Microwave and radio frequency integrated treatment system

Abstract

The invention discloses a microwave and radio frequency integrated treatment system which comprises a multi-source information acquisition module, a feature extraction module, a feature modeling module and an instruction generation module. The multi-source information acquisition module acquires internal temperature distribution data and contact impedance spectroscopy data of a focus through a sensor array, and forms focus materialization data in combination with focus space information; the feature extraction module performs space-time registration on the focus materialization data by using a space-time registration algorithm to obtain focus comprehensive thermal impedance features; the feature modeling module extracts space-time evolution features of the comprehensive thermal impedance features of the focus by using a lightweight convolutional neural network, and constructs a thermal impedance dynamic expansion model; the instruction generation module generates an initial control instruction sequence used for controlling the microwave and radio frequency integrated treatment equipment based on the thermal impedance dynamic expansion model in combination with the treatment scheme.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical devices, and in particular relates to a microwave radiofrequency integrated treatment system. Background Art

[0002] With the advancement of tumor hyperthermia and energy therapy technologies in the medical device field, microwave radiofrequency combined therapy, a minimally invasive treatment modality that achieves precise thermal damage to lesions through electromagnetic energy conversion, has gradually become a research hotspot in clinical treatment due to its non-invasiveness, controllability, and minimal damage to surrounding tissues. Currently, microwave radiofrequency therapy primarily achieves thermal effects on lesions by controlling energy output parameters such as power and duration. However, in practical applications, traditional treatment systems face key challenges, such as insufficient multimodal data fusion accuracy and weak dynamic control capabilities during the treatment process. Traditional technologies primarily rely on empirical parameter settings or open-loop control based on a single physical model to control thermal effects on lesion tissue. For example, single-point thermocouple temperature measurement combined with fixed power output makes it difficult to obtain the spatial distribution characteristics of the internal temperature field and impedance spectrum of the lesion in real time. Alternatively, simple feedback control algorithms are used to adjust energy output, lacking multi-objective optimization for temperature uniformity, impedance stability, and power fluctuations. Furthermore, traditional systems often overlook the correlation between in vitro experimental data and the physiological characteristics of living individuals when generating treatment plans. They also lack a hierarchical response and dynamic monitoring mechanism for potential safety risks such as temperature anomalies and excessive energy accumulation during treatment, making it difficult to balance treatment accuracy and safety. To address these issues, existing technologies have attempted to acquire multi-source data through sensor arrays, but have yet to develop a fusion model for the spatiotemporal characteristics of temperature distribution and impedance spectra. Control strategies lack dynamic model construction based on lightweight neural networks and reinforcement learning-driven instruction optimization, making it difficult to adapt to clinical treatment needs with significant individual differences. Summary of the Invention

[0003] Based on this, it is necessary to provide a microwave radiofrequency integrated treatment system that can solve the above problems.

[0004] In a first aspect, the present application provides a microwave radiofrequency integrated treatment system, comprising:

[0005] A multi-source information acquisition module is used to obtain the internal temperature distribution data and contact impedance spectrum data of the lesion through a sensor array, and to form the lesion materialization data by combining the lesion spatial information;

[0006] The feature extraction module is used to perform spatiotemporal registration of the lesion physical and chemical data using a spatiotemporal registration algorithm to obtain the comprehensive thermal impedance characteristics of the lesion;

[0007] The feature modeling module is used to extract the spatiotemporal evolution characteristics of the comprehensive thermal impedance characteristics of the lesion using a lightweight convolutional neural network and to construct a dynamic expansion model of thermal impedance;

[0008] The instruction generation module is used to generate an initial control instruction sequence based on the dynamic expansion model of thermal impedance and combined with the treatment plan; the initial control instruction sequence is used to control the microwave radio frequency integrated treatment equipment to achieve corresponding functions.

[0009] In one embodiment, the system further includes an instruction optimization module for:

[0010] Combining the dynamic expansion model of thermal impedance with the initial control instruction sequence, a reinforcement learning model is used to generate predictions of treatment process changes;

[0011] A multi-objective optimization model is constructed based on the deviation between the predicted changes in the treatment process and the real-time lesion physical and chemical data;

[0012] The multi-objective optimization model is used to dynamically adjust the initial control instruction sequence to obtain the optimized control instruction sequence.

[0013] In one embodiment, the instruction optimization module is further configured to:

[0014] Comparing the lesion's physical and chemical data with a preset matching threshold to determine the lesion's thermal impedance state;

[0015] Generate dual-frequency tuning parameters of RF or microwave based on thermal impedance state, and use the tuning parameters as dynamic constraints of the multi-objective optimization model;

[0016] Based on the multi-objective optimization model, the dynamic constraints are iteratively updated through the rolling horizon control algorithm, and the initial control instruction sequence is adjusted synchronously to obtain the optimized control instruction sequence.

[0017] In one embodiment, the instruction optimization module is further configured to construct a multi-objective optimization model using the following formula:

[0018]

[0019] in, represents the spatial variance of the temperature field in the k-th step of the prediction time domain, represents the square sum of the impedance spatial gradient, α represents the temperature uniformity weight coefficient, γ represents the impedance stability weight coefficient, μ||Δu k || ∞ =max||P k+1 -P k || represents the power fluctuation penalty term, μ represents the power fluctuation penalty coefficient, represents the constraints, A and B represent the state transfer matrix, E x represents the maximum allowed cumulative energy, x k Represents the state variable, u k represents the control variable.

[0020] In one embodiment, the system further includes an intelligent solution module for:

[0021] Associating in vitro experimental data with in vivo lesion characteristics through a preset knowledge graph;

[0022] Using a machine learning model, combined with knowledge graph matching results, historical treatment data, and a dynamic expansion model of thermal impedance, a candidate treatment parameter set containing radiofrequency or microwave power and action time is generated.

[0023] Based on the candidate treatment parameter set, the optimal parameter combination is determined using a multidimensional trade-off optimization model;

[0024] Generate dynamic treatment plans based on the optimal parameter combination.

[0025] In one embodiment, the system further includes a multi-level monitoring module for:

[0026] Generate dynamic reference temperature distribution based on knowledge graph matching results;

[0027] Using the thermal impedance dynamic expansion model, the spatial gradient distribution of the predicted temperature field is generated;

[0028] Based on the temperature field and impedance gradient data in the lesion physical and chemical data, the difference between the reference temperature distribution and the predicted gradient distribution is calculated;

[0029] Compare the difference value with the historical security boundary threshold in the knowledge graph to generate anomaly judgment results;

[0030] According to the abnormal judgment results, the corresponding control instructions are triggered using the hierarchical response control mechanism.

[0031] In one embodiment, the hierarchical response control mechanism includes:

[0032] Based on the deviation between the difference value in the abnormal judgment result and the safety boundary threshold, perform the following steps:

[0033] When the deviation is less than 15%, a control instruction is generated to attenuate the RF or microwave output power according to a preset ratio;

[0034] When the deviation reaches 15%-30%, a control instruction and an alarm instruction for pausing the treatment equipment are generated;

[0035] When the deviation exceeds 30%, an emergency stop command is generated and an abnormality log is created.

[0036] In a second aspect, the present application also provides a microwave radiofrequency integrated treatment method, comprising:

[0037] The sensor array is used to obtain the internal temperature distribution data and contact impedance spectrum data of the lesion, and the lesion spatial information is combined to form the lesion physical and chemical data;

[0038] The spatial and temporal registration algorithm is used to perform spatial and temporal registration on the physical and chemical data of the lesion to obtain the comprehensive thermal impedance characteristics of the lesion;

[0039] A lightweight convolutional neural network is used to extract the spatiotemporal evolution characteristics of the lesion's comprehensive thermal impedance features and to construct a dynamic expansion model of thermal impedance.

[0040] Based on the dynamic expansion model of thermal impedance, an initial control instruction sequence is generated in combination with the treatment plan; the initial control instruction sequence is used to control the microwave radio frequency integrated treatment equipment to achieve corresponding functions.

[0041] In a third aspect, the present application also provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned microwave radiofrequency integrated treatment system function when executing the computer program.

[0042] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of implementing the functions of a microwave radiofrequency integrated treatment system are implemented.

[0043] The aforementioned integrated microwave radiofrequency treatment system utilizes a multi-source information acquisition module, using a sensor array to collect internal lesion temperature distribution, contact impedance spectrum, and lesion spatial information to generate multidimensional physical data. This overcomes the limitations of traditional single-point detection and provides comprehensive input for precise modeling. The feature extraction module uses a spatiotemporal registration algorithm to perform spatiotemporal registration of multimodal data, resolving spatiotemporal misalignment issues and forming a comprehensive thermal impedance signature that integrates spatiotemporal features, thereby improving the accuracy of characterizing the lesion's thermal-electrical coupling characteristics. The feature modeling module utilizes a lightweight convolutional neural network to automatically learn spatiotemporal evolution features and construct a dynamic expansion model, overcoming the shortcomings of traditional fixed physical models or empirical formulas. This lightweight design meets real-time requirements and adapts to individual tissue differences. The instruction generation module generates an initial control instruction sequence based on a dynamic model combined with the treatment plan, achieving a transition from empirical open-loop control to model-driven closed-loop control. This allows energy output parameters to be dynamically adjusted based on the real-time status of the lesion. This systematically addresses the problems of insufficient multimodal data fusion accuracy, spatiotemporal feature fragmentation, poor model adaptability, and weak dynamic control capabilities in traditional technologies, providing key technical support for the precise and personalized development of microwave radiofrequency treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is one of the structural diagrams of a microwave radiofrequency integrated treatment system of the present invention;

[0046] Figure 2 This is the second structural diagram of a microwave radiofrequency integrated treatment system of the present invention;

[0047] Figure 3 This is a flow chart of a microwave and radio frequency integrated treatment method of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0049] The present invention provides an integrated microwave and radiofrequency treatment system, whose implementation environment includes a sensor array terminal, a computing terminal, microwave and radiofrequency treatment equipment, and a data storage system. The sensor array collects lesion temperature distribution, contact impedance spectrum, and tissue spatial information in real time, transmitting this information to the computing terminal for spatiotemporal alignment, feature modeling, and instruction generation. The processing device combines knowledge graphs, historical data, and dynamic models to generate initial control instructions, dynamically adjusting the output energy of the treatment equipment based on multi-objective constraints. Wired and wireless communication between the hardware implements a closed loop of data acquisition, modeling, and control, meeting the requirements of multimodal data fusion, personalized solution generation, and real-time safety monitoring in scenarios such as tumor hyperthermia.

[0050] In one embodiment, Figure 1 As shown, a microwave radiofrequency integrated treatment system is provided. This embodiment uses the system deployed on a computing terminal as an example for illustration. It is understandable that the system can also be deployed on a server, and can also be applied to a collaborative architecture including a computing terminal and a cloud server, and implemented through the interaction between the computing terminal and the cloud server. In this embodiment, the system includes the following modules:

[0051] The multi-source information acquisition module 101 is used to acquire the temperature distribution data and contact impedance spectrum data inside the lesion through the sensor array, and form the lesion physical and chemical data in combination with the lesion spatial information.

[0052] Among them, the sensor array includes temperature sensors and impedance sensors to realize real-time collection of spatial distribution data inside the lesion. The temperature sensor can be a miniature thermocouple or infrared sensor, covering different depths and positions of the lesion, and obtaining temperature distribution data such as the temperature value at each point and the temperature field gradient; the impedance sensor measures the electrical impedance characteristics of the tissue by applying a weak AC signal, and generates contact impedance spectrum data such as the impedance amplitude and phase change curve with frequency. At the same time, it can be combined with medical imaging such as MRI, CT or ultrasound positioning technology to obtain lesion spatial information, including the lesion's three-dimensional coordinates, geometric shape, boundary contour and relative position with surrounding tissue. Through the fusion of sensor array and spatial information, a multidimensional data set containing temperature distribution, impedance spectrum and anatomical positioning is constructed, laying a data foundation for subsequent spatiotemporal alignment, feature modeling and precise control.

[0053] The feature extraction module 102 is used to perform spatiotemporal registration on the lesion physical and chemical data using a spatiotemporal registration algorithm to obtain a comprehensive thermal impedance feature of the lesion.

[0054] Among them, different sensors, such as temperature sensors and impedance sensors, have different sampling frequencies and trigger times, resulting in a mismatch in the timestamps of temperature and impedance data at the same physical location. For example, temperature is collected every 0.1 seconds, and impedance is collected every 0.5 seconds. Physical deployment errors of the sensor array, tissue deformation, such as lesion displacement caused by respiratory movement, or spatial coordinate conversion errors, such as mapping deviations from the sensor's local coordinate system to the global coordinate system of medical imaging, lead to inaccurate spatial positioning of data points. By utilizing spatiotemporal registration algorithms, discrete, asynchronous multimodal data can be converted into structured data that is consistent in both time and space: time synchronization can use interpolation such as linear interpolation, spline interpolation, or time window alignment technology to unify time series data of different frequencies into the same time grid, such as a time series with an interval of 100ms, to ensure a one-to-one correspondence between temperature, impedance, and spatial information at the same moment. Spatial registration is based on the three-dimensional coordinate system constructed by medical images, such as CT / MRI. Through coordinate transformations such as rigid body transformation and elastic registration, the local spatial coordinates collected by the sensor are mapped to the global coordinate system, eliminating the effects of tissue deformation or positioning errors, ensuring that each data point accurately corresponds to the specific anatomical location within the lesion. After spatiotemporal registration, the temperature and impedance characteristics of each spatiotemporally consistent data point can be combined into a thermal impedance feature, which is then converted into a spatiotemporally coupled feature matrix to obtain the comprehensive thermal impedance feature of the lesion.

[0055] The feature modeling module 103 is used to extract the spatiotemporal evolution characteristics of the comprehensive thermal impedance characteristics of the lesion using a lightweight convolutional neural network and to construct a dynamic expansion model of the thermal impedance.

[0056] The lightweight convolutional neural network's hierarchical feature extraction mechanism captures multi-scale information. Bottom-level convolutional layers extract local thermal impedance features, such as the instantaneous correlation between single-point temperature and impedance. Higher-level convolutional layers integrate global features, such as the spatiotemporal correlation between temperature uniformity and impedance stability across the entire lesion. The shallow network structure can employ a shallow architecture, such as 2-3 convolutional blocks, to reduce parameter count, using small 3×3 convolutional kernels and avoiding complex residual connections to ensure real-time computational efficiency. In the spatial dimension, two-dimensional convolutional layers are used to extract local features, such as temperature gradients and impedance uniformity between adjacent points. The convolutional kernel weight sharing mechanism naturally adapts to spatial translation invariance, such as the similarity of thermal-electric response patterns at different locations. In the temporal dimension, the time series can be treated as a stack of channels or through sequential convolution. For example, feature maps from consecutive K frames are used as input, and 3D convolutional kernels are used to capture dynamic changes in the spatiotemporal cube. Alternatively, temporal pooling, such as mean / maximum pooling in the temporal dimension, can be applied after the 2D convolutional neural network to extract statistical features across time, such as the trend of temperature variance over time. At the same time, model compression techniques such as pruning, quantization, or efficient network architectures, such as the deep separable convolution of MobileNet (a lightweight convolutional neural network), are used to reduce computational complexity while maintaining feature extraction capabilities, adapting to the embedded computing environment of medical equipment. The convolutional layer calculates the weighted sum of thermal impedance features within the local receptive field through a sliding window, capturing the coupling relationship between adjacent points in space. For example, the local variance of the temperature distribution can be indirectly represented by the convolution operation of the temperature values of adjacent points; the spatial gradient of the impedance spectrum can be extracted through the gradient operator, and the lightweight convolutional neural network automatically learns more complex features in a data-driven manner. The dynamic expansion model of thermal impedance, constructed based on the spatiotemporal evolution characteristics of the comprehensive thermal impedance characteristics of the lesion, can use historical treatment data for supervised learning, and the loss function design takes into account both temperature prediction error and impedance prediction error. The deep modeling of the dynamic characteristics of the lesion provides core technical support for the precision and individualization of microwave radiofrequency treatment.

[0057] The instruction generation module 104 is used to generate an initial control instruction sequence based on the thermal impedance dynamic expansion model and in combination with the treatment plan; the initial control instruction sequence is used to control the microwave radiofrequency integrated treatment device to achieve corresponding functions.

[0058] The dynamic expansion model of thermal impedance can predict, based on the current lesion status and input control commands, how the temperature of various regions within the lesion will rise and how tissue impedance will change under different microwave powers and radiofrequency frequencies. Treatment plans are formulated based on the patient's specific condition, lesion characteristics such as size, location, and type, and treatment goals such as complete ablation of the lesion and avoiding damage to surrounding normal tissue. If the treatment goal is to raise the lesion temperature to a certain range within a specified timeframe to achieve ablation, the dynamic expansion model of thermal impedance is used to identify a combination of control commands that can achieve this temperature target. For example, the appropriate microwave power and radiofrequency frequency, as well as the application duration, are determined, generating a series of initial control commands. These initial control commands typically include various parameter settings for the microwave and radiofrequency devices, such as power, frequency, and application duration. These initial control commands are not limited to single parameter settings but can also include time-varying sequences. For example, at different stages of treatment, the microwave and radiofrequency power and frequency are adjusted based on changes in lesion status predicted by the dynamic expansion model of thermal impedance. A higher power may be required to rapidly increase the temperature at the beginning of treatment, while the power may be appropriately reduced as the target temperature is approached to maintain a stable therapeutic effect.

[0059] The above-mentioned microwave radiofrequency integrated treatment system uses a sensor array to collect the internal temperature distribution, contact impedance spectrum and spatial information of the lesion through a multi-source information acquisition module to form multi-dimensional physical data, breaking through the limitations of traditional single-point detection and providing comprehensive input for accurate modeling; the feature extraction module uses a spatiotemporal registration algorithm to align the multimodal data in time and space, solve the problem of spatiotemporal misalignment of data, form a comprehensive thermal impedance feature that integrates spatiotemporal features, and improve the characterization accuracy of the thermal-electrical coupling characteristics of the lesion; the feature modeling module uses a lightweight convolutional neural network to automatically learn the spatiotemporal evolution features and construct a dynamic expansion model, overcoming the traditional fixed physical model or experience. The shortcomings of the formula are overcome by using lightweight design to meet real-time requirements and adapt to individual tissue differences; the instruction generation module generates the initial control instruction sequence based on the dynamic model combined with the treatment plan, realizing the transformation from empirical open-loop control to model-driven closed-loop control, so that the energy output parameters can be dynamically adjusted according to the real-time status of the lesion, and systematically solving the problems of insufficient multimodal data fusion accuracy, spatiotemporal feature fragmentation, poor model adaptability and weak dynamic regulation ability in traditional technologies, providing key technical support for the precision and individualization of microwave radiofrequency treatment, effectively improving the control accuracy and safety of the thermal effect of the lesion during treatment, and reducing damage to surrounding tissues.

[0060] In one embodiment, Figure 2 As shown, the system further includes an instruction optimization module 105, which is used to:

[0061] Combining the dynamic expansion model of thermal impedance with the initial control instruction sequence, a reinforcement learning model is used to generate predictions of treatment process changes;

[0062] A multi-objective optimization model is constructed based on the deviation between the predicted changes in the treatment process and the real-time lesion physical and chemical data;

[0063] The multi-objective optimization model is used to dynamically adjust the initial control instruction sequence to obtain the optimized control instruction sequence.

[0064] Specifically, the reinforcement learning model learns the optimal policy through interaction between the agent and the environment, aiming to maximize cumulative rewards. It takes as input a dynamic expansion model of thermal impedance and an initial sequence of control commands. It simulates each time step during the treatment process, considering the impact of different commands on the lesion's thermal impedance state and how these state changes contribute to the treatment goal. For example, at a certain time point, the microwave power is increased, and the model-predicted temperature distribution and impedance changes in the lesion are observed to evaluate the contribution of this operation to the treatment outcome. Through continuous simulation and learning, the model predicts the trend of the lesion's thermal impedance state throughout the treatment process, including changes in temperature and impedance, based on different control commands. This generates a prediction of the treatment process. The predicted treatment process change is compared with the lesion's physical and chemical data, and the deviation between the two is calculated, reflecting the difference between the prediction and the actual situation. This difference may be due to model inaccuracies, individual patient differences, or external environmental factors. Multi-objective optimization refers to the process of finding the optimal solution among multiple conflicting objectives. In treatment scenarios, these objectives typically include maximizing the ablation effect of the lesion, minimizing damage to surrounding normal tissue, and ensuring the safety and efficiency of the treatment process. Based on the aforementioned deviations, a multi-objective optimization model is constructed, integrating these objectives and deviations into a unified framework. For example, the model considers how to adjust control instructions to minimize the deviation between prediction and actual results while meeting multiple treatment objectives. The treatment process is dynamic, and the initial control instruction sequence may not fully adapt to actual developments. By monitoring and analyzing deviations in real time and utilizing a multi-objective optimization model to dynamically adjust the initial control instruction sequence, treatment plans can be made more flexible and precise. For example, if the actual temperature rise is slower than predicted, the model may recommend increasing microwave power or extending heating time. Based on its internal algorithms and rules, the multi-objective optimization model adjusts and optimizes the initial control instruction sequence while meeting multiple treatment objectives. The resulting optimized control instruction sequence better adapts to the actual conditions during treatment, improving treatment efficacy and mitigating potential risks. This optimization process is iterative, and control instructions are continuously optimized as treatment progresses and new data is acquired. These three steps form a closed-loop feedback control system. Through continuous prediction, comparison, and optimization, precise control and dynamic adjustment of the treatment process are achieved, helping to improve the efficacy and safety of integrated microwave and radiofrequency therapy.

[0065] In one embodiment, the instruction optimization module 105 is further configured to:

[0066] Comparing the lesion's physical and chemical data with a preset matching threshold to determine the lesion's thermal impedance state;

[0067] Generate dual-frequency tuning parameters of RF or microwave based on thermal impedance state, and use the tuning parameters as dynamic constraints of the multi-objective optimization model;

[0068] Based on the multi-objective optimization model, the dynamic constraints are iteratively updated through the rolling horizon control algorithm, and the initial control instruction sequence is adjusted synchronously to obtain the optimized control instruction sequence.

[0069] For example, comparing the lesion's physical and chemical data with a preset matching threshold can accurately determine the lesion's current thermal impedance state. For example, if the detected tissue impedance value is below a preset lower limit, it indicates that the lesion may be overheating or carbonizing; if the temperature distribution exceeds the preset normal range, it indicates an abnormal thermal state. This comparison allows real-time understanding of the lesion's actual state during treatment, providing a basis for subsequent parameter adjustments and instruction optimization. After determining the lesion's thermal impedance state, corresponding radiofrequency or microwave tuning parameters are generated based on specific rules and algorithms. These tuning parameters are used to adjust the operating state of the microwave-radiofrequency integrated treatment device, such as the frequency and power of the radiofrequency or microwave, to ensure the effectiveness and safety of the treatment. The generated radiofrequency or microwave tuning parameters are incorporated into the multi-objective optimization model as dynamic constraints. During the optimization of control instructions, the execution results of the instructions are guaranteed to meet the radiofrequency or microwave tuning requirements under the current thermal impedance state. For example, if the lesion's thermal impedance indicates an excessively high temperature, the generated radiofrequency or microwave tuning parameters may require a reduction in output power. The multi-objective optimization model takes this constraint into account during optimization. The rolling horizon control algorithm is a control strategy that optimizes control instructions within each control cycle based on the current system state and predicted information within the time domain over a period of time. In this system, the algorithm, combined with a multi-objective optimization model, iteratively updates dynamic constraints. Within each control cycle, the thermal impedance state of the target tissue is reassessed, new RF or microwave tuning parameters are generated, and the dynamic constraints are updated. Simultaneously, the initial control instruction sequence is synchronously adjusted based on the updated dynamic constraints and the multi-objective optimization model. After multiple iterations, the algorithm converges to an optimal solution, resulting in an optimized control instruction sequence. This optimized control instruction sequence dynamically adjusts the operating parameters of the treatment device based on the real-time thermal impedance state of the target lesion, ensuring that the treatment process is always optimal and improving the accuracy and effectiveness of the treatment. By comparing, generating parameters, iteratively updating, and adjusting instructions, the treatment plan can be adjusted based on the real-time state of the target lesion.

[0070] In one embodiment, the instruction optimization module 105 is further configured to construct a multi-objective optimization model using the following formula:

[0071]

[0072]

[0073] in, represents the spatial variance of the temperature field in the k-th step of the prediction time domain, represents the square sum of the impedance spatial gradient, α represents the temperature uniformity weight coefficient, γ represents the impedance stability weight coefficient, μ||Δu k || ∞ =max||P k+1 -P k || represents the power fluctuation penalty term, μ represents the power fluctuation penalty coefficient, represents the constraints, A and B represent the state transfer matrix, E x represents the maximum allowed cumulative energy, x k Represents the state variable, u k represents the control variable.

[0074] Specifically, the spatial variance of the temperature field in the kth step of the prediction time domain is At each time step k in the prediction time domain, the temperature values of each point in the entire lesion area are statistically analyzed and their variance is calculated. This reflects the spatial uniformity of the temperature field. A smaller spatial variance means that the temperature is more evenly distributed within the lesion. For example, in tumor ablation treatment, if the temperature distribution is uneven, some tumor tissues may not be fully heated, affecting the treatment effect. Therefore, by introducing this indicator into the objective function, the spatial variance of the temperature field can be reduced during the optimization process. Impedance spatial gradient sum of squares Calculated based on contact impedance spectrum data. The spatial gradient of the impedance value at each point in the lesion is calculated, and then the squares of the gradients at each point are summed. It is used to measure the severity of the spatial change of impedance. A stable impedance distribution usually indicates that the physical state of the tissue is relatively uniform, while drastic changes in the impedance spatial gradient may indicate abnormalities in the tissue, such as tumor boundaries or areas of tissue damage. During the treatment process, it is hoped that the impedance distribution is relatively stable, and the stability of the tissue state during treatment is ensured by minimizing this indicator. The temperature uniformity weight coefficient α and the impedance stability weight coefficient γ are determined by the doctor based on experience or reference to the medical database, according to the specific treatment needs of the actual situation, to meet different temperature uniformity and impedance stability. Power fluctuation penalty term μ||Δu k || ∞ =max||P k+1 -P k||Calculated based on the power value set in the control command. Avoid drastic power fluctuations during treatment. Large power fluctuations may cause adverse effects on the treatment device and patient tissue, such as unstable operation of the device or additional damage to surrounding normal tissue. This penalty term is introduced to encourage the optimized control command to make power changes as smooth as possible. Constraints In, x k+1 =Ax k +Bu k This is the basic form of the state-space model, where the state transition matrices A and B are determined by the dynamic expansion model of thermal impedance constructed in the feature modeling module. It reflects how system states (such as temperature and impedance) change over time, as well as control inputs (such as power and frequency). It ensures that the execution results of control instructions conform to the state transition equations, thus ensuring the model's physical rationality and predictive accuracy. The safety upper limit is determined based on factors such as the patient's physical condition, lesion location, and treatment plan. It is used to limit the total energy input during the entire treatment process to avoid unnecessary damage to the patient due to excessive energy. k ∈[δ min ,δ max ] in the control variable u k The constraint range is [δ min ,δ max ] is used to avoid abnormal inputs that exceed a reasonable range. A multi-objective optimization model constructed using this formula comprehensively considers indicators such as temperature field spatial uniformity, impedance spatial stability, and power fluctuation. Combined with constraints such as system state transition patterns and maximum allowable cumulative energy, the model optimizes control variables to achieve precise regulation of the microwave radiofrequency treatment process, thereby improving treatment efficacy and safety.

[0075] In one embodiment, the system further includes an intelligent solution module 106 for:

[0076] Matching in vitro experimental data with in vivo lesion characteristics through a preset knowledge graph to obtain matching results;

[0077] Using a machine learning model, combined with matching results, historical treatment data, and a dynamic expansion model of thermal impedance, a candidate treatment parameter set including radiofrequency or microwave power and action time is generated;

[0078] Based on the candidate treatment parameter set, the optimal parameter combination is determined using a multidimensional trade-off optimization model;

[0079] Generate dynamic treatment plans based on the optimal parameter combination.

[0080] For example, in vitro experimental data includes thermal response parameters of different tissue types, such as tumors and normal tissues, under microwave / RF energy, such as thermal conductivity, impedance change threshold, critical temperature for thermal damage, and energy-time-temperature curves. In vivo lesion characteristics include anatomical features acquired through medical imaging such as CT / MRI, such as lesion 3D coordinates, volume, adjacent tissue location, and physiological indicators such as blood supply, tissue water content, and pH. Nodes and relationships are constructed using a graph database. Nodes can include in vitro experimental parameters such as ablation of a liver tumor at 45°C for 100 seconds, and in vivo characteristics such as a right liver tumor diameter of 3 cm and proximity to the portal vein. Edges can define association rules, such as increasing the initial power by 10% when the tumor diameter exceeds 2 cm, to achieve migration and adaptation of in vitro data to in vivo scenarios. Based on the knowledge graph matching results, the patient's lesion's anatomical features, such as location and size, are matched to the closest in vitro experimental case and corresponding basic parameters, such as the recommended power range and minimum application time. Historical treatment data includes successful treatment plans for similar lesions, such as 50W power, 120 seconds, and a 95% ablation rate, as well as failed cases, such as excessive power causing burns to adjacent tissue. A dynamic thermal impedance expansion model predicts the lesion's temperature field and impedance changes under different power-time combinations. For example, at 60W power for 60 seconds, the lesion's center temperature can reach 45°C and its edge temperature 42°C. Ensemble learning models, such as random forests or gradient boosting trees, can be used to target treatment efficacy (ablation rate) and safety (normal tissue temperature <45°C). The model outputs multiple candidate parameter sets, including RF or microwave output power (e.g., 20-40W) and application time (e.g., 60-300 seconds). The optimization objectives of a multidimensional trade-off optimization model can be: treatment efficacy, ensuring the core lesion temperature reaches the ablation threshold (e.g., above 42°C) and temperature uniformity (e.g., spatial variance <5°C²); safety, such as limiting normal tissue temperature to <45°C and controlling impedance gradients to avoid sudden impedance increases caused by tissue carbonization; energy efficiency, namely minimizing total energy output (power × time), to reduce treatment duration and device wear; and considering device physical limitations (e.g., maximum power of 100W and minimum time step of 1 second) and patient physiological limitations (e.g., avoiding high-frequency power fluctuations for lesions near the heart). A multi-objective genetic algorithm or Pareto optimization can be used to search for the optimal solution within a set of candidate parameters and generate a dynamic treatment plan. In a dynamic treatment plan, the treatment process can be divided into an initialization phase (e.g., rapid heating), a maintenance phase (e.g., stable ablation), and a cooling phase (e.g., safe exit), each corresponding to different power-time parameters. The treatment plan also includes a parameter adjustment interface. If the lesion edge temperature does not reach 42°C after 50 seconds of treatment, the power is automatically increased by 5W. Through cross-scenario association of knowledge graphs, personalized modeling of machine learning, and intelligent decision-making of multi-objective optimization, the transition from extensive empirical treatment to precise model-driven treatment has been achieved, providing core technical support for the individualization and safety of microwave radiofrequency treatment.

[0081] In one embodiment, the system further includes a multi-level monitoring module 107 for:

[0082] Generate dynamic reference temperature distribution based on knowledge graph matching results;

[0083] Using the thermal impedance dynamic expansion model, the spatial gradient distribution of the predicted temperature field is generated;

[0084] Based on the temperature field and impedance gradient data in the lesion physical and chemical data, the difference between the reference temperature distribution and the predicted gradient distribution is calculated;

[0085] Compare the difference value with the historical security boundary threshold in the knowledge graph to generate anomaly judgment results;

[0086] According to the abnormal judgment results, the corresponding control instructions are triggered using the hierarchical response control mechanism.

[0087] Specifically, the knowledge graph is used to match the patient's lesion with specific characteristics, such as anatomical features and physiological indicators, to identify the most similar cases. Based on the matching results, a dynamic reference temperature distribution tailored to the current treatment process is generated. This distribution dynamically adjusts as treatment progresses, corresponding to the temperature requirements of different treatment phases, such as warming, maintenance, and cooling. Using a dynamic extended thermal impedance model and combined with current treatment parameters, the spatial gradient distribution of the lesion's temperature field within the entire treatment area is predicted. The actual monitored temperature field and impedance gradient data are compared with the predicted spatial gradient distribution of the temperature field, and the difference between them is calculated to quantify the deviation between the actual treatment outcome and the expected outcome. Based on the safety margin thresholds of a large number of historical treatment cases in the knowledge graph—i.e., the reasonable range of difference values under different circumstances—the calculated difference value is compared with the historical safety margin threshold. If the difference value exceeds the threshold range, the treatment process is considered abnormal; otherwise, the treatment process is considered normal. The abnormality judgment result can be simply normal or abnormal, or further subdivided into different levels of abnormality, such as mild abnormality, moderate abnormality, and severe abnormality. Based on the severity of the abnormality judgment result, the abnormal situation is divided into different levels, and each level corresponds to different control instructions. When an abnormality is determined during the treatment process, the corresponding control instruction is triggered according to the level of the abnormality. For example, if the abnormality is mild, a warning message may be issued to alert the operator; if the abnormality is moderate, the treatment parameters may be automatically adjusted, such as reducing the output power of the radio frequency or microwave and shortening the action time; if the abnormality is severe, the treatment may be stopped immediately to ensure the safety of the patient.

[0088] In one embodiment, the hierarchical response control mechanism includes:

[0089] Based on the deviation between the difference value in the abnormal judgment result and the safety boundary threshold, perform the following steps:

[0090] When the deviation is less than 15%, a control instruction is generated to attenuate the RF or microwave output power according to a preset ratio;

[0091] When the deviation reaches 15%-30%, a control instruction and an alarm instruction for pausing the treatment equipment are generated;

[0092] When the deviation exceeds 30%, an emergency stop command is generated and an abnormality log is created.

[0093] For example, when the deviation is less than 15%, it indicates that while the treatment process has deviated from expectations, the deviation is minor and does not pose a serious threat to treatment safety and effectiveness. In this case, the system does not directly terminate treatment. Instead, it fine-tunes the treatment process by attenuating the RF or microwave output power by a preset ratio to restore the treatment process to a normal range. This avoids excessive intervention that may lead to poor treatment results, while also enabling timely response to the abnormality and preventing further escalation of the deviation. For example, during microwave ablation therapy for tumors, if the detected temperature distribution deviates from the reference value by less than 15%, appropriately reducing the RF or microwave output power can prevent local tissue overheating while still ensuring continued treatment. When the deviation is between 15% and 30%, indicating a severe abnormality and the possibility of continued treatment posing a risk to the patient, a control command to pause the treatment device is immediately generated, halting the current treatment operation. Simultaneously, an alarm is triggered to alert the operator and promptly address the abnormality. This prevents the abnormality from escalating further and gives the operator time to conduct inspections and adjustments. The alarm ensures that the operator is promptly notified of the abnormality so that appropriate measures can be taken, such as checking the device status and adjusting treatment parameters. When the deviation exceeds 30%, it indicates a serious anomaly in the treatment process, potentially posing a significant threat to the patient's safety and health. In this case, the system immediately generates an emergency stop command, forcibly halting the treatment equipment. Simultaneously, an exception log is generated, recording the time, specific circumstances, and relevant parameters of the anomaly. Emergency treatment halts maximize patient safety and prevent further harm. The anomaly log provides a crucial basis for subsequent analysis and improvement. Users can review the log to identify the cause of the anomaly, analyze lessons learned, and optimize treatment plans and equipment settings.

[0094] The above-mentioned microwave radiofrequency integrated treatment system collects lesion temperature distribution, contact impedance spectrum and spatial information through a multi-source information acquisition module to form multi-dimensional physical and chemical data, breaking through the limitations of traditional single-point detection and providing comprehensive input for accurate modeling; the feature extraction module uses a spatiotemporal registration algorithm to solve the spatiotemporal misalignment problem of multimodal data, forming a comprehensive thermal impedance feature that integrates spatiotemporal features, and improving the accuracy of characterization of the thermal-electrical coupling characteristics of the lesion; the feature modeling module uses a lightweight convolutional neural network to automatically learn the spatiotemporal evolution features and construct a dynamic expansion model, overcoming the shortcomings of traditional fixed physical models or empirical formulas, and meeting real-time requirements and adapting to individual tissue differences with a lightweight design; the instruction generation module generates an initial control instruction sequence based on a dynamic model combined with a treatment plan, realizing the transition from empirical open-loop control to model-driven closed-loop control, so that energy The output parameters can be dynamically adjusted according to the real-time status of the lesion; the instruction optimization module dynamically adjusts the control instructions through reinforcement learning models, multi-objective optimization models and rolling time domain control algorithms, combined with the thermal impedance dynamic expansion model and real-time data deviation, to take into account multi-objective optimization of temperature uniformity, impedance stability and power fluctuation; the intelligent solution module associates in vitro experiments with in vivo characteristics through knowledge graphs, and uses machine learning and multi-dimensional trade-off optimization to generate individualized candidate treatment parameters and optimal solutions, thereby realizing the dynamic and personalized treatment plan; the multi-level monitoring module generates reference temperature distribution and predicted gradient distribution based on knowledge graphs and dynamic models, combines the real-time data difference value with the safety boundary threshold, and triggers different control instructions through a hierarchical response control mechanism to perform hierarchical responses and dynamic monitoring of safety risks such as temperature anomalies and excessive energy accumulation. The synergistic effect of the above modules systematically solves the problems in traditional technologies such as insufficient accuracy of multimodal data fusion, fragmentation of spatiotemporal features, poor model adaptability, weak dynamic control capabilities, and insufficient response to safety risks. It effectively improves the control accuracy and safety of the thermal effects of lesions during treatment, reduces damage to surrounding tissues, and provides key technical support for the precision, individualization and safety of microwave radiofrequency treatment.

[0095] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0096] Based on the same inventive concept, the embodiments of the present application also provide a microwave radiofrequency integrated treatment method for implementing the aforementioned microwave radiofrequency integrated treatment system. The solution to the problem provided by this method is similar to the solution described in the aforementioned system. Therefore, the specific limitations of one or more microwave radiofrequency integrated treatment method embodiments provided below can be found in the above-mentioned limitations on the microwave radiofrequency integrated treatment system and will not be repeated here.

[0097] In an exemplary embodiment, Figure 3 As shown, a microwave radiofrequency integrated treatment method is provided, comprising:

[0098] S11, acquiring temperature distribution data and contact impedance spectrum data inside the lesion through a sensor array, and combining the lesion spatial information to form lesion physical and chemical data;

[0099] S12, using a spatiotemporal registration algorithm to perform spatiotemporal registration on the lesion physical and chemical data to obtain the comprehensive thermal impedance characteristics of the lesion;

[0100] S13, using a lightweight convolutional neural network to extract the spatiotemporal evolution characteristics of the lesion's comprehensive thermal impedance features and construct a dynamic expansion model of thermal impedance;

[0101] S14, based on the dynamic expansion model of thermal impedance, combined with the treatment plan, generates an initial control instruction sequence; the initial control instruction sequence is used to control the microwave radio frequency integrated treatment device to achieve corresponding functions.

[0102] In one embodiment, the method further comprises:

[0103] S21, combining the dynamic expansion model of thermal impedance with the initial control instruction sequence, and using the reinforcement learning model to generate predictions of treatment process changes;

[0104] S22, construct a multi-objective optimization model based on the deviation between the predicted treatment process changes and the real-time lesion physical and chemical data;

[0105] S23, dynamically adjusting the initial control instruction sequence using a multi-objective optimization model to obtain an optimized control instruction sequence.

[0106] In one embodiment, the method further comprises:

[0107] S31, comparing the lesion physical and chemical data with a preset matching threshold to determine the thermal impedance state of the lesion;

[0108] S32, generating dual-frequency tuning parameters of radio frequency or microwave according to the thermal impedance state, and using the tuning parameters as dynamic constraints of the multi-objective optimization model;

[0109] S33, based on the multi-objective optimization model, iteratively updates the dynamic constraints through the rolling horizon control algorithm, and synchronously adjusts the initial control instruction sequence to obtain the optimized control instruction sequence.

[0110] In one embodiment, the method further comprises:

[0111] S41, use the following formula to build a multi-objective optimization model:

[0112]

[0113] in, represents the spatial variance of the temperature field in the k-th step of the prediction time domain, represents the square sum of the impedance spatial gradient, α represents the temperature uniformity weight coefficient, γ represents the impedance stability weight coefficient, μ||Δu k || ∞ =max||P k+1 -P k || represents the power fluctuation penalty term, μ represents the power fluctuation penalty coefficient, represents the constraints, A and B represent the state transfer matrix, E x represents the maximum allowed cumulative energy, x k Represents the state variable, u k represents the control variable.

[0114] In one embodiment, the method further comprises:

[0115] S51, associate in vitro experimental data with in vivo lesion characteristics through a preset knowledge graph;

[0116] S52, using a machine learning model, combined with the matching results of the knowledge graph, historical treatment data, and the thermal impedance dynamic expansion model, to generate a candidate treatment parameter set including radiofrequency or microwave power and action time;

[0117] S53, based on the candidate treatment parameter set, the optimal parameter combination is determined using a multidimensional trade-off optimization model;

[0118] S54, generating a dynamic treatment plan based on the optimal parameter combination.

[0119] In one embodiment, the method further comprises:

[0120] S61, generating a dynamic reference temperature distribution based on the knowledge graph matching result;

[0121] S62, using the thermal impedance dynamic expansion model to generate the predicted spatial gradient distribution of the temperature field;

[0122] S63, based on the temperature field and impedance gradient data in the lesion physical and chemical data, calculating the difference between the reference temperature distribution and the predicted gradient distribution;

[0123] S64, comparing the difference value with the historical security boundary threshold in the knowledge graph to generate an abnormality judgment result;

[0124] S65: Based on the abnormality determination result, a hierarchical response control mechanism is used to trigger corresponding control instructions.

[0125] In one embodiment, the hierarchical response control mechanism includes:

[0126] S71, based on the deviation between the difference value in the abnormality judgment result and the safety boundary threshold, perform the following steps:

[0127] S72, when the deviation is less than 15%, generating a control instruction to attenuate the RF or microwave output power according to a preset ratio;

[0128] S73, when the deviation reaches 15%-30%, generating a control instruction and an alarm instruction for pausing the treatment device;

[0129] S74: When the deviation exceeds 30%, an emergency stop command is generated and an abnormality log is generated.

[0130] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the aforementioned microwave and radiofrequency integrated treatment system functions when executing the computer program.

[0131] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0132] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0133] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A microwave radiofrequency integrated treatment system, characterized in that: The system comprises: A multi-source information acquisition module is used to obtain the internal temperature distribution data and contact impedance spectrum data of the lesion through a sensor array, and to form the lesion materialization data by combining the lesion spatial information; A feature extraction module is used to perform spatiotemporal registration on the lesion physical and chemical data using a spatiotemporal registration algorithm to obtain a comprehensive thermal impedance feature of the lesion; A feature modeling module is used to extract the spatiotemporal evolution characteristics of the comprehensive thermal impedance characteristics of the lesion using a lightweight convolutional neural network and to construct a dynamic expansion model of thermal impedance; The instruction generation module is used to generate an initial control instruction sequence based on the thermal impedance dynamic expansion model and in combination with the treatment plan; the initial control instruction sequence is used to control the microwave radio frequency integrated treatment device to achieve corresponding functions.

2. The system according to claim 1, wherein: The system further includes an instruction optimization module for: Combining the thermal impedance dynamic expansion model with the initial control instruction sequence, a reinforcement learning model is used to generate a prediction of treatment process changes; Constructing a multi-objective optimization model based on the deviation between the predicted treatment process changes and the real-time lesion physical and chemical data; The multi-objective optimization model is used to dynamically adjust the initial control instruction sequence to obtain an optimized control instruction sequence.

3. The system according to claim 2, characterized in that The instruction optimization module is further configured to: Comparing the lesion physical and chemical data with a preset matching threshold to determine the thermal impedance state of the lesion; generating dual-frequency tuning parameters of radio frequency or microwave according to the thermal impedance state, and using the tuning parameters as dynamic constraints of the multi-objective optimization model; Based on the multi-objective optimization model, the dynamic constraints are iteratively updated through a rolling horizon control algorithm, and the initial control instruction sequence is synchronously adjusted to obtain the optimized control instruction sequence.

4. The system according to claim 2, wherein: The instruction optimization module is further configured to construct a multi-objective optimization model using the following formula: in, represents the spatial variance of the temperature field in the k-th step of the prediction time domain, represents the square sum of the impedance spatial gradient, α represents the temperature uniformity weight coefficient, γ represents the impedance stability weight coefficient, μ||Δu k || ∞ =max||P k+1 -P k || represents the power fluctuation penalty term, μ represents the power fluctuation penalty coefficient, represents the constraints, A and B represent the state transfer matrix, E x represents the maximum allowed cumulative energy, x k Represents the state variable, u k represents the control variable.

5. The system according to claim 1, wherein: The system also includes an intelligent solution module for: Associating in vitro experimental data with in vivo lesion characteristics through a preset knowledge graph; Using a machine learning model, combined with the matching results of the knowledge graph, historical treatment data, and the thermal impedance dynamic expansion model, a candidate treatment parameter set including radio frequency or microwave power and action time is generated; Based on the candidate treatment parameter set, determining the optimal parameter combination using a multidimensional trade-off optimization model; A dynamic treatment plan is generated according to the optimal parameter combination.

6. The system according to claim 5, characterized in that The system also includes a multi-level monitoring module for: Generate a dynamic reference temperature distribution based on the knowledge graph matching result; Generate a predicted spatial gradient distribution of the temperature field using the thermal impedance dynamic expansion model; Calculating the difference between the reference temperature distribution and the predicted gradient distribution based on the temperature field and impedance gradient data in the lesion physical and chemical data; Comparing the difference value with the historical security boundary threshold in the knowledge graph to generate an abnormality judgment result; According to the abnormality judgment result, a hierarchical response control mechanism is used to trigger corresponding control instructions.

7. The system according to claim 6, characterized in that The hierarchical response control mechanism includes: According to the deviation between the difference value in the abnormality judgment result and the safety boundary threshold, the following steps are performed: When the deviation is less than 15%, a control instruction is generated to attenuate the radio frequency or microwave output power according to a preset ratio; When the deviation reaches 15%-30%, a control instruction and an alarm instruction for pausing the treatment device are generated; When the deviation exceeds 30%, an emergency stop instruction is generated and an abnormality log is created.

8. A microwave radiofrequency integrated treatment method, characterized in that: The method comprises: The sensor array is used to obtain the internal temperature distribution data and contact impedance spectrum data of the lesion, and the lesion spatial information is combined to form the lesion physical and chemical data; Performing spatiotemporal registration on the lesion physical and chemical data using a spatiotemporal registration algorithm to obtain a comprehensive thermal impedance characteristic of the lesion; A lightweight convolutional neural network is used to extract the spatiotemporal evolution characteristics of the comprehensive thermal impedance characteristics of the lesion, and a dynamic expansion model of thermal impedance is constructed; Based on the thermal impedance dynamic expansion model, an initial control instruction sequence is generated in combination with a treatment plan; the initial control instruction sequence is used to control the microwave radio frequency integrated treatment device to achieve corresponding functions.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the system function steps according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the system function according to any one of claims 1 to 7 are realized.

Images