Therapeutic instrument temperature real-time monitoring method based on intelligent sensor

Abstract

The invention discloses a therapeutic instrument temperature real-time monitoring method based on an intelligent sensor. According to the method, a personalized finite element model is constructed by collecting a three-dimensional medical image and real-time sensing data of a patient; forward calculating a theoretical three-dimensional temperature field based on a biological heat transfer model; fusing and correcting the theoretical temperature field and the real-time treatment parameters by using a data assimilation algorithm to obtain an accurate real-time three-dimensional temperature field and a corrected blood perfusion rate; on this basis, the system further calculates a three-dimensional thermal dose field, carries out risk assessment in combination with a real-time temperature field, and finally generates a predictive alarm and closed-loop control instruction; the method has the core advantages that a physical model and sparse sensing data are deeply fused, a digital thermal field model synchronous with physiological characteristics of a patient is constructed, crossing from point temperature measurement to body monitoring is achieved, temperature distribution of a whole treatment area can be comprehensively and accurately predicted in real time, and a monitoring blind area is eliminated.

Description

Technical Field

[0001] This embodiment relates to the field of real-time temperature monitoring technology for therapeutic instruments, specifically a method for real-time temperature monitoring of therapeutic instruments based on intelligent sensors. Background Technology

[0002] In thermotherapy and ablation therapy, current temperature monitoring mainly relies on sensor arrays deployed at discrete spatial points to obtain sparse physical temperature readings.

[0003] This approach can only provide local, discrete temperature measurement data, and cannot present a complete three-dimensional temperature field within the treatment area. It is also difficult to quantitatively assess the overall distribution of cumulative heat dose in the target area and surrounding tissues. Due to the lack of full-field information, treatment decisions heavily rely on limited and lagging physical readings, resulting in the inability to accurately control the treatment energy in a closed loop. This poses fundamental difficulties in assessing treatment effectiveness and predicting potential safety risks.

[0004] Therefore, how to overcome the technical bottleneck of discrete point temperature measurement and achieve real-time accurate prediction and adaptive control of the complete three-dimensional temperature field and biothermal effects in vivo has become a key technical problem that urgently needs to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for real-time temperature monitoring of a therapeutic instrument based on a smart sensor. Specifically, the technical solution of this invention includes:

[0006] The three-dimensional medical image data and real-time sensor data of the patient's treatment area are collected and modeled to obtain a personalized finite element model and real-time treatment parameters.

[0007] Based on the bioheat transfer model and the real-time treatment parameters, the personalized finite element model is subjected to forward calculation to obtain the theoretical three-dimensional temperature field.

[0008] The theoretical three-dimensional temperature field and the real-time treatment parameters are fused and corrected using a data assimilation algorithm to obtain a posterior state estimation vector containing the real-time three-dimensional temperature field and the corrected blood perfusion rate.

[0009] Based on the real-time three-dimensional temperature field contained in the posterior state estimation vector, biological thermal dose calculation is performed to obtain a three-dimensional thermal dose field.

[0010] Based on the three-dimensional thermal dose field and the real-time three-dimensional temperature field, risk assessment is performed to obtain predictive alarms and closed-loop control commands.

[0011] Preferably, the step of acquiring and modeling the three-dimensional medical image data and real-time sensor data of the patient's treatment area to obtain a personalized finite element model and real-time treatment parameters includes:

[0012] Acquire three-dimensional medical imaging data of the patient's treatment area;

[0013] Real-time acquisition of temperature readings and energy output power of the therapeutic instrument at multiple spatial locations generates real-time treatment parameters;

[0014] The three-dimensional medical image data is subjected to image segmentation processing to construct a three-dimensional finite element mesh model;

[0015] The mesh elements of the three-dimensional finite element mesh model are assigned initial prior values ​​of thermophysical properties to generate a personalized finite element model.

[0016] Preferably, the step of performing forward calculations on the personalized finite element model based on the bio-heat transfer model and the real-time treatment parameters to obtain the theoretical three-dimensional temperature field includes:

[0017] Extract the energy output power from the real-time treatment parameters;

[0018] Based on the energy output power and the preset normalized power absorption distribution function, the spatially distributed heat source term is calculated;

[0019] The spatially distributed heat source terms and the personalized finite element model are input into the Penners biological heat transfer equation for numerical solution to obtain the theoretical three-dimensional temperature field.

[0020] Preferably, the step of fusing and correcting the theoretical three-dimensional temperature field and the real-time treatment parameters using a data assimilation algorithm to obtain a posterior state estimation vector containing the real-time three-dimensional temperature field and the corrected blood perfusion rate includes:

[0021] Construct an augmented state vector that includes overall temperature and regional blood perfusion rate;

[0022] An extended Kalman filter framework is used, with the Penners biological heat transfer equation as the state transition function, to predict the augmented state vector and obtain the prior state prediction vector.

[0023] Extract the temperature readings from the real-time treatment parameters to form a measurement vector;

[0024] Based on the measurement vector, the prior state prediction vector is corrected to obtain the posterior state estimation vector.

[0025] Preferably, the correction process is implemented using the following formula:

[0026]

[0027] Where, x t Let be the posterior state estimation vector. Let z be the prior state prediction vector. t Let H be the measurement vector, H be the observation matrix, and Kt be the Kalman gain.

[0028] Preferably, the calculation of the Kalman gain Kt depends on the process noise covariance matrix Q and the measurement noise covariance matrix R;

[0029] The measurement noise covariance matrix R is set according to the calibration accuracy data of the sensor.

[0030] The process noise covariance matrix Q is determined through debugging and optimization using previous clinical data or simulation experiments.

[0031] Preferably, the step of performing biological thermal dose calculation based on the real-time three-dimensional temperature field contained in the posterior state estimation vector to obtain the three-dimensional thermal dose field includes:

[0032] The estimated temperature value of each grid cell in the real-time three-dimensional temperature field is processed by time integration using the cumulative equivalent minute model.

[0033] Wherein, when the estimated temperature value is greater than or equal to 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.5;

[0034] When the estimated temperature value is less than 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.25;

[0035] The results of the time integration process are converted to obtain a three-dimensional thermal dose field in minutes.

[0036] Preferably, the step of performing risk assessment processing based on the three-dimensional thermal dose field and the real-time three-dimensional temperature field to obtain predictive alarms and closed-loop control commands includes:

[0037] The real-time three-dimensional temperature field and the three-dimensional heat dose field are superimposed on the personalized finite element model in the form of a pseudo-color cloud map;

[0038] The three-dimensional thermal dose field is quantitatively evaluated based on preset treatment target thresholds and safety constraint thresholds.

[0039] Based on the results of the quantitative assessment, predictive alerts are generated;

[0040] Based on the results of the quantitative evaluation, a closed-loop control command is generated to adjust the energy output power.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. This method constructs a digital thermal field model that is synchronized with the patient's real-time physiological characteristics by deeply fusing physical models with sparse sensor data, and realizes real-time and accurate prediction of the complete three-dimensional temperature field in the entire treatment area. This overcomes the limitation of existing technologies that rely on discrete point temperature measurement and cannot comprehensively assess the thermal field distribution, and elevates the monitoring from point to volume, eliminating monitoring blind spots.

[0043] 2. This method can transform the physical temperature history into a biologically meaningful cumulative heat dose, generating a dynamically evolving three-dimensional heat dose field. This solves the problem that existing technologies cannot quantify the cumulative killing effect of treatment on biological tissues, enabling treatment decisions to shift from relying on instantaneous temperature readings to being based on quantitative predictions of future biological effects, thereby fundamentally improving the effectiveness of treatment.

[0044] 3. By utilizing data assimilation algorithms, this method enables online dynamic identification and correction of key physiological parameters that are highly individualized and difficult to measure directly. This addresses the limitation of existing technologies in real-time correction of personalized physiological parameters, ensuring that the model accurately reflects the real-time individual characteristics of patients and laying a solid foundation for achieving truly personalized precision treatment.

[0045] 4. Based on the quantitative assessment of the complete three-dimensional temperature field and the three-dimensional thermal dose field, this method can generate predictive alarms and automated closed-loop control commands to adjust the treatment energy in real time. This makes up for the shortcomings of existing technologies that lack accurate prediction and thus fail to adequately assess risks. By introducing forward-looking risk warnings and intelligent automatic control, the risk of damaging surrounding healthy tissues is greatly reduced, and the safety of treatment is significantly improved. Attached Figure Description

[0046] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0047] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0049] Example 1:

[0050] Please see Figure 1 A method for real-time temperature monitoring of a therapeutic instrument based on a smart sensor, comprising:

[0051] The three-dimensional medical image data and real-time sensor data of the patient's treatment area are collected and modeled to obtain a personalized finite element model and real-time treatment parameters.

[0052] Based on the bio-heat transfer model and real-time treatment parameters, a forward calculation is performed on the personalized finite element model to obtain the theoretical three-dimensional temperature field.

[0053] By using a data assimilation algorithm, the theoretical three-dimensional temperature field and real-time treatment parameters are fused and corrected to obtain a posterior state estimation vector containing the real-time three-dimensional temperature field and the corrected blood perfusion rate.

[0054] Based on the real-time three-dimensional temperature field contained in the posterior state estimation vector, biological thermal dose calculation is performed to obtain the three-dimensional thermal dose field.

[0055] Based on the three-dimensional thermal dose field and the real-time three-dimensional temperature field, risk assessment is performed to obtain predictive alarms and closed-loop control commands.

[0056] This invention discloses a method for real-time temperature monitoring of a therapeutic instrument based on intelligent sensors. Its core technology lies in constructing a digital thermal field model that characterizes the real-time physiological characteristics of the patient. Through deep fusion of the physical model and sparse sensor data, it achieves accurate prediction and control of the complete three-dimensional temperature field and cumulative heat dose within the treatment area. This method constitutes a complete and self-consistent technical closed loop, solving the problem that existing technologies rely solely on discrete point temperature measurements and cannot assess the overall treatment effect and safety risks.

[0057] In this embodiment, the complete process of the method is embodied in the following interrelated steps:

[0058] The process involves acquiring and modeling three-dimensional medical imaging data and real-time sensor data of the patient's treatment area to obtain a personalized finite element model and real-time treatment parameters. The personalized finite element model refers to a three-dimensional mesh model constructed based on the specific patient's CT or MRI image data, perfectly mirroring the patient's anatomical structure, and endowed with initial tissue thermophysical properties. The real-time treatment parameters refer to the dynamically acquired data set during treatment, primarily including the energy output power of the treatment device and sparse temperature readings measured by sensor arrays implanted or applied to the patient's body surface / body. The purpose of this step is to create a precise, patient-specific digital carrier and real-time data input source for all subsequent calculations and simulations.

[0059] Based on the bio-heat transfer model and real-time treatment parameters, a forward calculation is performed on the personalized finite element model to obtain the theoretical three-dimensional temperature field. The essence of this step is physical modeling and prediction. The bio-heat transfer model, in this embodiment, specifically refers to the Pennes bio-heat transfer equation, which is a partial differential equation describing the heat transfer law in living tissue. Using the real-time energy output power obtained in the previous step as the heat source term, the personalized finite element model is numerically solved under the constraint of this equation. The output of this step is the theoretical three-dimensional temperature field, which represents an idealized, a priori temperature distribution prediction without considering model parameter errors and individual physiological dynamic changes.

[0060] The core step of this invention involves fusing and correcting the theoretical three-dimensional temperature field and real-time treatment parameters using a data assimilation algorithm to obtain a posterior state estimation vector containing the real-time three-dimensional temperature field and the corrected blood perfusion rate. The data assimilation algorithm is a mathematical method that optimally fuses model predictions with actual observation data. Its purpose is to use sparse but realistic observation data to correct the uncertain full-field model predictions. The key output of this step is the posterior state estimation vector, which is an augmented data vector containing two core pieces of information: first, a high-fidelity real-time three-dimensional temperature field covering the entire treatment area after correction by sensor data; and second, a dynamically identified and optimized key physiological parameter, namely the corrected blood perfusion rate. Blood perfusion rate is a core factor affecting the efficiency of heat removal from tissues and is also the parameter with the greatest individual variation and the most difficult to measure. Online correction of this parameter is crucial for achieving truly personalized treatment.

[0061] Based on the real-time three-dimensional temperature field contained in the posterior state estimation vector, biological thermal dose calculation is performed to obtain a three-dimensional thermal dose field. This step aims to transform the physical temperature history into a biologically meaningful assessment of therapeutic effect. Biological thermal dose is a quantitative indicator used to measure the cumulative killing effect of thermal energy on biological tissues, and the commonly used unit is cumulative equivalent minutes. By integrating the temperature history of each point in the real-time three-dimensional temperature field over time, a dynamically evolving three-dimensional thermal dose field can be obtained, which intuitively reflects the degree to which each part of the treatment area has reached the therapeutic dose.

[0062] Based on the three-dimensional thermal dose field and the real-time three-dimensional temperature field, risk assessment is performed to obtain predictive alarms and closed-loop control commands. This step is the exit point for decision-making and control. It compares the real-time temperature field and thermal dose field calculated in the previous step with the preset clinical goals and safety boundaries. Once the system detects that the current or predicted future state may deviate from the preset goals or touch the safety red line, it will immediately generate a predictive alarm to remind the operator, or directly generate a closed-loop control command to automatically adjust the energy output of the treatment device, thereby realizing intelligent and adaptive control of the treatment process.

[0063] This invention achieves a dimensional shift from point measurement to volume monitoring through the aforementioned technological closed loop. It not only measures temperature but also quantifies and predicts the complete spatiotemporal effects of therapeutic energy within the body by constructing a digital model that is synchronized with the patient in real time. This transforms treatment decisions from relying on discrete and lagging physical readings to being based on quantitative predictions of future biological effects, thereby improving the effectiveness of thermotherapy or ablation therapy and greatly reducing the risk of damaging surrounding healthy tissues, thus achieving personalized and precise treatment.

[0064] Example 2:

[0065] The steps for acquiring and modeling three-dimensional medical image data and real-time sensor data of the patient's treatment area to obtain a personalized finite element model and real-time treatment parameters include:

[0066] Acquire three-dimensional medical imaging data of the patient's treatment area;

[0067] Real-time acquisition of temperature readings and energy output power of the therapeutic instrument at multiple spatial locations generates real-time treatment parameters;

[0068] Image segmentation processing is performed on 3D medical image data to construct a 3D finite element mesh model;

[0069] Assign initial thermophysical property prior values ​​to the mesh elements of the three-dimensional finite element mesh model to generate a personalized finite element model.

[0070] This embodiment is a specific implementation of the steps in Embodiment 1 for collecting and modeling three-dimensional medical image data and real-time sensor data of the patient's treatment area. Its purpose is to construct a high-fidelity personalized digital model that can carry all subsequent physical calculations.

[0071] This step specifically includes:

[0072] Acquire three-dimensional medical imaging data of the patient's treatment area; this step provides an anatomical basis for model construction; in this embodiment, during the treatment planning stage, sequence image data of the patient's treatment target area and its surrounding tissues are acquired using high-resolution computed tomography or magnetic resonance imaging equipment, and the data format can be the DICOM standard format;

[0073] The system collects temperature readings and the energy output power of the therapeutic device at multiple spatial locations in real time to generate real-time treatment parameters. These parameters serve as the real-time data input driving the model's real-time evolution. In this embodiment, a sensor array containing four fiber optic temperature probes is implanted or applied to preset monitoring points to continuously collect temperature readings at a sampling frequency of 1Hz. Simultaneously, the output acoustic power P of the high-intensity focused ultrasound therapeutic device is read in real time from its controller via a communication interface. in(t); These two sets of data together constitute the real-time treatment parameter vector;

[0074] The three-dimensional medical image data is segmented to construct a three-dimensional finite element mesh model. This step aims to transform the medical images into a mathematical model that can be physically calculated. In this embodiment, medical image processing software is used to semi-automatically segment the acquired CT images to accurately identify and delineate the boundaries of different tissues such as tumors, liver, and large blood vessels. Subsequently, based on these segmented three-dimensional contours, a tetrahedral mesh partitioning algorithm is used to generate an unstructured three-dimensional finite element mesh model containing approximately 500,000 elements.

[0075] Initial thermophysical property prior values ​​are assigned to the mesh elements of the three-dimensional finite element mesh model to generate a personalized finite element model; this step sets the initial conditions for the physical behavior of the model. Based on the image segmentation results, the tissue type to which each mesh element belongs is labeled. Then, according to the biophysical property database recognized in the field, corresponding initial thermophysical properties are assigned to the mesh elements of different tissues, including density ρ, specific heat capacity c, thermal conductivity k, and blood perfusion rate ω. b The prior values; after completing this step, a personalized finite element model integrating the patient's precise anatomical structure and initial physical properties is constructed;

[0076] Compared to general modeling methods, this embodiment ensures the dual personalization of the computational model through the above-described specific steps: first, personalization of anatomical structure, accurately reproducing the unique spatial relationship between the patient's tumor and surrounding organs; second, personalization of real-time input, using the actual energy applied during treatment and the resulting temperature response as the driving force of the model. This highly customized model lays a solid foundation for the accuracy of all subsequent predictions and corrections.

[0077] Example 3:

[0078] Based on the bio-heat transfer model and real-time treatment parameters, a forward calculation process is performed on the personalized finite element model to obtain the theoretical three-dimensional temperature field. The steps include:

[0079] Extract the energy output power from real-time treatment parameters;

[0080] Based on the energy output power and the preset normalized power absorption distribution function, the spatially distributed heat source term is calculated;

[0081] By inputting the spatially distributed heat source terms and the personalized finite element model into the Penners biological heat transfer equation, numerical solutions are obtained to obtain the theoretical three-dimensional temperature field.

[0082] This embodiment is a concrete implementation of the steps in Embodiment 1, which involves forward calculation of a personalized finite element model based on a bio-heat transfer model and real-time treatment parameters to obtain a theoretical three-dimensional temperature field. Its core lies in transforming the input energy parameters into a physical prediction with a full-field distribution.

[0083] This step specifically includes: extracting the energy output power from the real-time treatment parameters; the starting point for the calculation is the energy input; the system accurately extracts the energy output power value P of the treatment device at the current time t from the real-time acquired treatment parameter vector. in (t), the unit is watt;

[0084] Based on the energy output power and the preset normalized power absorption distribution function, the spatially distributed heat source term is calculated. The purpose of this step is to reasonably distribute a total power value to every point in three-dimensional space. The normalized power absorption distribution function, η(x,y,z), is a normalized function describing the spatial distribution of energy within the tissue. Its volume fraction over the entire computational domain is equal to 1, and its dimension is m. -3 Therefore, the dimension of this function is m. -3 The form of the function η(x,y,z) depends on the type of energy source; for example, in the high-intensity focused ultrasound therapy used in this embodiment, the function η(x,y,z) can typically be modeled as a three-dimensional Gaussian function or a sound intensity distribution field pre-calculated based on acoustic simulation to describe the energy deposition morphology in the focal region; the spatially distributed heat source term Q r The formula for calculating (x,y,z,t) is:

[0085] Q r (x,y,z,t)=P in (t)·η(x,y,z)

[0086] Its physical meaning is the energy power absorbed per unit volume of tissue at position (x,y,z) and time t, with the unit being watts per meter. 3 ;

[0087] The spatially distributed heat source terms and the personalized finite element model are input into the Pennesian biological heat transfer equation for numerical solution to obtain the theoretical three-dimensional temperature field; this is the core calculation for performing physical predictions; the Pennesian biological heat transfer equation is the biological heat transfer model used in this embodiment, and its standard form is:

[0088]

[0089] Where ρ, c, and k are the density, specific heat capacity, and thermal conductivity of the tissue, respectively, and their values ​​are derived from the prior values ​​assigned based on the tissue type in the personalized finite element model.

[0090] T(x,y,z,t): is the tissue temperature to be solved;

[0091] ω b : This is the blood perfusion rate, whose initial value also comes from the prior settings of the model and is the key object to be corrected in subsequent steps;

[0092] c b The specific heat capacity of blood is usually set as a constant based on physiological data.

[0093] T a Arterial blood temperature is usually set as the body core temperature based on physiological data;

[0094] Q m Tissue metabolic heat production rate, which is relatively small and can usually be ignored or set as a baseline value;

[0095] Q r That is, the spatially distributed heat source term obtained from the preceding steps;

[0096] By using a finite element method to solve the equation transiently, the theoretical three-dimensional temperature field T at the current moment can be obtained. pred ;

[0097] This embodiment concretizes forward calculation into a clear, first-principles-based physical prediction process through explicit steps; it not only specifies the core governing equations but also clarifies how to transform the engineering parameter of device output power into the heat source term Q. r The construction of the model is seamlessly coupled with the biophysical model, which makes the theoretical predictions not only based on evidence, but also transparent and reproducible, providing reliable prior information for subsequent data assimilation and correction.

[0098] Example 4:

[0099] The theoretical three-dimensional temperature field and real-time treatment parameters are fused and corrected using a data assimilation algorithm to obtain a posterior state estimation vector containing the real-time three-dimensional temperature field and the corrected blood perfusion rate. The steps include:

[0100] Construct an augmented state vector that includes overall temperature and regional blood perfusion rate;

[0101] An extended Kalman filter framework is used, with the Penners biological heat transfer equation as the state transition function, to predict the augmented state vector and obtain the prior state prediction vector.

[0102] Extract temperature readings from real-time treatment parameters to form a measurement vector;

[0103] Based on the measurement vector, the prior state prediction vector is corrected to obtain the posterior state estimation vector.

[0104] The correction process is achieved using the following formula:

[0105]

[0106] Where, x t This is the posterior state estimation vector. Let z be the prior state prediction vector. t Let H be the measurement vector, H be the observation matrix, and Kt be the Kalman gain.

[0107] The calculation of the Kalman gain Kt depends on the process noise covariance matrix Q and the measurement noise covariance matrix R;

[0108] Among them, the measurement noise covariance matrix R is set according to the calibration accuracy data of the sensor;

[0109] The process noise covariance matrix Q is determined through debugging and optimization using previous clinical data or simulation experiments.

[0110] This embodiment is a concretization and optimization of the steps in Embodiment 1, which use a data assimilation algorithm to fuse and correct the theoretical three-dimensional temperature field and real-time treatment parameters. Together, they constitute a robust and efficient state estimation algorithm module. Its core purpose is to use sparse real temperature measurement data to correct the prediction bias of the theoretical model in real time and to deduce key unknown physiological parameters.

[0111] In this embodiment, the data assimilation algorithm adopts the extended Kalman filter framework, and its implementation steps are as follows:

[0112] An augmented state vector containing the overall temperature and regional blood perfusion rate is constructed. To simultaneously estimate the temperature field and correct physiological parameters, this step defines an augmented state vector x. t The structure of this vector is x. t =[T1,T2,...,T N ,ω b,1 ,...,ω b,M ] T ; where [T1,...,T N ] T This represents the temperature values ​​of all N nodes in the finite element model, [ω b,1 ,...,ω b,M ] T This represents the average blood perfusion rate of the M different regions to be identified; the M regions here are predefined sub-regions with different physiological or anatomical features through image segmentation during the personalized finite element model construction stage, and each region contains several finite element nodes.

[0113] An extended Kalman filter (EKF) framework is employed, using the Penners biological heat transfer equation as the state transition function to predict the augmented state vector, yielding a priori state prediction vector. The EKF algorithm follows a prediction-correction loop. During the prediction phase, the system uses the final optimal estimate x from the previous time step. t-1 As initial conditions, the state at the current time t is predicted by solving the Penners biological heat transfer equation, thus obtaining the prior state prediction vector.

[0114] The temperature readings from the real-time treatment parameters are extracted to form a measurement vector; this step prepares the measured data for calibration. The system extracts temperature readings from k sensors from the real-time treatment parameters to form a measurement vector z. t =[T meas,1 ,...,T meas,k ] T ; in forming the measurement vector z t Previously, the system first preprocessed the raw temperature readings collected by the sensors, including effective range checks and outlier detection. If a sensor reading is found to exceed a preset physical threshold or the difference between the reading and the model prediction or other sensor readings exceeds the statistical limit, the reading is marked as an anomaly and will be ignored or given a very low weight in the correction step at the current moment. This step ensures the robustness of the system and prevents the failure of the entire temperature field estimation due to the failure of a single sensor.

[0115] Based on the measurement vector, the prior state prediction vector is corrected to obtain the posterior state estimation vector; this is the correction stage of EKF and the core of fusion; this correction process is implemented through the following formula:

[0116]

[0117] Where, x t The final output is the posterior state estimation vector, which contains the optimal estimate of the total field temperature T at the current moment. est and the corrected blood perfusion rate ω b ;

[0118] The prior state prediction vector obtained from the aforementioned prediction steps;

[0119] z t : Measurement vectors obtained in real time by sensors;

[0120] H: Observation matrix, its function is to obtain the complete state vector Extract the theoretically predicted temperature values ​​corresponding to the k sensor locations, so as to compare them with the measured value z. t Compare;

[0121] Kt Kalman gain is a dynamically calculated weight matrix used to determine the extent to which new measurement data is accepted to correct the model's predictions.

[0122] Kalman gain K t The calculation relies on two key noise covariance matrices: the process noise covariance matrix Q and the measurement noise covariance matrix R;

[0123] The measurement noise covariance matrix R represents the uncertainty of the sensor itself. In this embodiment, R is set as a diagonal matrix, and its diagonal elements are set according to the sensor's factory calibration accuracy data. For example, if the sensor's calibration accuracy is the standard deviation σ... T =0.1℃, then the corresponding diagonal element value is

[0124] The process noise covariance matrix Q represents the uncertainty of the physical model itself, mainly reflecting the blood perfusion rate ω. b Unpredictable physiological fluctuations in parameters such as Q; the determination of Q is achieved through playback analysis of previously accumulated clinical treatment data or through extensive simulation experiments for debugging and optimization, in order to achieve the best filtering and parameter identification effect; specifically, the Q matrix is ​​usually set as a diagonal matrix, with its diagonal elements Q ii This represents the variance that the i-th state in the state vector may generate within a single time step. Its value can be determined by the following offline analysis method: using a set of historical case data that has been treated, the forward calculation model of this invention is run in open-loop mode without applying Kalman filter correction; the temperature prediction value of the model at the sensor location is compared with the actual sensor temperature measurement value, and the residual sequence between the two at each time step is calculated; the variance of this residual sequence, after statistical analysis and empirical adjustment, can be used as the basis for setting the process noise variance of the corresponding temperature state in the Q matrix; for the blood perfusion rate state, its process noise variance can be set with an initial range based on physiological knowledge, and iteratively fine-tuned by analyzing the convergence and stationarity of the blood perfusion rate after filtering, with the goal of making its estimated value change smoothly within a physiologically reasonable range rather than oscillating violently; this statistical analysis and iterative fine-tuning process is solidified into a standardized procedure to ensure the objectivity and repeatability of the Q matrix setting when treating different cases;

[0125] This invention constructs a mathematically optimal and physically meaningful data assimilation framework. It not only solves the problem of how to fuse models and data, but also provides a specific mathematical expression for achieving the fusion, and further clarifies the source and setting basis of key parameters in the mathematical expression. This makes the core algorithm no longer an unknown module, but a transparent, robust and engineerable module. The direct effect is that it can accurately infer the complete temperature distribution of the entire three-dimensional treatment area from a few extremely sparse temperature measurement points, and at the same time identify key personalized physiological parameters that cannot be directly measured, such as blood perfusion rate, which determine the success or failure of treatment.

[0126] Example 5:

[0127] Based on the real-time three-dimensional temperature field contained in the posterior state estimation vector, the biological thermal dose calculation process is performed to obtain the three-dimensional thermal dose field, including:

[0128] The estimated temperature value of each grid cell in the real-time three-dimensional temperature field is processed by time integration using the cumulative equivalent minute model.

[0129] Specifically, when the estimated temperature value is greater than or equal to 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.5;

[0130] When the estimated temperature is less than 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.25;

[0131] The results of the time integration process are converted to obtain a three-dimensional thermal dose field in minutes.

[0132] This embodiment is a concrete implementation of the step in Embodiment 1 to calculate the three-dimensional thermal dose field based on the real-time three-dimensional temperature field contained in the posterior state estimation vector; its purpose is to transform instantaneous physical temperature information into a cumulative biological effect indicator that can directly guide clinical decision-making.

[0133] This step uses a cumulative equivalent minute model to analyze the real-time three-dimensional temperature field T output from the previous step. est The estimated temperature value of each grid cell in (t) is processed by time integration; the calculation method is to discretize and sum over time:

[0134]

[0135] To obtain results in minutes, unit conversion is required. The final calculation formula is:

[0136]

[0137] Wherein, CEM43: the output caloric dose value, in minutes;

[0138] T est (t i ): The estimated temperature value of a specific grid cell at the i-th time step, which is derived from the posterior state estimation vector;

[0139] Δt i : The time step for calculation, in seconds;

[0140] 1 / 60: is the conversion factor for converting seconds to minutes;

[0141] R cem : is a dimensionless constant whose value reflects the nonlinear relationship between temperature and biological damage effects. The rule for determining this value is based on empirical relationships summarized from cell survival experimental data, and is specifically set as follows:

[0142] When the estimated temperature value is greater than or equal to 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.5;

[0143] When the estimated temperature is less than 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.25;

[0144] This calculation is performed independently for each grid cell in the real-time three-dimensional temperature field, and is continuously accumulated from the start of treatment. Finally, at each moment, a three-dimensional matrix with the same spatial resolution as the real-time three-dimensional temperature field is obtained, which is the three-dimensional thermal dose field.

[0145] This embodiment solidifies radiobiology theory into a clear and executable computational procedure; it introduces the CEM43 model and clarifies its key parameter R. cem By establishing a set of value rules, this invention elevates the evaluation criteria for the treatment process from physical quantities to biological effect quantities. This allows doctors to intuitively determine whether the treatment has achieved its objectives. For example, clinical practice shows that for most solid tumors, a CEM43 value exceeding 240 minutes usually indicates complete tissue necrosis. Therefore, this quantitative evaluation method provides a scientific basis for setting clear treatment endpoints and safety thresholds, enhancing the controllability of treatment.

[0146] Example 6:

[0147] Based on the three-dimensional thermal dose field and the real-time three-dimensional temperature field, risk assessment is performed to obtain predictive alarm and closed-loop control command steps, including:

[0148] The real-time three-dimensional temperature field and the three-dimensional heat dose field are superimposed on the personalized finite element model in the form of a pseudo-color cloud map;

[0149] Based on preset treatment target thresholds and safety constraint thresholds, the three-dimensional thermal dose field is quantitatively evaluated;

[0150] Based on the results of the quantitative assessment, predictive alerts are generated;

[0151] Based on the results of the quantitative evaluation, closed-loop control commands are generated to adjust the energy output power.

[0152] This embodiment is a concrete implementation of the steps of predictive alarm and closed-loop control instructions obtained by risk assessment based on the three-dimensional thermal dose field and real-time three-dimensional temperature field in Embodiment 1. Its purpose is to transform the calculation results into visual information and decision instructions that have direct guiding significance for clinical operation.

[0153] This step specifically includes:

[0154] The real-time three-dimensional temperature field and the three-dimensional thermal dose field are superimposed on the personalized finite element model in the form of pseudo-color cloud maps; this is the visualization output. On a three-dimensional interactive interface, the patient's anatomical structure is displayed in a semi-transparent manner, and then the real-time temperature field and thermal dose field are rendered separately or simultaneously in the form of pseudo-color cloud maps and superimposed on the anatomical structure.

[0155] Based on preset treatment target thresholds and safety constraint thresholds, the three-dimensional thermal dose field is quantitatively evaluated; this is a quantitative analysis. The treatment target thresholds and safety constraint thresholds are key performance indicators set according to clinical guidelines and previous studies; in this embodiment, the threshold logic is set as follows:

[0156] Treatment goal: The percentage of tumor tissue with CEM43 > 240 minutes should reach 95% or more; this threshold is determined based on clinical data statistics that lead to complete tumor necrosis.

[0157] Safety constraints: Nerve tissue within 5mm of the tumor boundary, CEM43 should not exceed 4 minutes; adjacent large blood vessel walls, instantaneous temperature T est Temperature must not exceed 55°C; this threshold is based on industry safety standards to prevent irreversible damage to heat-sensitive tissues.

[0158] Based on the results of the quantitative assessment, predictive alerts are generated; this is a risk warning. The system not only assesses the current state but also considers temperature change trends. The system extrapolates and predicts the evolution of heat dose within the next 30 seconds; when the system predicts that the heat dose in a critical area will exceed the limit based on the quantitative assessment results, it will trigger a predictive alarm.

[0159] Based on the quantitative evaluation results, a closed-loop control command is generated to adjust the energy output power; this is closed-loop control. In automatic control mode, the system uses the quantitative evaluation results as feedback input and automatically generates a control command for adjusting the energy output power P of the therapeutic device through a proportional-integral-derivative controller. inThe adjustment instructions; specifically, the controller's input error signal e(t) is defined as the difference between the core performance index of the target area and the preset treatment target.

[0160] This embodiment constructs a complete information closed loop from calculation to decision-making. Through intuitive visualization, clear quantitative assessment, forward-looking risk warning, and intelligent automatic control, it transforms complex background calculation results into decision support information that is of great value to doctors. This not only reduces the cognitive burden on operators, but also raises the safety and accuracy of treatment to a new level by introducing prediction and closed-loop control, ensuring the achievement of the core treatment goal of ablating lesions and protecting normal tissue.

[0161] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for real-time monitoring of the temperature of an intelligent sensor-based therapeutic instrument, characterized in that, The method comprises the following steps: Collecting and modeling the three-dimensional medical image data and real-time sensing data of the patient's treatment area to obtain a personalized finite element model and real-time treatment parameters; Based on the biological heat transfer model and the real-time treatment parameters, the personalized finite element model is calculated to obtain a theoretical three-dimensional temperature field; Through a data assimilation algorithm, the theoretical three-dimensional temperature field and the real-time treatment parameters are fused and corrected to obtain a posterior state estimation vector containing a real-time three-dimensional temperature field and a corrected blood perfusion rate; Based on the real-time three-dimensional temperature field contained in the posterior state estimation vector, biological heat dose calculation processing is performed to obtain a three-dimensional heat dose field; Based on the three-dimensional heat dose field and the real-time three-dimensional temperature field, risk assessment processing is performed to obtain a predictive alarm and closed-loop control instructions.

2. The intelligent sensor-based therapeutic apparatus temperature real-time monitoring method according to claim 1, characterized in that, The step of collecting and modeling the three-dimensional medical image data and real-time sensing data of the patient's treatment area to obtain a personalized finite element model and real-time treatment parameters comprises: Obtaining three-dimensional medical image data of the patient's treatment area; Real-time collection of temperature readings and energy output power of the treatment instrument at multiple spatial positions to generate real-time treatment parameters; Image segmentation processing is performed on the three-dimensional medical image data to construct a three-dimensional finite element grid model; Initial thermal physical property prior values are assigned to the grid elements of the three-dimensional finite element grid model to generate a personalized finite element model.

3. The intelligent sensor based temperature monitoring method of therapeutic apparatuses in real time as claimed in claim 1 wherein, The step of calculating the personalized finite element model based on the biological heat transfer model and the real-time treatment parameters to obtain a theoretical three-dimensional temperature field comprises: Extracting the energy output power from the real-time treatment parameters; Based on the energy output power and a pre-set normalized power absorption distribution function, a spatially distributed heat source term is calculated; The spatially distributed heat source term and the personalized finite element model are input into the Pennes biological heat transfer equation for numerical solution to obtain a theoretical three-dimensional temperature field.

4. The intelligent sensor based therapeutic apparatus temperature real time monitoring method as claimed in claim 1 wherein, The step of fusing and correcting the theoretical three-dimensional temperature field and the real-time treatment parameters by a data assimilation algorithm to obtain a posterior state estimation vector containing a real-time three-dimensional temperature field and a corrected blood perfusion rate comprises: Constructing an augmented state vector containing the full-field temperature and regional blood perfusion rate; Using an extended Kalman filter framework, the Pennes biological heat transfer equation is used as a state transition function to predict the augmented state vector to obtain a prior state prediction vector; Extracting the temperature readings from the real-time treatment parameters to form a measurement vector; Based on the measurement vector, the prior state prediction vector is corrected to obtain a posterior state estimation vector.

5. The intelligent sensor based therapeutic apparatus temperature real time monitoring method as claimed in claim 4, wherein, The correction processing is realized by the following formula: where x t is the posterior state estimate vector, is the prior state prediction vector, z t is the measurement vector, H is the observation matrix, and Kt is the Kalman gain.

6. The intelligent sensor-based therapeutic apparatus temperature real-time monitoring method according to claim 5, wherein, The calculation of the Kalman gain Kt depends on the process noise covariance matrix Q and the measurement noise covariance matrix R; The measurement noise covariance matrix R is set according to the calibration accuracy data of the sensor; The process noise covariance matrix Q is determined by adjusting and optimizing the previous clinical data or simulation experiments.

7. The smart sensor based therapeutic temperature real-time monitoring method as claimed in claim 1, wherein, The step of performing a biological thermal dose calculation process based on the real-time three-dimensional temperature field included in the posterior state estimation vector to obtain a three-dimensional thermal dose field comprises: A cumulative equivalent minute model is used to perform a time integration process on the estimated temperature value of each grid cell in the real-time three-dimensional temperature field; When the estimated temperature value is greater than or equal to 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.5; When the estimated temperature value is less than 43 degrees Celsius, the calculation constant of the cumulative equivalent minute model is set to 0.25; The result of the time integration process is unit-converted to obtain a three-dimensional thermal dose field in minutes.

8. The smart sensor based therapeutic apparatus temperature real-time monitoring method as claimed in claim 1, wherein, The step of performing a risk assessment process based on the three-dimensional thermal dose field and the real-time three-dimensional temperature field to obtain a predictive warning and a closed-loop control instruction comprises: The real-time three-dimensional temperature field and the three-dimensional thermal dose field are superimposed in the form of a pseudo-color cloud chart on the personalized finite element model; The three-dimensional thermal dose field is quantitatively evaluated based on preset treatment target thresholds and safety constraint thresholds; A predictive warning is generated based on the result of the quantitative evaluation; A closed-loop control instruction for adjusting the energy output power is generated based on the result of the quantitative evaluation.

Images