively absorbed in tissues with different impedances, reducing energy loss and the risk of burns; Third, through the AI ​​recognition system, the patient's impedance changes are monitored and analyzed in real time, and the radio frequency energy and frequency are dynamically adjusted to achieve the best treatment effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a structural block diagram of the present invention. DETAILED DESCRIPTION

[0021] Example 1 like Figure 1 As shown, an AI RF detection, verification and identification system includes electrodes and RF power supplies, wherein a terminal AI processing module is also provided. The terminal AI processing module is used to process and analyze data collected by the electrodes, and dynamically adjust specific RF output parameters according to the feedback data, and send the optimal RF output parameters to the RF power supply through instructions, and the RF power supply outputs specific RF output parameters according to the instructions.

[0022] Specifically, the RF power supply is a key device used to convert electrical energy into RF energy and transmit it to the electrode. After receiving the instruction, the RF power supply can accurately adjust the frequency and power of the output RF signal to meet the needs of different tissue treatments. RF power supply mainly involves parameter output related to RF frequency band (frequency range) and frequency point (specific frequency value). The RF frequency range is high-frequency alternating current from 3kHz to 300GHz. The relationship between frequency and wavelength is expressed as follows: ; Where c is the speed of light and f is the frequency; Furthermore, RF frequency output: The RF power supply can output specific RF frequencies according to the instructions of the terminal AI processing module, which can be determined according to the treatment needs of different tissues. For example, for epidermal tissue treatment, the terminal AI instructs the RF power supply to output a 27MHz frequency based on tissue characteristics analysis. This frequency is suitable for the epidermis and superficial dermis. After receiving the instructions, the RF power supply uses internal frequency synthesizers and other components to accurately adjust the output frequency to 27MHz, thereby providing suitable RF energy for epidermal tissue treatment.

[0023] RF band output: RF power supply can also output RF signals within a specific frequency band, requiring the RF power supply to output in a specific frequency band, such as 110MHz. This frequency band selection is based on the deep scanning of human tissue and electrical characteristics analysis instructions to ensure that the RF energy is effectively transmitted to the target tissue area.

[0024] In the preferred solution, a cloud database is also provided which is electrically connected to the terminal AI processing module, and is responsible for storing the individual impedance value data collected by the terminal AI processing into the cloud database, and classifying and labeling the uploaded data, and using the cloud database for data storage, model training and evaluation optimization; Among them, each piece of classified and labeled data corresponds to the patient's information, tissue type, and ablation site; The terminal AI processing unit and the cloud database work together to exchange data to match different tissue states and determine the corresponding impedance patterns.

[0025] In a preferred embodiment, the electrode is electrically connected to the terminal AI processing module to collect and feed back the RF signal source from the electrode to the terminal AI processing module, and the electrode is electrically connected to the RF power supply to contact the target area of ​​the tissue and send the RF signal.

[0026] Specifically, frequency modulation through algorithms is the core manifestation of the intelligence of radio frequency treatment systems. The same tissue lesions in different individuals have different optimal absorption points for radio frequency due to individual differences, such as tissue electrical properties and density. Algorithm frequency modulation can find the most suitable frequency for each patient and improve the degree of personalized treatment.

[0027] Frequency and treatment effect: Frequency directly affects the effect of RF energy in tissues. Different frequencies change the absorption, reflection, and penetration characteristics of tissues to RF energy. During treatment, the system dynamically adjusts between frequencies such as 6.1 MHz, 6.3 MHz, and 6.5 MHz based on the energy absorption of the tissues. For example, tissues with high water content absorb high frequencies better, while tissues with high fat content may absorb low frequencies better. The tissue electrical property data collected by the acquisition electrodes plays a decisive role in judging the tissue's response to different frequencies. The changes in the collected voltage and current data can reflect the changing trend of the tissue's energy absorption efficiency at different frequencies, providing an accurate reference for frequency adjustment.

[0028] Characteristics of bands and wavelengths and basis for selection: In the field of radio frequency, frequency is related to wavelength. Different bands (meter wave, decimeter wave, millimeter wave) have different physical characteristics. Long wavelength (low frequency) radio frequency waves have deep penetration but dispersed energy distribution, while short wavelength (high frequency) has concentrated energy but shallow penetration. In actual applications, the appropriate wavelength and band are indirectly selected by adjusting the frequency according to the treatment needs, such as high-frequency short wavelength (millimeter wave band) for shallow tissue treatment, and low-frequency long wavelength (decimeter wave band) for deep tissue heating or ablation. The system automatically completes the selection by the algorithm based on comprehensive factors such as tissue depth, type and treatment purpose. The acquisition of electrode data is the basis for the algorithm to make accurate decisions to achieve the optimal radiofrequency treatment.

[0029] Frequency scanning to determine wavelength is based on the frequency-wavelength relationship of the RF signal, that is, the frequency and wavelength are still inversely proportional in media such as human tissue, and the different response characteristics of tissues to RF signals of different frequencies. The core is to change the frequency of the RF signal and monitor the tissue response to find a wavelength suitable for specific treatment or detection.

[0030] The RF signal can be output in a preset manner (linear or logarithmic scanning) to continuously change the RF frequency within a set frequency range, such as 1MHz10MHz. The frequency change method and range are determined according to the experimental or treatment requirements to obtain measurement data at different frequencies.

[0031] It should be noted that the measured data changes with the sweep frequency. Since the efficiency of electrical processes such as tissue ion movement and dipole rotation is different at different frequencies, the impedance frequency curve can be obtained by measuring tissue impedance at different frequencies. The impedance changes significantly at certain frequencies, which can provide clues for determining the appropriate wavelength; Radiofrequency energy is absorbed by tissues and converted into heat energy, causing tissue temperature to rise. When sweeping the frequency, use thermocouples, fiber optic temperature sensors, etc. to monitor temperature changes and understand the thermal effects of radiofrequency energy at different frequencies. For treatments such as radiofrequency ablation, find the frequency (wavelength) that can quickly heat the target tissue to the treatment temperature, such as a local tissue ablation temperature of 50~100℃. The RF signal is partially reflected when it enters the tissue. By analyzing the intensity and phase of the reflected signal, the tissue reflection characteristics can be inferred, which are related to the tissue interface characteristics (tissue layer boundaries) and the overall electrical uniformity. By monitoring the frequency changes of the reflected signal, the weak reflected signal at certain frequencies means that the RF energy can more easily penetrate the tissue at this frequency, which helps to determine the wavelength suitable for penetrating deep into the tissue.

[0032] When sweeping the frequency, the measured current, voltage, electrical impedance, electric power, capacitive reactance, inductive reactance, temperature, reflected signal and other data and the corresponding frequency information are recorded to obtain the measurement data, which are collected and stored by the computer system. The measurement data are then analyzed using a specific algorithm, such as drawing a curve of parameter changes over frequency, finding characteristic points such as the peak value, valley value, and turning point of the curve, and obtaining the special response frequency of the radio frequency signal of the corresponding tissue, and selecting the optimal frequency based on the special response frequency.

[0033] Among them, the treatment selects the frequency that makes the target tissue achieve the best treatment effect, such as the highest temperature, the best impedance matching and the efficient energy absorption; the detection selects the frequency that makes the tissue characteristics, such as the difference between lesions and normal tissues, the best frequency is determined from data analysis, and the best wavelength is calculated based on the frequency-wavelength relationship to determine the most suitable treatment wavelength The electrodes include radio frequency electrodes and collection electrodes; The radiofrequency electrode transmits radiofrequency energy to tissues of different depths and ranges by changing the current path and distribution in different working modes; The acquisition electrode is responsible for real-time acquisition of measurement data generated during the interaction between the RF electrode and the tissue in different working modes, reflecting the changes in the internal state of the tissue, helping the system to determine tissue characteristics, evaluate treatment effects, and perform mode switching or parameter adjustments when necessary; Among them, different working modes of radiofrequency electrodes include monopolar mode and bipolar mode; In monopolar mode, the RF electrode acts as an active electrode, and the current flows out of the RF electrode, passes through the human tissue, and then returns to the device through the large-area dispersed electrode. Due to the long current path, the RF energy can penetrate deep into the tissue. It can be figuratively described as the energy spreading outward from the active electrode as the center, similar to the water waves propagating deep into the water when a stone is dropped into the water. It is suitable for deep tissue treatment, such as local tissue ablation, and can deliver energy to local tissue to destroy cells. In bipolar mode, two adjacent radio frequency electrodes are used as the positive and negative poles to form a current loop. The current is concentrated in the tissue area between the two electrodes, just like focusing energy in a small range to avoid excessive diffusion of energy to deep tissues. For example, in the treatment of skin lesions, it can accurately act on the target tissue, control the treatment depth, and reduce damage to deep normal tissues.

[0034] In the preferred solution, the terminal AI processing unit includes: The data acquisition module is responsible for collecting the measurement data generated during the interaction between electrodes and tissues in real time; Specifically, the data acquisition module is responsible for collecting various measurement data generated during the interaction between electrodes and human tissues in real time. It continuously collects data at a certain frequency to ensure the timeliness and integrity of the data. The collected data will be passed to the subsequent preprocessing unit for processing. Furthermore, during the operation, the data acquisition module collects data several times per second so that it can capture the dynamic changes of tissues under the action of radio frequency in a timely manner, such as the changing trend of impedance value after the tissue is heated.

[0035] A preprocessing unit, which preprocesses the collected measurement data; Specifically, the collected raw measurement data is preprocessed, including removing noise and interference data from the data. The noise and interference come from electromagnetic interference in the surrounding environment and the electrical noise of the equipment itself. Through filtering and other technical means, the quality and accuracy of the data can be improved, making subsequent analysis and processing more reliable. Furthermore, digital filters are used to remove noise signals in a specific frequency range to extract effective signals that truly reflect tissue characteristics.

[0036] The feature extraction unit uses a deep learning algorithm to extract features from the preprocessed data, identify and extract the impedance characteristics of different tissues; Specifically, deep learning algorithms are used to extract features from preprocessed measurement data. Deep learning algorithms have powerful automatic feature learning capabilities and can mine hidden feature patterns from complex data. Furthermore, deep learning algorithms can identify the unique features of different tissues in measurement data by constructing a convolutional neural network (CNN), such as the impedance spectrum characteristics of different tissues and the specific patterns of impedance changes over time. This can accurately identify and extract the impedance characteristics of different tissues, which can be used to distinguish different types of tissues and provide key basis for subsequent treatment decisions.

[0037] A trend analysis unit, used to calculate the trend line of the impedance characteristics of different tissues after changing over time using a linear regression model; It should be noted that the linear regression model is used to determine the linear relationship between impedance value and time through analysis of historical data, that is, to calculate parameters such as the rate of change of impedance value over time. Furthermore, during the radiofrequency treatment process, the impedance of the tissue may change with the passage of time and the input of energy. The trend analysis unit can predict the trend of this change, and judge in advance the tissue's absorption of radiofrequency energy and possible changes, such as whether the impedance will rise sharply due to overheating, so as to adjust the treatment parameters in time.

[0038] The feature recognition unit compares and analyzes the calculated impedance characteristics with the annotated historical data in the cloud database, identifies the characteristics of the current tissue, obtains the comparative analysis results, and sends the optimal RF output parameters to the RF power supply through instructions.

[0039] Specifically, the calculated impedance characteristics are compared and analyzed with the annotated historical data in the cloud database. The cloud database stores a large amount of patient data, including information such as tissue types, impedance characteristics, and corresponding treatment effects of different patients. By comparing the impedance characteristics of the current tissue with the historical data, the specific characteristics of the current tissue can be identified, and the currently collected tissue measurement data can be matched and compared with the known tissue measurement data in the database to determine the degree of similarity, thereby obtaining comparative analysis results. Based on these results, the optimal RF output parameters can be determined, and instructions can be sent to the RF power supply to achieve personalized precision treatment.

[0040] The dynamic adjustment module dynamically adjusts the RF output parameters between different frequencies based on the comparative analysis results and through a specific algorithm to match the characteristics of different tissues.

[0041] The radio frequency output parameter is a radio frequency point or a frequency band or a specific wavelength; When the RF output parameter is a specific wavelength, the selection of the specific wavelength is automatically determined based on the optimal frequency and through a specific algorithm; Among them, long-wavelength radiofrequency waves are used for deep tissue heating or ablation, and short-wavelength radiofrequency waves are used for shallow tissue treatment.

[0042] Furthermore, since different tissues have different absorption and response characteristics to RF energy, for example, different tissues have different electrical characteristics and require different RF output parameters to achieve the best treatment effect, the dynamic adjustment module will update the RF parameters in real time according to the changes in tissue characteristics to ensure that the best energy transfer and treatment effect are always maintained during the treatment process. For example, when it is found that the impedance value of the tissue changes, indicating that it does not absorb the current RF frequency well, the dynamic adjustment module will adjust to a more appropriate frequency in time to improve the treatment efficiency.

[0043] In the preferred solution, the cloud database is also provided with a model training module, which uses a clustering algorithm to classify similar patient data to form different impedance patterns, and uses the classified and labeled data as a data set, which is divided into a training set, a validation set, and a test set; The model is then trained multiple times based on the training set data. During the training process, the model continuously adjusts its parameters by learning the features and patterns in the data to improve its ability to recognize the impedance characteristics of different tissues. The training results are monitored on the validation set to promptly detect problems such as overfitting or underfitting in the model and make adjustments. Finally, the model is tested on the test set to evaluate the generalization ability and accuracy of the model, and the optimized model is obtained and sent to the terminal AI processing unit to improve its recognition and prediction capabilities, enabling it to more accurately analyze and process newly collected data, providing more reliable decision support for clinical treatment.

[0044] In the preferred embodiment, the cloud database is also provided with an evaluation feedback module, which is used to evaluate and compare the current data and historical data of the target area of ​​the same tissue to obtain the evaluation results. Further, the measurement data during the current tissue treatment process is compared with the data of the patient's previous treatment stage, or with the historical data of other similar cases, to analyze the changes in the tissue and the progress of the treatment effect, and generate a verification and identification report based on the evaluation results. The verification and identification report records in detail the content, results and related analysis conclusions of the evaluation, and uploads the generated verification and identification report to the cloud database. The verification and identification report can provide feedback information for subsequent model training to optimize the model training module. Furthermore, if it is found that the evaluation results of certain types of tissues under specific treatment conditions deviate from the model predictions, then the feedback information can be used to adjust and optimize the model to improve the accuracy and reliability of the model.

[0045] Example 2 Further illustrate with reference to Example 1, Figure 1 The present invention also provides a method for applying the above-mentioned AI radio frequency detection verification and identification system, which includes the following steps: S1. Preliminary testing: After the electrode contacts the target tissue area, the contact status monitoring is started. When the contact pressure between the electrode and the target tissue area is within the preset pressure value range and lasts for 1-2 seconds, it is determined to be effective contact. At this time, the data acquisition module starts to collect data, including current, voltage, and electrical impedance information; The initial acquisition frequency setting rules are as follows: If the target tissue is epidermal tissue or a smaller tissue area, the data acquisition frequency of the data acquisition module is initially set to once every 0.1 seconds, so as to accurately capture its impedance characteristics that may change rapidly; If the target tissue is deep tissue or a large tissue area, the data acquisition frequency is initially set to once every 0.3 seconds, which ensures data validity while reducing unnecessary data redundancy.

[0046] S2. Data analysis: The data acquisition module transmits the collected data to the preprocessing unit; The preprocessing unit preprocesses the data, including using a specific filtering algorithm to remove noise data, using a data smoothing algorithm to process interference data, and using a normalization formula Normalize the data; in is the original data, and are the minimum and maximum values ​​of the feature in the data set; Convolutional neural network CNN is used to extract features from preprocessed data. The architecture and operation of convolutional neural network CNN are as follows: Assume the input data is , set as a two-dimensional tensor, such as the representation of electrical impedance data in time and space dimensions, the convolution kernel is , and set multiple convolution kernels of different sizes. The output after convolution is: Perform convolution operations; Specifically, the convolution kernel is like a filter that can extract local features from the input data by sliding on the input data and performing the above multiplication and summation operations. For example, different convolution kernels can capture tissue impedance characteristics corresponding to different frequency components. If the size of the convolution kernel is 3×3, it will operate on a small 3×3 area of ​​the input data each time. By adjusting the weight value in the convolution kernel, it can be made more sensitive to specific patterns, where the specific pattern is the impedance change pattern of a certain tissue within a specific time and space range.

[0047] And connect the ReLU activation function after the convolution layer To introduce nonlinear characteristics; in, is the output of the convolutional layer; Specifically, with the introduction of nonlinear characteristics, if there is no activation function, the stacking of multiple convolutional layers is still just a linear transformation and cannot learn complex feature representations. The ReLU activation function truncates all values ​​less than 0 to 0, allowing the convolutional neural network CNN to learn more complex nonlinear relationships. For example, in the impedance characteristics, there may be some threshold characteristics, and certain specific tissue responses will only appear when the impedance value changes by more than a certain degree. ReLU can help the convolutional neural network CNN capture this nonlinear feature.

[0048] Assume the pooling window size is 2×2, for the input data , the output after pooling adopts the maximum pooling operation, and its formula is expressed as follows: Perform downsampling operation; in, , The pooling step size is dynamically adjusted according to the data characteristics to reduce the data dimension and retain the key features; Specifically, the output of the convolutional layer is downsampled to reduce the dimension of the data while retaining the most important features. For example, when processing measurement data, there may be a large amount of local similar information. Max pooling can select the most significant feature in each small area, that is, the maximum value, which can reduce the noise and redundant information in the data and make the convolutional neural network CNN more focused on representative feature areas. For example, if there are impedance values ​​at multiple time points in a small area, max pooling can highlight the features corresponding to the time points with the most significant impedance changes in the area, which helps the convolutional neural network CNN focus on the key moments of tissue response.

[0049] In the fully connected layer is the input vector, is the weight matrix, is the bias vector, the output vector ; After passing through multiple convolutional and pooling layers, the data is flattened into a one-dimensional vector and then input into the fully connected layer. The neurons in the fully connected layer are connected to all neurons in the previous layer. It can combine the local features extracted previously to form a high-level abstract representation of the entire input data. For example, the impedance feature fragments at different time and spatial positions have been extracted in the previous convolutional and pooling layers. The fully connected layer combines these fragments to construct a complete abstract representation of tissue impedance characteristics. The output dimension of the last fully connected layer can be based on the number of categories of tissue impedance characteristics that need to be identified. To identify the impedance characteristics of fat, muscle, blood, and epidermal tissue, the output dimension of the last fully connected layer can be set to , the output can be expressed as a probability vector, and the output can be converted into the probability corresponding to each category through an appropriate activation function, that is, the output is converted into the probability corresponding to each category through the softmax function Convert to probability vector ,in is the output vector of the last fully connected layer. elements, probability vector Indicates the probability that the input data belongs to each tissue type, thereby identifying and extracting the impedance characteristics of different tissues.

[0050] Compare and analyze the extracted impedance characteristics of different tissues with the classified and labeled data stored in the cloud database to identify the characteristics of the current tissue, including fat distribution, muscle density, blood concentration, and epidermal thickness; Let time point be the independent variable , the impedance value is the dependent variable , the slope is , the intercept is , is a random error term, which represents the influence of other factors on the impedance value other than time. In actual situations, the change of impedance value may be affected by a variety of complex factors, including tissue heterogeneity, individual physiological fluctuations, and measurement errors. It can make the model more in line with the actual situation, improve the generalization ability and accuracy of the model, avoid the wrong estimation of the impedance value change trend due to failure to consider other factors, and finally obtain the analysis results; Slope It is expressed as the change in impedance value per unit time in the impedance characteristic, and its calculation formula is: ; in, is the sample size, and They are Sample and value, and Represented as time points and impedance value The average value of The formula can quantify the rate of change of impedance value over time; In radiofrequency treatment, if the slope , which means that the impedance value increases over time, indicating that the tissue's absorption of RF energy causes its electrical properties to change, so that the RF energy can be adjusted accordingly, such as increasing the RF energy to maintain the treatment effect; If the slope , it means that the impedance value is reduced and the RF energy needs to be reduced to avoid overtreatment or adverse effects.

[0051] intercept It represents the impedance value when the time point is 0, and its calculation formula is: ; Specifically, the intercept Determining the impedance level in the initial state provides a starting benchmark for the entire impedance value change trend line, which helps to more comprehensively describe the relationship between impedance value and time. It is of great significance for analyzing the electrical characteristics of tissues in the initial stage and comparing them with subsequent changes.

[0052] The linear regression model is used to calculate the trend line of the impedance value over time, and the formula is as follows: ; Among them, the impedance value It is the target variable for predicting or analyzing its changing rules. In the AI ​​RF detection verification and identification system, the impedance value Changes over time or other factors are critical for determining tissue properties and subsequent adjustments in RF treatment parameters; Time When studying the impedance value change trend, it is a key factor. With time point By establishing a relationship, the dynamic changes in the electrical properties of tissues under the action of radiofrequency over time can be observed.

[0053] S3, Differentiated settings: Determine the patient's current impedance value change rate based on the analysis results , and obtain multi-factor information of patients, including age ,gender , physical condition ; Constructing Judgment Matrix by AHP : ; in, Indication factors Relative to factors The degree of importance can be determined through expert verification. The formula reflects the subjective assessment of the relative importance of different factors in determining RF output parameters. For example, when judging the importance of impedance value and age for RF settings, if impedance value is considered to be three times more important than age; Then let the impedance value be factor 1 and age be factor 2, then: ; at the same time , to ensure the consistency requirements of the judgment matrix; And calculate the weight vector according to the following steps : First, calculate the judgment matrix The product of each row element : ; Among them, the element product Used to combine multiple relative importance evaluation values ​​into an intermediate calculation quantity; Next, calculate of Second root: ; in, For element-wise multiplication The initial operation of normalizing the product of , so that the importance values ​​of different factors are at the same level; Finally, Normalize to get the weight vector ; in, ; According to the final weight vector The relative importance of multiple factors in determining RF output parameters is clarified; Comprehensive calculation of RF output parameter indicators The formula is as follows: ; According to the indicator Match, search and determine the optimal RF output parameter combination in the pre-set parameter database to ensure that energy can be effectively absorbed in tissues with different impedances, reducing energy loss and the risk of burns; Send instructions to the RF power supply according to the optimal RF output parameter combination.

[0054] S4. Output execution: The RF power supply receives the instruction and recognizes the instruction as an execution event. It outputs the set RF signal through the electrode according to the execution event, and performs stability and safety tests on the RF signal before output. The frequency fluctuation range of the RF signal is required to be within ±3%. This is because radio frequency (RF) is an electromagnetic wave, and the accuracy of its frequency has a vital impact on the treatment effect. Different frequencies can produce different heating effects on tissues, and then act on different tissue layers. At the same time, the overvoltage protection threshold is set to 110% of the rated voltage, and the overcurrent protection threshold is set to 120% of the rated current to ensure that the output RF signal is stable and safe.

[0055] Specifically, radio frequency (RF) can heat tissues at specific frequencies, and different frequencies can affect different tissue layers. For example, the deeper fat layer may require lower frequency radio frequency waves, while the epidermis may be suitable for higher frequencies. Due to its tissue characteristics and deep location, the deeper subcutaneous fat layer may require lower frequency radio frequency waves, such as 1MHz, which can more effectively penetrate into the deep fat tissue and produce appropriate thermal effects, promote the metabolism of fat cells or change their physical state, thereby achieving treatment goals such as fat reduction; For the dermis and part of the subcutaneous fat layer, the frequency of 448kHz is more suitable. It can accurately transmit energy to this layer of tissue while taking into account a certain depth. The higher frequency of 27MHz is suitable for the epidermis and superficial dermis. This frequency can produce relatively concentrated energy in the epidermis and superficial dermis, stimulate the production of collagen, improve skin texture, etc. The epidermis layer uses a frequency of 7MHz alone, which is mainly used to treat some epidermal lesions or cosmetic needs, such as improving the texture and firmness of the epidermis, while ensuring effective effects on the epidermis and avoiding unnecessary effects on deep tissues.

[0056] Among them, strict protection thresholds are also set in terms of voltage and current. The overvoltage protection threshold is set to 110% of the rated voltage, which means that once the voltage output by the RF power supply exceeds 1.1 times the rated voltage, the overvoltage protection mechanism will be activated immediately to prevent excessive voltage from causing damage to the equipment and tissue. Excessive voltage may cause problems such as excessive heating of the tissue, increased risk of electric shock, and damage to the internal components of the device. The overcurrent protection threshold is set to 120% of the rated current. When the current exceeds 1.2 times the rated current, the overcurrent protection function takes effect. Excessive current may cause abnormal heating and burns of the tissue and overload failure of the equipment. By limiting the current within a safe range, it can be ensured that energy can be delivered stably and safely during the treatment of tissues at different depths, whether for high impedance or low impedance tissues. For example, in the process of voltage regulation, the output power is changed by adjusting the voltage, which has good adaptability to tissues with different impedances. High impedance tissues may require higher voltages to drive enough current to achieve effective energy input, while low impedance tissues require relatively low voltages to avoid excessive energy input. In terms of current modulation, the output power is changed by adjusting the current, which is mainly suitable for tissues at different depths. Deeper tissues may require larger currents to ensure that the energy can penetrate to the corresponding depth and produce the expected therapeutic effect, while shallow tissues need to properly control the current size to prevent overheating. Frequency modulation optimizes energy input by adjusting the frequency. According to the sensitivity and demand differences of different types of tissues to the frequency, the most appropriate frequency is selected, such as the frequency selection for different levels of tissues mentioned above, to achieve precise treatment.

[0057] It is further expressed as executing the RF frequency point by voltage regulation, current regulation and frequency regulation; Voltage regulation: The output power can be changed by adjusting the voltage, which is suitable for tissues with different impedances; Current regulation: The output power can be changed by adjusting the current, which is suitable for tissues at different depths; Frequency modulation: Optimize energy input by adjusting the frequency, suitable for different types of tissues; Specifically, in laser therapy, the principle is to use light energy to act on the skin, and different wavelengths are highly targeted.

[0058] The wavelength of 1064nm is usually used for the treatment of dark pigmentation or deep tissue. For some deep spots, such as nevus of Ota, the pigment particles are often located in the deep layer of the skin. The 1064nm laser can penetrate to the corresponding depth and be selectively absorbed by the pigment particles. The pigment particles are broken up through the photothermal effect and then gradually removed by the body's own metabolic system. At the same time, in some treatments involving deep tissue repair or stimulation, such as the improvement of deep wrinkles, this wavelength can also play a role, promoting the regeneration and remodeling of deep collagen through thermal stimulation; For shallow melanin, a wavelength of 755nm is used. This wavelength has a good effect on shallow melanin deposition problems such as freckles and sun spots. It can effectively decompose shallow melanin particles and promote skin whitening and spot fading.

[0059] The wavelength of 532nm is more suitable for the treatment of epidermal red pigments, such as facial red blood streaks. Since red pigments are mainly concentrated in the epidermis, the 532nm laser can accurately act on the red pigments in the epidermis, causing physical or chemical changes after absorbing light energy, thereby reducing or eliminating the red blood streak phenomenon. Deep hemoglobin uses either 585nm or 590nm. For example, when treating certain deep vascular lesions, lasers in this wavelength range can penetrate into tissues to a certain depth, acting on the hemoglobin in the diseased blood vessels to coagulate and seal the blood vessels, thereby improving related skin problems such as port-wine stains. Compared with other wavelengths, it has better safety and effectiveness in the treatment of deep hemoglobin, and reduces damage to surrounding normal tissues.

[0060] S5. Data monitoring: During the treatment process, determine the appropriate data acquisition frequency, adjust the data acquisition frequency of the data acquisition module to collect data once every 0.1 seconds, and use the data acquisition module to obtain various parameters during the treatment process in real time, including temperature, current, and voltage; The real-time monitoring data is input into the terminal AI processing module, which recalculates the impedance value change trend and related parameters in multiple factors based on the new data and dynamically adjusts the indicators. , and based on the adjusted The value is matched again in the frequency band and frequency database to find and update the RF frequency, thereby realizing real-time dynamic adjustment of the RF frequency.

[0061] S6. Data evaluation: Construct a screening and verification mechanism for historical data, use the terminal AI processing module to evaluate and compare the current data of the target area of ​​the same organization with the screened and verified historical data to obtain the evaluation results, and input the evaluation results into the evaluation feedback module for analysis to generate a verification and identification report, and upload the generated verification and identification report to the cloud database. The screening mechanism for historical data is to remove data that is more than 1.5 years away from the current time and data that differs from the basic physical characteristics of the current patient by more than the preset difference value. The verification mechanism uses an independent test set to verify the accuracy of the historical data, and the test set accounts for 25%~35% of the total historical data.

[0062] S7, classification and model optimization: The K-Means clustering algorithm is used and appropriate parameters are set. The K-Means clustering algorithm is used to identify unlabeled data with similar impedance value change patterns to trigger labeling events. The unlabeled data are classified according to the labeling events to obtain correctly labeled data, and the correctly labeled data are organized into a data set. The data set is divided into a training set, a validation set, and a test set. The clustering effectiveness evaluation formula is the sum of squared errors It is expressed as follows: ; in, is the number of clusters, different The value will divide the data into different numbers of groups. Indicates Clusters are formed, and the unlabeled data with similar impedance value change patterns are classified into the same cluster through the clustering algorithm. Is a cluster The data points in the equation are each specific impedance value data sample. It is The center of the cluster is All data points in The mean of the cluster centers in the K-Means algorithm It is a core concept that is constantly updated and adjusted, and the data points According to the cluster center The distances are assigned to different clusters, and the cluster centers Also based on the data point The error sum of squares is recalculated according to the change of It is used to evaluate the effectiveness of clustering. The calculation principle is to calculate each data point With the cluster center The smaller the value, the closer the data points are to the cluster. The more closely clustered they are in the center of the cluster, The better the clustering effect is, the more accurately the similar impedance value change pattern data are classified into one category. In this embodiment, the error sum of squares is used. The effectiveness of the K-Means clustering algorithm in classifying data with similar impedance value change patterns can be evaluated to ensure the formation of reasonable impedance patterns so that these patterns can be used to make personalized treatment decisions in the future, such as selecting appropriate frequency bands and frequency points, thereby realizing automated data analysis and clustering based on individual impedance databases, improving the system's ability to identify and process differences in impedance between different individuals, and achieving the best treatment effect and reducing risks; The model is trained multiple times based on the training set data, the training results are monitored on the validation set, and tested on the test set to obtain the optimized model, which is then sent to the terminal AI processing unit to improve the recognition and prediction capabilities of the model. The cross entropy loss function is used in the training process. : ; in, is the number of samples, which indicates the size of the data set used to train the convolutional neural network. A sufficient number of samples is crucial for the effective training of the network because it can provide rich information for the network to learn the characteristic patterns of impedance characteristics of different tissues. Too few samples may lead to overfitting of the network, that is, the network is too adapted to the training data and cannot be generalized to new data. Too many samples may increase the training time and computing resource requirements. is the number of categories, for example, the number of categories of different tissue types. In this embodiment, different tissue types include fat, muscle, blood, epidermis, etc. The value of is the total number of tissue types. The output of the convolutional neural network CNN needs to correspond to these categories in order to classify and predict the input data. Indicates The samples belong to The true label of the class, that is, the pre-labeled information in the training data, is the learning target of the convolutional neural network CNN. For example, if a sample actually belongs to muscle tissue, then the label position corresponding to the muscle tissue category , other category labels are 0, Indicates The sample is predicted to be The probability of the class, during the training process, will predict the results With the true label For comparison, through the cross entropy loss function Measures the difference between two probability distributions; Specifically, for each sample, its true label distribution is a discrete distribution with only one category as 1, which indicates the category to which the sample belongs, and other categories as 0, while the predicted output is a probability distribution , this difference can be quantified by calculating the cross entropy loss function.

[0063] Cross Entropy Loss Function It is used to train the convolutional neural network (CNN), measure the difference between the prediction results of the convolutional neural network (CNN) and the actual label, and minimize this loss function. It uses optimization algorithms such as stochastic gradient descent (SGD) or its variants to update the parameters of the convolutional neural network (CNN), such as the weights of the convolution kernel, the weights and biases of the fully connected layers, etc. During the training process, the learning rate is adaptively adjusted so that the network can converge faster and more stably. In each iteration, the parameters of the convolutional neural network CNN are adjusted by back propagation according to the gradient of the loss function to the parameters, so that the convolutional neural network CNN gradually learns the parameter settings that can accurately classify the tissue impedance characteristics, thereby improving the recognition accuracy of the convolutional neural network CNN for different tissue impedance characteristics. For example, the prediction result of a certain tissue sample With the true label If the difference is large, the loss function value will be large. Through back propagation, the convolutional neural network CNN will adjust the parameters to reduce this difference. After multiple iterations of training, the convolutional neural network CNN can better extract and identify the impedance characteristics of the tissue; The labeled training data, including known tissue types and their corresponding current, voltage, and electrical impedance data, are input into the convolutional neural network CNN for multiple iterations of training. In each iteration, the value of the loss function is calculated, and the network parameters are updated according to the optimization algorithm. Through continuous training, the convolutional neural network CNN gradually learns the ability to extract different tissue impedance characteristics from the input data. The number of training rounds and batch size can be set. For example, 100 rounds of training with a batch size of 32 means that 32 samples are used as a group of input network training each time. After the training of the convolutional neural network CNN is completed, new unlabeled samples containing current, voltage, and electrical impedance data are input into the trained convolutional neural network CNN. Through a series of operations of the previously constructed convolutional layer, pooling layer, and fully connected layer, the last layer output of the convolutional neural network CNN is the prediction result of the impedance characteristics of different tissues. For example, the output may be a probability vector, which represents the probability that the sample belongs to fat, muscle, blood, epidermis and other tissues, thereby realizing the extraction and identification of impedance characteristics of different tissues. These extracted tissue impedance characteristics are then compared and analyzed with the classified and labeled data stored in the cloud database, and further used for subsequent treatment decisions, such as the setting of radio frequency points.

[0064] The following is a table showing the effects of personalized treatment using the present invention. 21 cases were randomly selected and their names were hidden.

[0065] Table 1 shows the personalized treatment effects of wrinkles Clinical trial personnel gender Skin problems Individual resistance value (variation range) Impedance change rate before treatment α Number of treatments RF frequency Treatment duration User Experience A001 female Crow's feet 45-85Ω 0.05 2 times 6MHz 30min Slight improvement in skin firmness A002 female Frown lines 48-88Ω 0.06 4 times 6.2MHz 30min Slight improvement in skin firmness A003 female Nasal folds 50-75Ω 0.03 3 times 6.78MHz 30min Slight improvement in skin firmness A004 male Fine lines around eyes 48-70Ω 0.02 10 times 6.5MHz 30min Significant improvement in skin firmness A005 female Forehead wrinkles 39-65Ω 0.05 14 times 6.3MHz 30min Skin firmness is visibly improved A006 female Frown lines 46-92Ω 0.03 17 times 5.9MHz 30min Skin firmness is significantly improved A007 male Fine lines around eyes 45-87Ω 0.02 2 times 6.48MHz 30min Slight improvement in skin firmness A008 female Frown lines 42-85Ω 0.01 9 times 6.1MHz 30min Skin firmness is visibly improved A009 male Frown lines 40-75Ω 0.02 3 times 6.6MHz 30min Slight improvement in skin firmness A0010 female Fine lines around eyes 45-85Ω 0.05 1 time 6.3MHz 30min Slight improvement in skin firmness A0011 male Fine lines around eyes 43-82Ω 0.03 11 times 6.3MHz 30min Skin firmness is visibly improved A0012 male Nasal folds 45-95Ω 0.07 10 times 5.8MHz 30min Skin firmness is visibly improved A0013 female Forehead wrinkles 44-76Ω 0.03 4 times 6.4MHz 30min Slight improvement in skin firmness A0014 female Forehead wrinkles 43-80Ω 0.05 13 times 6.3MHz 30min Improved skin firmness A0015 female Fine lines around eyes 45-88Ω 0.01 16 times 5.9MHz 30min Significant improvement in skin firmness A0016 female Nasal folds 38-86Ω 0.02 2 times 6.48MHz 30min Slight improvement in skin firmness A0017 male Nasal folds 41-80Ω 0.05 10 times 6.1MHz 30min Skin firmness is visibly improved A0018 female Fine lines around eyes 35-70Ω 0.03 3 times 6.6MHz 30min Slight improvement in skin firmness A0019 female Fine lines around eyes 50-86Ω 0.07 17 times 6.3MHz 30min Significant improvement in skin firmness A0020 female Forehead wrinkles 51-89Ω 0.08 14 times 6.7MHz 30min Skin firmness is visibly improved A0021 female Nasal folds 48-85Ω 0.03 13 times 6.3MHz 30min Skin firmness is visibly improved Table 2 shows the effects of personalized treatment of acne Clinical trial personnel gender Acne Problems Individual resistance value (variation range) Impedance change rate before treatment α Number of treatments RF frequency band Treatment duration User Experience B001 female Comedonal acne 75-150Ω 0.09 3 times 2MHz 45min Mild redness of the skin B002 female Inflammatory acne 70-130Ω 0.06 10 times 2.1MHz 45min High comfort B003 female Cystic acne 72-120Ω 0.07 6 times 2.05MHz 45min Mild redness of the skin B004 male Acne conglobata 70-118Ω 0.1 4 times 2.1MHz 45min High comfort B005 female Nodular acne 75-150Ω 0.09 6 times 2MHz 45min High comfort B006 female Inflammatory acne 70-130Ω 0.06 1 time 2.1MHz 45min Mild redness of the skin B007 male Nodular acne 72-120Ω 0.07 4 times 2.05MHz 45min High comfort B008 female Inflammatory acne 70-118Ω 0.1 5 times 2.1MHz 45min Mild redness of the skin B009 male Nodular acne 75-150Ω 0.09 2 times 2MHz 45min High comfort B0010 female Acne conglobata 70-130Ω 0.06 4 times 2.1MHz 45min No obvious skin discomfort B0011 male Nodular acne 72-120Ω 0.07 1 time 2.05MHz 45min High comfort B0012 male Comedonal acne 70-118Ω 0.1 3 times 2.1MHz 45min High comfort B0013 female Comedonal acne 70-130Ω 0.06 1 time 2.1MHz 45min No obvious skin discomfort B0014 female Acne conglobata 72-120Ω 0.07 4 times 2.05MHz 45min Moderate redness of the skin B0015 female Nodular acne 70-118Ω 0.1 2 times 2.1MHz 45min High comfort B0016 female Comedonal acne 75-150Ω 0.09 6 times 2MHz 45min High comfort B0017 male Acne conglobata 70-130Ω 0.06 4 times 2.1MHz 45min No obvious skin discomfort B0018 female Nodular acne 70-130Ω 0.06 6 times 2.1MHz 45min High comfort B0019 female Comedonal acne 72-120Ω 0.07 2 times 2.05MHz 45min No obvious skin discomfort B0020 female Acne conglobata 70-118Ω 0.1 6 times 2.1MHz 45min High comfort B0021 female Inflammatory acne 70-130Ω 0.06 3 times 2.1MHz 45min Mild redness of the skin According to the above data, the solution of the present invention can greatly improve the user experience and treatment effect of radio frequency treatment.

[0066] Example 3 Further illustrate with reference to Examples 1 and 2. Figure 1 The structure shown in the figure, the embodiment of the present application provides an electronic device, which may include: A memory, a processor, and a computer program stored in the memory and executable on the processor.

[0067] When the processor executes the program, an AI radio frequency detection, verification and identification system and method provided in the above embodiment is implemented.

[0068] Furthermore, the electronic device further comprises: Communication interface, used for communication between memory and processor.

[0069] Memory is used to store computer programs that can be run on the processor.

[0070] The memory may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0071] If the memory, processor and communication interface are implemented independently, the communication interface, memory and processor can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0072] The preprocessing unit is a processor, and the processor may include one or more processing units, for example: the processor may include an application processor (application processor, AP), a modem processor, a graphics processor (graphics processing unit, GPU), an image signal processor (image signal processor, ISP), a controller, a memory, a video codec, a digital signal processor (digital signal processor, DSP), a baseband processor, and / or a neural-network processing unit (neural-network processing unit, NPU), etc. Among them, different processing units can be independent devices or integrated in one or more processors. Among them, the controller can be a neural center and a command center. The controller can generate an operation control signal according to the instruction opcode and the timing signal to complete the control of fetching and executing instructions. A memory can also be set in the processor for storing instructions and data. In some embodiments, the memory in the processor is a high-speed cache memory. The memory can save instructions or data that the processor has just used or circulated. If the processor needs to use the instruction or data again, it can be directly called from the memory. Repeated access is avoided, the waiting time of the processor is reduced, and the efficiency of the system is improved.

[0073] Display, for displaying images, videos, etc. The display may include a display panel, which may be a liquid crystal display (LCD), an organic light-emitting diode (OLED), an active-matrix organic light emitting diode or an active-matrix organic light emitting diode (AMOLED), a flexible light-emitting diode (FLED), Miniled, MicroLed, Micro-oLed, a quantum dot light emitting diode (QLED), etc.

[0074] Optionally, in a specific implementation, if the memory, processor and communication interface are integrated on a chip, the memory, processor and communication interface can communicate with each other through an internal interface.

[0075] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned AI radio frequency detection, verification and identification system and method.

[0076] An embodiment of the present application also provides a computer program product, which can run computer instructions, and when the computer instructions are executed by a processor, an AI radio frequency detection, verification and identification system and method as described above are implemented.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0078] Although the present application has been described in conjunction with specific features and embodiments thereof, it is obvious that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and the drawings are merely exemplary illustrations of the present application as defined by the appended claims, and are deemed to have covered any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. An AI radio frequency detection, verification and identification system, comprising electrodes and a radio frequency power supply, characterized in that: A terminal AI processing module is also provided, which is used to process and analyze the data collected by the electrodes, and dynamically adjust specific RF output parameters according to the feedback data, and send the optimal RF output parameters to the RF power supply through instructions. The RF power supply outputs specific RF output parameters according to the instructions.

2. The AI ​​radio frequency detection, verification and identification system according to claim 1 is characterized in that: There is also a cloud database connected to the terminal AI processing module electrical signal, which is responsible for storing the individual impedance value data collected by the terminal AI processing into the cloud database, and classifying and labeling the uploaded data; Among them, each piece of classified and labeled data corresponds to the patient's information, tissue type, and ablation site; The terminal AI processing unit and the cloud database work together to exchange data to match different tissue states and determine the corresponding impedance patterns.

3. The AI ​​radio frequency detection, verification and identification system according to claim 2 is characterized in that: The electrode is electrically connected to the terminal AI processing module to collect and feed back the RF signal source from the electrode to the terminal AI processing module, and the electrode is electrically connected to the RF power supply to contact the target area of ​​the tissue and send the RF signal; The RF signal is output in a preset manner with a continuously changing RF frequency to obtain measurement data at different frequencies, including current, voltage, electrical impedance, electric power, capacitive reactance, inductive reactance, temperature and reflected signal. The terminal AI processing module analyzes the measurement data to obtain the special response frequency of the corresponding RF signal, and selects the best frequency based on the special response frequency. The preset mode is one of linear scanning or logarithmic scanning; The electrodes include radio frequency electrodes and collection electrodes; The radiofrequency electrode transmits radiofrequency energy to tissues of different depths and ranges by changing the current path and distribution in different working modes; The acquisition electrode is responsible for real-time acquisition of measurement data generated during the interaction between the radiofrequency electrode and the tissue; Among them, different working modes of radiofrequency electrodes include monopolar mode and bipolar mode; In the monopolar mode, the RF electrode acts as the energy transmitting end and cooperates with a specific return electrode to construct a long current path, so that the RF energy penetrates deep into the tissue for therapeutic intervention in deep tissues. In bipolar mode, two adjacent radio frequency electrodes serve as positive and negative electrodes to form a local current loop, focusing energy on the area between the electrodes for superficial or local lesions to control the depth and range of treatment.

4. The AI ​​radio frequency detection, verification and identification system according to any one of claims 1 to 3, characterized in that: The terminal AI processing unit includes: The data acquisition module is responsible for receiving the measurement data transmitted by the acquisition electrodes; A preprocessing unit, which preprocesses the collected measurement data; The feature extraction unit uses a deep learning algorithm to extract features from the preprocessed data, identify and extract the impedance characteristics of different tissues; A trend analysis unit, used to calculate the trend line of the impedance characteristics of different tissues after changing over time using a linear regression model; The feature recognition unit compares and analyzes the calculated impedance characteristics with the annotated historical data in the cloud database, identifies the characteristics of the current tissue, obtains the comparison and analysis results, generates the optimal RF output parameters based on the comparison and analysis results, and sends the optimal RF output parameters to the RF power supply through instructions. Dynamic adjustment module, which dynamically adjusts the RF output parameters between different frequencies based on the comparative analysis results and through a specific algorithm to match the characteristics of different tissues; The radio frequency output parameter is a radio frequency point or a frequency band or a specific wavelength; When the RF output parameter is a specific wavelength, the selection of the specific wavelength is automatically determined based on the optimal frequency and through a specific algorithm; Among them, long-wavelength radiofrequency waves are used for deep tissue heating or ablation, and short-wavelength radiofrequency waves are used for shallow tissue treatment.

5. The AI ​​radio frequency detection verification and identification system according to claim 2 or 3, characterized in that: The cloud database also has a model training module, which uses a clustering algorithm to classify similar patient data to form different impedance patterns, and uses the classified and labeled data as a data set, which is divided into a training set, a validation set, and a test set; The model is trained multiple times based on the training set data, the training results are monitored on the validation set, and tested on the test set to obtain the optimized model, which is then sent to the terminal AI processing unit to improve the model's recognition and prediction capabilities.

6. The AI ​​radio frequency detection, verification and identification system according to claim 5 is characterized in that: The cloud database also has an evaluation feedback module, which is used to compare the current data and historical data of the target area of ​​the same organization to obtain evaluation results, and generate a verification and identification report based on the evaluation results. The generated verification and identification report is uploaded to the cloud database to optimize the model training module.

7. A method applied to the AI ​​radio frequency detection, verification and identification system according to any one of claims 1 to 6, characterized in that: The following steps are involved: S1. Preliminary detection: If the electrode contacts the target tissue area and the contact state lasts for a preset time, it is determined to be effective contact, and the data acquisition module starts to collect data, including current, voltage, and electrical impedance; S2, data analysis: the data acquisition module transmits the collected data to the preprocessing unit; The preprocessing unit preprocesses the data, including removing noise and interference data; A specific deep learning algorithm is used to extract features from the preprocessed data to obtain impedance characteristics of different tissues, wherein the specific deep learning algorithm is a convolutional neural network (CNN); Compare and analyze the extracted impedance characteristics of different tissues with the classified and labeled data stored in the cloud database to identify the characteristics of the current tissue, including fat distribution, muscle density, blood concentration, and epidermal thickness; Calculate the trend line of impedance value changing over time according to the characteristics of the current tissue to obtain the analysis result; S3. Differentiation setting: Calculate and determine the patient's current impedance value change rate based on the analysis results , and obtain multi-factor information of patients, including age ,gender , physical condition , according to the rate of change of the patient's impedance value The relative importance weights of the RF output parameters are determined by using multi-factor information, and the relative importance weights of the RF output parameters are calculated comprehensively to obtain the indicators of the RF output parameters. , according to the indicator Match and determine the optimal RF output parameter combination in the pre-set parameter database to ensure that energy is effectively absorbed in tissues with different impedances, reducing energy loss and the risk of burns; Send instructions to the RF power supply according to the optimal RF output parameter combination. S4, output execution: the RF power supply receives the instruction and identifies the instruction as an execution event, outputs the set RF signal through the electrode according to the execution event, and detects the RF signal before executing the output; Among them, the frequency fluctuation range of the RF signal is within ±3%; S5. Data monitoring: Determine the appropriate data acquisition frequency and use the data acquisition module to obtain various parameters during the treatment process in real time to obtain monitoring data, including temperature, current, and voltage; Input real-time monitoring data into the terminal AI processing module to dynamically adjust indicators , and based on the adjusted indicators The values ​​are matched in the parameter database and the RF output parameters are updated. S6. Data evaluation: Establish a historical data screening and verification mechanism, use the terminal AI processing module to evaluate and compare the current data of the target area of ​​the same organization with the screened and verified historical data to obtain the evaluation results, and input the evaluation results into the evaluation feedback module for analysis to generate a verification and identification report, and upload the generated verification and identification report to the cloud database; S7, classification and model optimization: adopt a specific clustering algorithm and set appropriate parameters, use the clustering algorithm to identify unlabeled data with similar impedance value change patterns to trigger labeling events, classify the unlabeled data according to the labeling events to obtain correctly labeled data, and organize the correctly labeled data into a data set, adopt the back propagation algorithm in deep learning and optimize the model training module according to the data set; Among them, the specific clustering algorithm is the K-Means clustering algorithm. The initial clustering center of the K-Means clustering algorithm is determined by combining random selection with a density-based method, and the number of clusters is dynamically determined according to the distribution characteristics of the impedance value change pattern.

8. The method of the AI ​​radio frequency detection verification and identification system according to claim 7 is characterized in that: In step S2, the preprocessing unit preprocesses the data, including using a specific filtering algorithm to remove noise data, using a data smoothing algorithm to process interference data, and using a normalization formula Normalize the data; in is the original data, and are the minimum and maximum values ​​of the feature in the data set; The convolutional neural network CNN is used to extract features from the preprocessed data, and the output is processed by the softmax function. Convert to probability vector ; in, is the output vector of the fully connected layer in the convolutional neural network CNN. elements, probability vector Indicates the probability that the input data belongs to each tissue type to identify and extract the impedance characteristics of different tissues. Compare and analyze the extracted impedance characteristics of different tissues with the classified and labeled data stored in the cloud database to identify the characteristics of the current tissue, including fat distribution, muscle density, blood concentration, and epidermal thickness; Let time point be the independent variable , the impedance value is the dependent variable , the slope is , the intercept is , is a random error term, representing the influence of other unconsidered factors on the impedance value except the time factor, and finally the analysis result is obtained; Slope It is expressed as the change in impedance value per unit time in the impedance characteristic, and its calculation formula is: ； in, is the sample size, and They are Sample and value, and Represented as time points and impedance value The average value of The formula can quantify the rate of change of impedance value over time; In radiofrequency treatment, if the slope , it means the impedance value increases, and the RF energy is increased; If the slope , it means the impedance value is reduced, and the RF energy is reduced; intercept It represents the impedance value when the time point is 0, and its calculation formula is: ； Among them, the intercept Used to determine the impedance level in the initial state; The linear regression model is used to calculate the trend line of impedance value changing over time, and its formula is as follows: ； Among them, the impedance value The target variable used to predict or analyze its changing pattern, the time point is , used to convert the impedance value With time point Establish a relationship to observe the changing trend of impedance value.

9. The method of the AI ​​radio frequency detection verification and identification system according to claim 8 is characterized in that: In step S3, first, a judgment matrix is ​​constructed by using the hierarchical analysis method. : ； in, Indication factors Relative to factors The degree of importance can be determined through expert verification; And calculate the weight vector according to the following steps : First, calculate the judgment matrix The product of each row element : ； Among them, the element product Used to combine multiple relative importance evaluation values ​​into an intermediate calculation quantity; Next, calculate of Second root: ； in, For element-wise multiplication The initial operation of normalizing the product of , so that the importance values ​​of different factors are at the same level; Finally, Normalize to get the weight vector ; in, ; According to the final weight vector Determining the relative importance weights of multiple factor information in determining RF output parameters; Comprehensive calculation of RF output parameter indicators The formula is as follows: ； Among them, the indicator It is used to adaptively determine the optimal radiofrequency treatment parameters based on multiple individual factors, reducing the risk of energy loss and burns caused by inadequate consideration of a single factor.

10. The method of the AI ​​radio frequency detection verification and identification system according to claim 9 is characterized in that: In step S7, the clustering effectiveness evaluation formula is the sum of squared errors It is expressed as follows: ； in, is the number of clusters, Indicates clusters, Is a cluster The data points in the equation are each specific impedance value data sample. It is The center of the cluster is All data points in The mean, sum of squared errors Used to evaluate the effectiveness of clustering and calculate each data point With the cluster center The smaller the value, the smaller the data points in the cluster. The more closely clustered they are in the center of the cluster, The better the clustering effect is, that is, similar impedance value change pattern data are more accurately classified into one category to achieve automated data analysis and clustering based on individual impedance databases; Next, the model is trained in multiple rounds according to the training set data, the training results are monitored on the validation set, and tested on the test set to obtain the optimized model, which is then sent to the terminal AI processing unit to improve the recognition and prediction capabilities of the model. The cross entropy loss function is used in the training process. : ； in, is the number of samples, the size of the dataset used to train the convolutional neural network, is the number of categories, indicating the total number of different tissue types, and the output of the convolutional neural network CNN corresponds to the category , in order to make classification predictions on the input data, Indicates The samples belong to The true label of the class, Indicates The sample is predicted to be Class probability; cross entropy loss function Used in the training process to predict the results With the true label Perform a comparison to measure the difference between two probability distributions.

Images