A tumor dose monitoring method and system based on radiotherapy

Abstract

The present invention discloses a method and system for monitoring tumor dose based on radiotherapy, which relates to the technical field of medical care informatics. The method and system for monitoring tumor dose based on radiotherapy include the following steps: S1, using an implantable fiber optic dose detector to collect tumor dose data, and performing data cleaning and normalization processing on the tumor dose data; S2, performing preliminary dose estimation and physical modeling compensation on the pre-processed tumor dose data, and constructing a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution; S3, analyzing the deviation intensity based on the comparison result between the tumor dose assessment value and the reference surface dose, making a graded judgment on the deviation intensity and generating adjustment suggestions, forming a control process for closed-loop feedback; S4, focusing on the statistical deviation distribution characteristics of the deep estimation point of the target area, and generating structured correction data for subsequent reconstruction input and risk prompt generation. It solves the problem that surface dose monitoring is difficult to accurately restore the three-dimensional dose distribution of the tumor area in the body.

Description

Technical Field

[0001] The present invention relates to the technical field of medical care informatics, and in particular to a method and system for monitoring tumor dose based on radiotherapy. Background Art

[0002] As radiotherapy technology evolves toward high-dose, precisely targeted, and multi-angle irradiation, the role of dose monitoring in ensuring treatment accuracy and safety is becoming increasingly prominent. In recent years, the research focus of tumor dose monitoring methods and systems has gradually shifted from verifying plan execution consistency to real-time feedback and individualized dose control.

[0003] For example, the invention with publication number: CN118588235A discloses a precise control system for radioactive drug dose, including: a microfluidic drug reservoir, a high-precision micropump, a precision dose control unit, a drug concentration monitoring module, a data analysis module, a dose adjustment module and a user interface: the microfluidic drug reservoir is loaded with radioactive drugs, the high-precision micropump releases the drugs according to a preset dose, and the precision dose control unit is connected to the microfluidic drug reservoir, the drug concentration monitoring module and the dose adjustment module.

[0004] For example, the invention with publication number: CN119517302A discloses an intelligent medical management method and system based on radiotherapy of malignant tumors, including the following steps: obtaining holographic image data corresponding to the patient's clinical image data; inputting the holographic image data into a trained deep learning model to identify the malignant tumor target area and its attributes; calculating the dose of radioactive particles based on the malignant tumor target area and its attributes; generating 3D printing instructions according to the malignant tumor target area; obtaining the dose distribution monitoring data of the radioactive particles in the treatment area; analyzing the uniformity of the radioactive particles based on the dose distribution monitoring data; generating a treatment plan and a treatment risk assessment based on the malignant tumor target area, attributes, and uniformity of the radioactive particles.

[0005] The dose information obtained by these methods is limited to the body surface and cannot reflect the actual energy deposition process of the radiation in the complex tissue structure in the body. Especially when treating deep tumors, due to changes in tissue density, scattering effects and uncertainty in the beam path, there may be a significant deviation between the surface dose and the actual target dose, resulting in a decrease in the accuracy of dose reconstruction.

[0006] Therefore, in response to the above problems, there is an urgent need for a tumor dose monitoring method and system based on radiotherapy. Summary of the Invention

[0007] Technical problems solved

[0008] In response to the shortcomings of the existing technology, the present invention provides a tumor dose monitoring method and system based on radiotherapy, which solves the problem that surface dose monitoring is difficult to accurately restore the three-dimensional dose distribution of the tumor area in the body.

[0009] Technical Solution

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a tumor dose monitoring method and system based on radiotherapy, comprising the following steps: S1, using an implantable fiber optic dose detector to collect tumor dose data, and performing data cleaning and normalization processing on the tumor dose data; S2, performing preliminary dose estimation and physical modeling compensation on the preprocessed tumor dose data, and constructing a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution; S3, analyzing the deviation intensity based on the comparison results of the tumor dose assessment value and the reference surface dose, performing graded judgment on the deviation intensity and generating adjustment suggestions, forming a control process for closed-loop feedback; S4, focusing on the statistical deviation distribution characteristics of the deep estimation points in the target area, and generating structured correction data for subsequent reconstruction input and risk prompt generation.

[0011] Furthermore, the specific steps for collecting tumor dose data using an implantable fiber optic dose detector are as follows: an implantable fiber optic dose detector is used, which is deployed in the tumor target area and its periphery to sense radiation-induced light signals in real time and collect tumor dose data; the tumor dose data includes instantaneous radiation dose, detection point coordinates, beam source coordinates, estimated point coordinates, photon beam energy, tissue CT value and reference surface dose; wherein, the instantaneous radiation dose of each detection point is obtained through photoelectric conversion, signal amplification and analog-to-digital conversion; the spatial position of the fiber optic probe is calibrated through preoperative CT scanning, the patient coordinate system is constructed, and the detection point coordinates, beam source coordinates and estimated point coordinates in the patient's body are obtained; the photon beam energy of the radiation used for treatment is obtained through the radiotherapy equipment control system; the tissue CT value of the tissue corresponding to the detection point is read from the patient's preoperative CT image; the tissue angle of the current beam relative to the patient coordinate system is read in real time by connecting to the control system of the radiotherapy accelerator; and the reference surface dose is obtained through the TLD detector attached to the patient's body surface.

[0012] Furthermore, the specific steps for cleaning and normalizing the tumor dose data are as follows: using the sliding window method combined with the triple standard difference constant detection algorithm to eliminate tumor dose data that obviously deviates from the normal range; using millisecond-level time alignment and linear interpolation methods to fill in the instantaneous missing tumor dose data; using the rigid registration algorithm to align the detector space coordinates with the patient's preoperative CT image coordinates to ensure three-dimensional spatial consistency; using the Kalman filter algorithm to denoise the continuously collected instantaneous radiation dose; using unit conversion and maximum and minimum normalization algorithms to normalize the tumor dose data and unify the data scale.

[0013] Furthermore, the specific steps for performing preliminary dose estimation and physical modeling compensation on the preprocessed tumor dose data are as follows: take N detection points, calculate the square of the coordinate difference between the estimated point and the i-th detection point in the transverse direction, the square of the coordinate difference in the longitudinal direction, and the square of the coordinate difference in the vertical direction, add a minimum term, and finally square the result to obtain the spatial distance between the i-th detection point and the estimated point in the patient coordinate system; multiply the spatial distance by the exponential decay factor and take the negative value as the exponent, and then calculate the value of this power function with the natural constant as the base to obtain the weight of the current detection point; multiply the weight by the instantaneous radiation dose as the weighted value of the point, repeat the above process to sum the N detection points, and use the sum result as the numerator; only weight the N detection points The power function is calculated repeatedly to obtain the weight of the current estimated point, which is not multiplied by the instantaneous radiation dose, and the sum is used as the denominator; the numerator is divided by the denominator to obtain the spatial interpolation dose assessment value, and the preliminary dose estimation of the area without detectors is completed; the photon beam energy, irradiation angle, detection point coordinates, beam source coordinates and tissue CT value are obtained, and the photon beam energy, the cosine value of the tissue angle and the tissue CT value are multiplied together, and the product of the three is divided by the square of the spatial distance between the estimated point coordinates and the beam source coordinates, and a minimum term is added. The sum is then multiplied by the physical weight to obtain the first part; the instantaneous radiation dose is multiplied by the measured weight to obtain the second part; the sum of the two parts is calculated to obtain the physical modeling dose assessment value, and the dose value of each estimated point in three-dimensional space is physically compensated by modeling.

[0014] Furthermore, the specific steps of constructing a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution are as follows: weighted fusion of the spatial interpolation dose assessment value and the physical modeling dose assessment value to obtain the tumor dose assessment value, and use the tumor dose assessment value to generate an intermediate dose reconstruction map driven by physical and observation data; the patient's preoperative CT image, tumor dose data and intermediate dose reconstruction map are used as joint input to construct a three-dimensional convolutional neural network model and carry out training and prediction to complete the dose information of the area where no detection points are deployed and refine the overall dose distribution; weighted fusion of the network prediction results and the intermediate dose reconstruction map to generate a unified final three-dimensional dose map; use a rigid registration algorithm to align the coordinates of the final three-dimensional dose map with the patient's preoperative CT image to achieve accurate matching of the spatial image and the dose field.

[0015] Furthermore, the specific steps for analyzing the deviation intensity based on the comparison results of the tumor dose assessment value and the reference surface dose are as follows: the tumor dose assessment value of each estimated point in the final three-dimensional dose map is compared with the reference surface dose one by one, and the comprehensive deviation intensity value of each estimated point is calculated; the tumor dose assessment value of the estimated point is subtracted from the reference surface dose and divided by the reference surface dose to express the relative dose deviation; then, the relative dose deviation is multiplied by the tissue CT value of the estimated point as the first part; the sum of the squares of the coordinate differences between the estimated point and the beam source in the transverse, longitudinal and vertical directions in the spatial coordinates is calculated, and the result is divided by the square of the spatial scale normalization constant to obtain the second part; the results of the above two parts are added together to obtain the comprehensive deviation intensity value.

[0016] Furthermore, the deviation intensity is graded and an adjustment suggestion is generated to form a control process of closed-loop feedback. The specific steps are as follows: the absolute value of the comprehensive deviation intensity value is compared with the deviation threshold. The deviation threshold includes the first-level deviation threshold and the second-level deviation threshold. A graded judgment is made according to the comparison result: if the comprehensive deviation intensity value is less than or equal to the first-level deviation threshold, it is marked as a safe zone; if the comprehensive deviation intensity value is greater than the first-level deviation threshold and less than the second-level deviation threshold, it is marked as a mild deviation zone, and a recommended adjustment plan is generated: fine-tune the irradiation direction and readjust the incident angle; increase or decrease the current position according to the actual situation of the patient. Tumor dose; if the comprehensive deviation intensity value is greater than or equal to the secondary deviation threshold, it is marked as a severe deviation area, and a recommendation to suspend irradiation is issued, and the clinician is prompted to re-evaluate the treatment plan; based on the grading results and recommended strategies, an irradiation control parameter adjustment recommendation table is automatically generated and sent to the clinician; the clinician adjusts the irradiation control parameters according to the adjustment recommendation table, and recalculates the comprehensive deviation intensity value, compares the comprehensive deviation intensity value with the deviation threshold, and verifies whether the adjustment achieves the expected control effect; if the deviation control target is still not met, it automatically enters the next round of regulation and continues to optimize the parameters until the comprehensive deviation intensity value is in the safe zone.

[0017] Furthermore, the specific steps for focusing on the statistical deviation distribution characteristics of deep estimation points in the target area are as follows: after completing the output of the final three-dimensional dose map and the calculation of the comprehensive deviation intensity value, the deep estimation points located deep in the target area are screened out, and their corresponding tumor dose assessment values, tissue CT values, estimation point coordinates and beam source coordinates are extracted; for these deep estimation points, their comprehensive deviation intensity values ​​are aggregated, and combined with the tissue CT value distribution and spatial depth information, the deviation characteristics under each CT interval and distance level are statistically analyzed to establish a mapping relationship between deviation behavior and tissue structure properties.

[0018] Furthermore, the specific steps for generating structured correction data for subsequent reconstruction input and risk prompt generation are as follows: utilizing the mapping relationship between deviation behavior and tissue structure attributes, constructing a spatial correction factor based on the existing tumor dose assessment value and the comprehensive deviation intensity value, which is used to compensate for the error of the tumor dose assessment value of the deep estimation point; the spatial correction factor is used as an incremental input to participate in the subsequent reconstruction process of the final three-dimensional dose map; the tissue CT value, detection point coordinates, beam source coordinates, photon beam energy and comprehensive deviation intensity value in the deep area are reorganized as training samples to further optimize the weight configuration of the spatial dose reconstruction network; the reconstructed final three-dimensional dose map is named as the three-dimensional dose map multi-level dose distribution visualization The internal space grid is divided with the tumor target area as the boundary, and the mean value of the comprehensive deviation intensity value and the difference between the comprehensive deviation intensity values ​​before and after correction in each grid are marked to generate a target area response stability distribution map, which intuitively reflects the response differences of various areas of the target area to changes in irradiation parameters; the spatial correction factors generated in the deep area of ​​the current patient are bound to the patient's preoperative CT image to form structured index data for subsequent loading and calling before treatment; if the correction area is close to the reference surface dose corresponding area and the boundary of the sensitive anatomical structure, the restricted enhancement recommendation mark is automatically added when generating the irradiation control parameter adjustment recommendation table, which is used to assist doctors in making more information-based judgments when making dose enhancement or compensation decisions.

[0019] The second aspect of the present invention provides a tumor dose monitoring system based on radiotherapy, including: a tumor dose data acquisition and preprocessing module, a dose assessment and three-dimensional mapping module, a deviation identification and closed-loop feedback module, and a deep dose error backtracking module, wherein: the tumor dose data acquisition and preprocessing module is used to use an implantable fiber optic dose detector to collect tumor dose data, and perform data cleaning and normalization on the tumor dose data; the dose assessment and three-dimensional mapping module is used to perform preliminary dose estimation and physical modeling compensation on the preprocessed tumor dose data, and construct a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution; the deviation identification and closed-loop feedback module is used to analyze the deviation intensity based on the comparison result of the tumor dose assessment value and the reference surface dose, perform graded judgment on the deviation intensity and generate adjustment suggestions to form a control process for regulating the closed-loop feedback; the deep dose error backtracking module is used to focus on the statistical deviation distribution characteristics of the deep estimation point of the target area, and generate structured correction data for subsequent reconstruction input and risk prompt generation.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This method and system for monitoring tumor dose based on radiotherapy performs preliminary dose estimation and physical modeling compensation on preprocessed tumor dose data, and constructs a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution, thereby forming a tumor dose assessment mechanism that integrates physics and observation, and improving the estimation accuracy of the dose in non-detected areas.

[0023] (2) This method and system for monitoring tumor dose based on radiotherapy, by inputting the patient's preoperative CT image, tumor dose data and intermediate dose reconstruction map into a three-dimensional convolutional neural network model, trains and generates a spatially continuous final three-dimensional dose map, thereby filling the blind spot of data collection and effectively improving the complete modeling capability of dose distribution in deep tissues.

[0024] (3) This method and system for monitoring tumor dose based on radiotherapy compares the estimated tumor dose with the reference surface dose, combines the tissue CT value and spatial position to build a deviation intensity quantification model, and calculates the comprehensive deviation intensity value. The system can automatically generate control parameter recommendations based on the deviation level and dynamically adjust the irradiation settings, thus realizing a dose control feedback closed loop.

[0025] (4) This method and system for monitoring tumor dose based on radiotherapy generates spatial correction factors by combining the distribution characteristics of tissue CT values ​​and spatial position statistical deviations at deep estimation points in the target area and uses them for incremental training of the neural network. The correction factors are bound to preoperative CT images to construct an individualized correction template, thereby improving the ability to identify and adaptively correct deep dose errors and achieving precise control of the system's tumor dose for patients.

[0026] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 The following is a flow chart of a tumor dose monitoring method based on radiotherapy;

[0028] Figure 2 This is a structural diagram of a tumor dose monitoring system based on radiotherapy;

[0029] Figure 3 It is a three-dimensional dose map and a multi-level dose distribution visualization diagram;

[0030] Figure 4 is the target area response stability distribution diagram. DETAILED DESCRIPTION

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0032] See also Figures 1-4 The embodiment of the present invention provides a technical solution: a tumor dose monitoring method and system based on radiotherapy, comprising the following steps: S1, using an implantable fiber optic dose detector to collect tumor dose data, and performing data cleaning and normalization processing on the tumor dose data; S2, performing preliminary dose estimation and physical modeling compensation on the preprocessed tumor dose data, and constructing a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution; S3, analyzing the deviation intensity based on the comparison result between the tumor dose assessment value and the reference body surface dose, performing graded judgment on the deviation intensity and generating adjustment suggestions, forming a control process of closed-loop feedback; S4, focusing on the statistical deviation distribution characteristics of the deep estimation point of the target area, and generating structured correction data for subsequent reconstruction input and risk prompt generation.

[0033] Specifically, the specific steps of using an implantable fiber optic dose detector to collect tumor dose data are as follows: an implantable fiber optic dose detector is used, which is placed in the tumor target area and its periphery, and the dose variation range of the target area is ensured to be covered by precise positioning, and the radiation-induced light signal is sensed in real time to collect tumor dose data; the tumor dose data includes instantaneous radiation dose, detection point coordinates, beam source coordinates, estimated point coordinates, photon beam energy, tissue CT value and reference surface dose; wherein, the radiation-excited fluorescence signal is converted into a digital electrical signal through photoelectric conversion, signal amplification and analog-to-digital conversion to obtain the instantaneous radiation dose of each detection point; the spatial position of the fiber optic probe is calibrated through preoperative CT scanning, the patient coordinate system is constructed, and the rigid registration method is used to compare the CT structure. The system obtains structural information, acquires the coordinates of the detection point, beam source and estimated point in the patient's body, and realizes the matching of dose acquisition and spatial position; obtains the photon beam energy of the radiation used for treatment through the control system of the radiotherapy equipment, which is used as the key physical parameter for simulating radiation penetration and energy deposition in physical modeling; reads the tissue CT value of the tissue where the corresponding detection point is located in the patient's preoperative CT image, which reflects the tissue density characteristics; connects with the control system of the radiotherapy accelerator, reads the tissue angle of the current beam relative to the patient's coordinate system in real time, and provides a spatial reference for the subsequent dose direction factor; obtains the reference surface dose through the TLD detector attached to the patient's body surface, which serves as a benchmark indicator for dose deviation evaluation and control calibration, and assists in constructing a comprehensive deviation intensity value function.

[0034] In this implementation, an implantable fiber optic dose detector is used to collect real-time dose data at multiple points and deep locations in the tumor target area and its surroundings, breaking through the limitations of traditional surface monitoring in spatial depth. Preoperative CT registration is used to obtain the spatial coordinates of the detection point, beam source, and estimation point, enabling precise matching of dose data with the patient's anatomical structure to ensure data consistency and integrity. At the same time, key physical information such as photon beam energy, tissue CT value, and irradiation angle are introduced to provide a physical basis for subsequent dose modeling. Combined with surface reference dose data, a fusion foundation is laid for deviation analysis and closed-loop control, thereby enhancing the system's ability to restore the true dose distribution in the body and the precision control level of individualized radiotherapy.

[0035] Specifically, the steps for data cleaning and normalization of tumor dose data are as follows: using the sliding window method combined with the triple standard difference constant detection algorithm, local statistical analysis of the dose changes of each detection point in a continuous time period is performed, tumor dose data that significantly deviates from the normal range is eliminated, and the interference of transient noise and abnormal jumps on model training is reduced; using millisecond-level time alignment and linear interpolation methods, the instantaneous missing tumor dose data caused by detection delays, communication interference and other factors during the acquisition process is supplemented to maintain the integrity and continuity of the time series data; using the rigid registration algorithm to align the detector spatial coordinates with the patient's preoperative CT image coordinates, and by constructing a unified three-dimensional spatial mapping relationship, ensure the one-to-one correspondence between all tumor dose data and tissue structure information; using the Kalman filter algorithm to perform time series modeling and denoising on the continuously acquired instantaneous radiation dose, reduce high-frequency fluctuation interference, and enhance the dynamic stability of the data; using unit conversion and maximum and minimum normalization algorithms to normalize the tumor dose data, unify the values ​​of different sources and different scales into a standard dimension range, and improve the model calculation efficiency and data comparability.

[0036] In this implementation, the quality and consistency of tumor dose data are improved by introducing multiple data cleaning and normalization strategies: a sliding window and outlier detection algorithm are used to effectively remove local noise and outliers, ensuring stable and reliable data; a time alignment and interpolation completion mechanism solves the problem of short-term data loss during the acquisition process, ensuring the continuity of the time series; a rigid registration algorithm achieves a one-to-one mapping of dose data and CT space, providing a spatial foundation for subsequent 3D modeling; a Kalman filter enhances the smoothness of instantaneous dose values ​​and suppresses high-frequency fluctuations; and normalization processing unifies the data dimension, optimizing the computational efficiency of subsequent model training and fusion input. These processing methods jointly construct high-quality, structured tumor dose data, providing a solid data foundation for the system's subsequent 3D reconstruction and deviation identification.

[0037] Specifically, the specific steps for preliminary dose estimation and physical modeling compensation of preprocessed tumor dose data are as follows: take N detection points, calculate the square of the coordinate difference between the estimated point and the i-th detection point in the transverse direction, the square of the coordinate difference in the longitudinal direction, and the square of the coordinate difference in the vertical direction, add a minimum term, and finally square the result to obtain the spatial distance between the i-th detection point and the estimated point in the patient coordinate system; multiply the spatial distance by the exponential decay factor and take the negative value as the exponent, and then calculate the value of this power function with the natural constant as the base to obtain the weight of the current detection point; wherein, the exponential decay factor is obtained by minimizing the mean square error between the spatial interpolation dose assessment value and the instantaneous radiation dose in historical cases, and is fitted and optimized on multiple patient data to reflect the optimal attenuation rate of the influence of spatial distance on dose attenuation; multiply the weight by the instantaneous radiation dose as the weighted value of the point, repeat the above process to sum the N detection points, and use the sum result as the numerator; only weight the N detection points The power function is calculated repeatedly to obtain the weight of the current estimated point, which is not multiplied by the instantaneous radiation dose, and the sum is used as the denominator; the numerator is divided by the denominator to obtain the spatial interpolation dose assessment value, completing the preliminary dose estimation for the area without detector deployment; the photon beam energy, irradiation angle, detection point coordinates, beam source coordinates, and tissue CT value are obtained, and the photon beam energy, the cosine value of the tissue angle, and the tissue CT value are multiplied. The product of the three is divided by the square of the spatial distance between the estimated point coordinates and the beam source coordinates, and a minimum term is added. The sum is then multiplied by the physical weight to obtain the first part; the instantaneous radiation dose is multiplied by the measured weight to obtain the second part; the sum of the two parts is calculated to obtain the physical modeling dose assessment value, and the dose value of each estimated point in three-dimensional space is physically modeled and compensated; among them, based on the minimum mean square error fitting algorithm of historical patient tumor cases, the measured weight and physical weight are repeatedly trained and optimized between the known instantaneous radiation dose distribution and the three-dimensional dose distribution output, and both have a value range of [0,1].

[0038] Among them, the specific calculation formula of the spatial interpolation dose assessment value is:

[0039]

[0040] Where, K (x,y,z) Denotes the spatial interpolation dose assessment value, D i Indicates the instantaneous radiation dose, N indicates the total number of detection points used for interpolation calculation, x, y, z indicate the coordinates of the current estimated point, x i ,y i ,z i Represents the detection point coordinates of the i-th detection point, k represents the exponential decay factor, and ∈ represents the minimum term.

[0041] The specific calculation formula for the physical modeling dose assessment value is:

[0042]

[0043] Where W (x,y,z) represents the dose evaluation value of the physical model, E represents the photon beam energy, θ represents the tissue angle, x, y, z represent the detection point coordinates of the current estimation point, x0, y0, z0 represent the beam source coordinates, H represents the tissue CT value of the estimation point, D s It represents the instantaneous radiation dose obtained by weighting adjacent detector points, α represents the physical weight, β represents the measured weight, and ∈ represents the minimum term.

[0044] In this implementation, the accuracy of dose restoration for areas without detectors is effectively improved by performing spatial interpolation estimation and physical modeling compensation on the preprocessed tumor dose data: first, a weight model is constructed using spatial distance and exponential decay function to achieve continuous dose completion of the estimated points, avoiding spatial gaps caused by insufficient detection coverage; second, a dose compensation model based on radiation transmission characteristics is constructed by combining key physical parameters such as photon beam energy, irradiation angle, and tissue CT value to further enhance the dose distribution's responsiveness to anatomical structure and incident direction; the dual-path fusion of spatial interpolation and physical modeling significantly improves the system's accuracy in characterizing complex dose fields in the body.

[0045] Specifically, the specific steps of constructing a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution are as follows: weighted fusion of spatial interpolation dose assessment values ​​and physical modeling dose assessment values, and integration of the two types of estimated information according to the set fusion ratio to obtain a tumor dose assessment value as a unified representation compatible with structural and physical characteristics; using the tumor dose assessment value to generate an intermediate dose reconstruction map driven by both physical and observational data, providing a dose input basis with spatial continuity and preliminary credibility for network training; using the patient's preoperative CT image, tumor dose data and intermediate dose reconstruction map as joint input to construct a three-dimensional convolutional neural network model, and conduct training and prediction on a preset sample set Through the deep feature extraction capability of the network, the dose information of the area without detection points is completed, the ability to restore spatial details and boundary transition areas is improved, and the overall dose distribution is further refined; the network prediction results are weightedly fused with the intermediate dose reconstruction map, making full use of the detail advantages of AI prediction and the global continuity of the original modeling map to generate a unified final three-dimensional dose map, ensuring local accuracy and global consistency; the rigid registration algorithm is used to align the coordinates of the final three-dimensional dose map with the patient's preoperative CT image, establish the correspondence between the image space and the dose space, realize the precise matching of the spatial image and the dose field, and provide structured input for subsequent deviation analysis and feedback control.

[0046] In this implementation, an intermediate dose reconstruction map is constructed by fusing the spatial interpolation dose assessment value with the physical modeling dose assessment value, providing a structurally complete and physically credible training basis for the three-dimensional convolutional neural network; the patient's preoperative CT image and tumor dose data are jointly input to enable the network to have multi-dimensional perception capabilities of spatial structure and tissue characteristics, thereby achieving accurate completion of the dose in areas where no detection points are deployed and refined optimization of the overall distribution; finally, a high-resolution three-dimensional dose map is output by fusing the AI ​​prediction results with the intermediate map, and rigid registration is used to align it with the preoperative image to achieve spatial matching of the dose field and the patient's anatomical structure, thereby significantly improving the integrity and accuracy of the system's dose restoration in deep and sparse areas.

[0047] Specifically, the specific steps for analyzing the deviation intensity based on the comparison results of the tumor dose assessment value and the reference surface dose are as follows: the tumor dose assessment value of each estimated point in the final three-dimensional dose map is compared with the reference surface dose one by one to ensure that the two are in a unified spatial coordinate system, and the comprehensive deviation intensity value of each estimated point is calculated to quantify the dose deviation risk of the point; the tumor dose assessment value of the estimated point is subtracted from the reference surface dose and divided by the reference surface dose to express the relative dose deviation of the point, which can reflect the degree of deviation of the dose output from the standard level; then, the relative dose deviation is multiplied by the tissue CT value of the estimated point as the first part to introduce the influence of tissue density. factor to enhance the sensitivity of the deviation value to the response characteristics of different tissues; the sum of the squares of the coordinate differences between the estimation point and the beam source in the lateral, longitudinal and vertical directions in the spatial coordinates is calculated to reflect the relative spatial relationship between its position and the radiation path, and the result is divided by the square of the spatial scale normalization constant to make different depth positions comparable, thereby obtaining the second part; among them, the spatial scale normalization constant is obtained by calculating the average spatial distance between the tumor target estimation point and the beam source in historical patients, and is used to unify the position difference measurement standard under different treatment scenarios; the results of the above two parts are added together to obtain the comprehensive deviation intensity value, which provides a structured numerical basis for subsequent deviation grading and response control.

[0048] Among them, the specific calculation formula of the comprehensive deviation intensity value is:

[0049]

[0050] Where, P (x,y,z) Indicates the comprehensive deviation strength value of the estimated point, D f(x,y,z) represents the tumor dose assessment value, D r represents the reference surface dose, H represents the tissue CT value of the estimation point, x, y, z represent the detection point coordinates of the estimation point, x0, y0, z0 represent the beam source coordinates, and L represents the spatial scale normalization constant.

[0051] In this implementation plan, a comprehensive deviation intensity value is constructed to systematically quantify the degree of deviation between the tumor dose assessment value and the reference surface dose. The tissue CT value and spatial depth information are integrated so that the deviation assessment not only reflects the absolute dose difference, but also takes into account the combined effects of tissue density and position factors, thereby enhancing the tissue sensitivity and spatial resolution of deviation identification. This deviation intensity indicator provides an accurate quantitative basis for subsequent high-risk area identification, dose grading judgment, and automatic control strategy generation, significantly improving the system's perception of abnormal dose behavior and the reliability of the control response.

[0052] Specifically, the following steps are used to grade the deviation intensity and generate adjustment recommendations to form a closed-loop feedback control process: the absolute value of the comprehensive deviation intensity value is compared with the deviation threshold, which includes a first-level deviation threshold and a second-level deviation threshold. A graded judgment is performed based on the comparison result: if the comprehensive deviation intensity value is less than or equal to the first-level deviation threshold, it is marked as a safe zone; if the comprehensive deviation intensity value is greater than the first-level deviation threshold but less than the second-level deviation threshold, it is marked as a mild deviation zone, and a recommended adjustment plan is generated: fine-tune the irradiation direction and readjust the incident angle; increase or decrease the tumor dose at the current location based on the patient's actual situation; if the comprehensive deviation intensity value is greater than or equal to the second-level deviation threshold, it is marked as a severe deviation zone, and a recommendation to pause irradiation is issued, prompting the clinician to re-evaluate the treatment plan; based on the grading results and recommended strategy, an irradiation control parameter adjustment recommendation table is automatically generated and sent to the clinician; the clinician adjusts the irradiation control parameters according to the adjustment recommendation table, recalculates the comprehensive deviation intensity value, and compares the comprehensive deviation intensity value with the deviation threshold to verify whether the adjustment achieves the expected control effect; if the deviation control target is still not met, the next round of regulation is automatically entered, and the parameters are continuously optimized until the comprehensive deviation intensity value is in the safe zone.

[0053] In this implementation plan, a hierarchical judgment mechanism based on the comprehensive deviation intensity value is established to achieve accurate identification and response classification of dose deviations of different degrees; combined with the set deviation threshold, corresponding adjustment strategies are formulated for the safe zone, mild deviation zone and severe deviation zone respectively, and an irradiation control parameter adjustment recommendation table is automatically generated and the clinician decision-making link is introduced to construct a human-machine collaborative control feedback process; the system supports verification of the adjustment effect and multiple rounds of iterative optimization, effectively improving the ability to control abnormal dose behavior during radiotherapy and the individualized adaptability of the treatment plan.

[0054] Specifically, the specific steps for focusing on the statistical deviation distribution characteristics of deep estimation points in the target area are as follows: after completing the output of the final three-dimensional dose map and the calculation of the comprehensive deviation intensity value, based on the coordinate depth threshold and the target area boundary conditions, the deep estimation points located deep in the target area and far away from the body surface are screened out, and the dose distribution accuracy and potential deviation risks in these areas are focused on; the corresponding tumor dose assessment values, tissue CT values, estimation point coordinates and beam source coordinates are extracted to ensure that the data comes from a unified coordinate system and retain the spatial directionality of the structural information; for these deep estimation points, their comprehensive deviation intensity values ​​are aggregated, and by summarizing the deviation magnitudes of multiple points, a structurally consistent deep deviation data set is formed, and combined with the tissue CT value distribution and spatial depth information, a multidimensional parameter matrix is ​​established for classification analysis; the deviation characteristics under each CT interval and distance level are statistically analyzed to identify the common error types and spatial distribution trends in high-density and low-density tissues, and finally a mapping relationship between deviation behavior and tissue structure properties is established to provide a model basis and data support for the subsequent correction factor construction.

[0055] In this implementation plan, by performing aggregate analysis on the deviation intensity of deep estimation points in the target area, combined with tissue CT values ​​and spatial depth information, the deviation distribution patterns under different tissue densities and spatial levels are systematically explored to make up for the problem that traditional dose monitoring has insufficient ability to identify deep areas; by constructing a mapping relationship between deviation behavior and tissue structure properties, a data foundation is laid for establishing a structure-dependent error compensation mechanism, which helps to improve the system's dose estimation accuracy in complex tissue environments and its ability to perceive and model deep dose anomalies.

[0056] Specifically, the specific steps for generating structured correction data for subsequent reconstruction input and risk prompt generation are as follows: using the mapping relationship between deviation behavior and tissue structure attributes, a spatial correction factor is constructed based on the existing tumor dose assessment value and the comprehensive deviation intensity value to compensate for the error of the tumor dose assessment value of the deep estimation point; the spatial correction factor is used as an incremental input to participate in the subsequent reconstruction process of the final three-dimensional dose map; the tissue CT value, detection point coordinates, beam source coordinates, photon beam energy and comprehensive deviation intensity value in the deep area are reorganized as training samples to further optimize the weight configuration of the spatial dose reconstruction network; the reconstructed final three-dimensional dose map is named as the three-dimensional dose map multi-level dose distribution visualization The internal space grid is divided with the tumor target area as the boundary, and the mean value of the comprehensive deviation intensity value and the difference between the comprehensive deviation intensity values ​​before and after correction in each grid are marked to generate a target area response stability distribution map, which intuitively reflects the response differences of various areas of the target area to changes in irradiation parameters; the spatial correction factors generated in the deep area of ​​the current patient are bound to the patient's preoperative CT image to form structured index data for subsequent loading and calling before treatment; if the correction area is close to the reference surface dose corresponding area and the boundary of the sensitive anatomical structure, the restricted enhancement recommendation mark is automatically added when generating the irradiation control parameter adjustment recommendation table, which is used to assist doctors in making more information-based judgments when making dose enhancement or compensation decisions.

[0057] In this case, the data table of the three-dimensional dose map multi-level dose distribution visualization diagram in Table 1 records the tumor dose evaluation of five estimated samples under different spatial positions and tissue density conditions. Each sample contains the estimated point coordinates (x, y, z), the tissue CT value, and the corresponding calculated tumor dose evaluation value, and the deviation level is divided accordingly. The estimated point coordinates of sample 1 are (0.1, 0.15, 0.05), the tissue CT value is 55, and the corresponding tumor dose evaluation value is 1.20, which belongs to the severe deviation area; sample 2 is located at (-0.12, -0.08, -0.1), the tissue CT value is -40, and the dose evaluation value is 0.90, which is judged to be a mild deviation area;

[0058] The estimated point coordinates of sample 3 are (0.054, 0.1, 0.12), the tissue CT value is 15, and the corresponding dose assessment value is 0.65, which is also classified as the mild deviation area; sample 4 is located at (-0.07, -0.1, 0.0), the tissue CT value is -80, and the tumor dose assessment value is 0.45, which is judged to be in the safe zone; the estimated point coordinates of sample 5 are (0.02, 0.0, -0.15), the tissue CT value is 90, and the tumor dose assessment value is 0.80, which also falls in the mild deviation area.

[0059] Table 1 Data table of three-dimensional dose map and multi-level dose distribution visualization diagram

[0060]

[0061]

[0062] like Figure 3 The figure below shows a multi-level visualization of the three-dimensional dose distribution. Purple indicates areas of severe deviation; orange indicates areas of mild deviation; and yellow indicates safe areas, which may be in the dose transition or marginal area. This graded visualization method helps doctors intuitively determine whether the dose covers the target area and whether it overflows into risk areas. It can also be used for subsequent deviation analysis and adjustment of control strategies.

[0063] In addition, as described in the target area response stability distribution diagram data table in Table 2, in the embodiment, the dose response of five target area samples under different spatial positions and tissue density conditions is recorded, and each sample includes the estimated point coordinates, tumor dose assessment value, reference surface dose, tissue CT value and comprehensive deviation intensity value. The estimated point coordinates of sample 1 are (0.10, -0.10, 0.20), the tumor dose assessment value is 1.15, the reference surface dose is 1.0, the tissue CT value is 50, and the comprehensive deviation intensity value is 0.78; the estimated point coordinates of sample 2 are (-0.40, 0.30, -0.30), in the low-density tissue with a tissue CT value of -80, the tumor dose assessment value is 0.85, the reference surface dose is 0.9, and the comprehensive deviation intensity value is 0.62; sample 3 is located at (0.35, -0.25, 0.10), and the tissue CT value is 20 , the tumor dose assessment value is 1.05, the reference surface dose is 1.0, and the comprehensive deviation intensity value is 0.45; the estimated point coordinates of sample 4 are (-0.20, 0.15, -0.05), the tissue CT value is -30, the corresponding tumor dose assessment value is 0.65, the reference surface dose is 0.7, and the comprehensive deviation intensity value is 0.35; the estimated point coordinates of sample 5 are (0.05, 0.00, 0.30), the tissue CT value is 100, the tumor dose assessment value is 1.25, the reference surface dose is 1.0, and the comprehensive deviation intensity value is 0.81.

[0064] Table 2 Target area response stability distribution data table

[0065] Target area samples x y z <![CDATA[D f(x,y,z) ]]> <![CDATA[D r ]]> H <![CDATA[P (x,y,z) <!-- 9 -->]]> 1 0.1 -0.1 0.2 1.15 1.0 50 0.78 2 -0.4 0.3 -0.3 0.85 0.9 -80 0.62 3 0.35 -0.25 0.1 1.05 1.0 20 0.45 4 -0.2 0.15 -0.05 0.65 0.7 -30 0.35 5 0.05 0.0 0.3 1.25 1.0 100 0.81

[0066] like Figure 4 The figure below shows the target response stability distribution map. Each dot represents a spatial grid cell within the tumor target volume. Colors closer to yellow indicate greater differences in response before and after dose adjustment, indicating lower stability. Darker purple indicates a more stable response. This map can be used to analyze areas within the target volume that are sensitive to or fluctuate dramatically with irradiation parameters, assisting physicians in identifying potential unstable areas or blind spots in dose control.

[0067] In this implementation scheme, by constructing a spatial correction factor based on the mapping relationship between deviation behavior and tissue structure, directional error compensation of tumor dose estimation values ​​in deep areas of the target area is achieved, and it is included as an incremental input in the reconstruction process of the final three-dimensional dose map, significantly improving the accuracy of tumor dose estimation in deep areas; at the same time, the tissue CT value, estimated point coordinates and photon beam energy are integrated as training samples to optimize the weight configuration of the spatial dose reconstruction network and enhance the model's adaptability to complex tissue structures; by constructing a three-dimensional dose map multi-level dose distribution visualization map and a target area response stability distribution map, the sensitivity of different areas to changes in irradiation parameters is intuitively presented; the spatial correction factor is bound to the patient's preoperative CT image to form structured index data, which provides support for rapid loading and retrospective calls before treatment; and a restricted enhancement recommendation mark is automatically added near the risk area to assist doctors in making more robust decisions based on data when performing dose enhancement or compensation operations, thereby improving the system's individualized regulation capabilities and clinical safety.

[0068] like Figure 2 As shown, the second aspect of the present invention provides a tumor dose monitoring system based on radiotherapy, including: a tumor dose data acquisition and preprocessing module, a dose assessment and three-dimensional mapping module, a deviation identification and closed-loop feedback module, and a deep dose error backtracking module, wherein: the tumor dose data acquisition and preprocessing module is used to use an implantable fiber optic dose detector to collect tumor dose data, and perform data cleaning and normalization on the tumor dose data; the dose assessment and three-dimensional mapping module is used to perform preliminary dose estimation and physical modeling compensation on the preprocessed tumor dose data, and construct a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution; the deviation identification and closed-loop feedback module is used to analyze the deviation intensity according to the comparison result of the tumor dose assessment value and the reference surface dose, perform graded judgment on the deviation intensity and generate adjustment suggestions, forming a control process for regulating the closed-loop feedback; the deep dose error backtracking module is used to focus on the statistical deviation distribution characteristics of the deep estimation point of the target area, and generate structured correction data for subsequent reconstruction input and risk prompt generation.

[0069] In this implementation plan, by constructing a four-level structured functional module covering data acquisition, dose modeling, deviation identification and deep backtracking, a complete process from deep multi-point real-time acquisition, three-dimensional dose fine reconstruction, to intelligent deviation identification and feedback control is realized; the system not only improves the ability to restore the true dose distribution in the tumor area, but also has the ability to dynamically respond to high-risk areas and adaptively correct deep errors, forming a data-driven, structurally correlated, closed-loop controlled individualized radiotherapy monitoring and regulation mechanism, which significantly enhances the accuracy, safety and intelligence level of tumor radiotherapy.

[0070] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0071] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for monitoring tumor dose based on radiotherapy, characterized in that: The following steps are involved: S1, using an implantable fiber optic dose detector to collect tumor dose data, and then perform data cleaning and normalization on the tumor dose data; S2, performs preliminary dose estimation and physical modeling compensation on the preprocessed tumor dose data, and constructs a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution; The specific steps of performing preliminary dose estimation and physical modeling compensation on the pre-processed tumor dose data are as follows: Take N detection points, calculate the square of the coordinate difference between the estimated point and the i-th detection point in the horizontal direction, the square of the coordinate difference in the longitudinal direction, and the square of the coordinate difference in the vertical direction, add a minimum term, and finally square the result to obtain the spatial distance between the i-th detection point and the estimated point in the patient coordinate system; multiply the spatial distance by the exponential attenuation factor and take the negative value as the exponent, and then calculate the value of this power function with the natural constant as the base to obtain the weight of the current detection point; multiply the weight by the instantaneous radiation dose as the weighted value of the point, repeat the above process to sum the N detection points, and use the sum result as the numerator; only repeatedly calculate the power function for the N detection points to obtain the weight of the current estimation point, without multiplying it by the instantaneous radiation dose, and use the sum result as the denominator; divide the numerator by the denominator to obtain the spatial interpolation dose assessment value, and complete the preliminary dose estimation for the area where no detectors are deployed; Obtain the photon beam energy, irradiation angle, detection point coordinates, beam source coordinates, and tissue CT value. Multiply the photon beam energy, the cosine value of the tissue angle, and the tissue CT value. Divide the product by the square of the spatial distance between the estimated point coordinates and the beam source coordinates, add a minimum term, and then multiply the sum by the physical weight to obtain the first part. Multiply the instantaneous radiation dose by the measured weight to obtain the second part. Calculate the sum of the two parts to obtain the physical modeling dose assessment value, and perform physical modeling compensation on the dose value of each estimated point in three-dimensional space. S3: Analyze the deviation intensity based on the comparison results of the tumor dose assessment value and the reference body surface dose, grade the deviation intensity and generate adjustment suggestions, forming a control process of closed-loop feedback; S4, focusing on the statistical deviation distribution characteristics of the target area deep estimation points, generates structured correction data for subsequent reconstruction input and risk prompt generation.

2. The method for monitoring tumor dose based on radiotherapy according to claim 1, characterized in that: The specific steps of collecting tumor dose data using the implantable fiber optic dose detector are as follows: An implantable fiber optic dose detector is used and placed in the tumor target area and its periphery to sense radiation-induced light signals in real time and collect tumor dose data; the tumor dose data includes instantaneous radiation dose, detection point coordinates, beam source coordinates, estimated point coordinates, photon beam energy, tissue CT value and reference surface dose; the instantaneous radiation dose of each detection point is obtained through photoelectric conversion, signal amplification and analog-to-digital conversion; the spatial position of the fiber optic probe is calibrated through preoperative CT scanning, the patient coordinate system is constructed, and the detection point coordinates, beam source coordinates and estimated point coordinates in the patient's body are obtained; the photon beam energy of the radiation used for treatment is obtained through the radiotherapy equipment control system; the tissue CT value of the tissue corresponding to the detection point is read from the patient's preoperative CT image; the tissue angle of the current beam relative to the patient coordinate system is read in real time by connecting to the control system of the radiotherapy accelerator; and the reference surface dose is obtained through the TLD detector attached to the patient's body surface.

3. The method for monitoring tumor dose based on radiotherapy according to claim 1, characterized in that: The specific steps of data cleaning and normalization processing of tumor dose data are as follows: A sliding window method combined with a three-fold standard deviation constant detection algorithm was used to eliminate tumor dose data that significantly deviated from the normal range. Millisecond-level time alignment and linear interpolation methods were used to supplement instantaneously missing tumor dose data. A rigid registration algorithm was used to align the detector spatial coordinates with the patient's preoperative CT image coordinates to ensure three-dimensional spatial consistency. A Kalman filter algorithm was used to denoise the continuously collected instantaneous radiation dose. Unit conversion and maximum and minimum normalization algorithms were used to normalize the tumor dose data to unify the data scale.

4. The method for monitoring tumor dose based on radiotherapy according to claim 1, characterized in that: The specific steps of constructing a three-dimensional convolutional neural network model to refine the three-dimensional dose distribution are as follows: The spatial interpolation dose assessment value and the physical modeling dose assessment value are weightedly fused to obtain the tumor dose assessment value, and the tumor dose assessment value is used to generate an intermediate dose reconstruction map driven by both physical and observation data. The patient's preoperative CT image, tumor dose data and intermediate dose reconstruction map are used as joint input to construct a three-dimensional convolutional neural network model and carry out training and prediction to complete the dose information of the area where no detection points are deployed and refine the overall dose distribution. The network prediction results are weightedly fused with the intermediate dose reconstruction map to generate a unified final three-dimensional dose map. The final three-dimensional dose map is aligned with the patient's preoperative CT image using a rigid registration algorithm to achieve precise matching of the spatial image and the dose field.

5. The method for monitoring tumor dose based on radiotherapy according to claim 1, characterized in that: The specific steps of analyzing the deviation intensity based on the comparison results of the tumor dose assessment value and the reference body surface dose are as follows: The tumor dose assessment value at each estimated point in the final 3D dose map was compared with the reference surface dose one by one, and the comprehensive deviation intensity value of each estimated point was calculated. The tumor dose assessment value at the estimated point was subtracted from the reference surface dose and then divided by the reference surface dose to express the relative dose deviation. The relative dose deviation was then multiplied by the tissue CT value at the estimated point as the first part. Calculate the sum of the squares of the differences in the horizontal, longitudinal, and vertical coordinates between the estimated point and the beam source in the spatial coordinates, and divide the result by the square of the spatial scale normalization constant to obtain the second part; add the results of the above two parts to obtain the comprehensive deviation intensity value.

6. The method for monitoring tumor dose based on radiotherapy according to claim 1, characterized in that: The specific steps of the control process for determining the deviation intensity by grade and generating adjustment suggestions to form a closed-loop feedback control are as follows: The absolute value of the comprehensive deviation intensity value is compared with the deviation threshold. The deviation threshold includes the first-level deviation threshold and the second-level deviation threshold. A graded judgment is made based on the comparison result: if the comprehensive deviation intensity value is less than or equal to the first-level deviation threshold, it is marked as a safe zone; If the comprehensive deviation intensity value is greater than the first-level deviation threshold and less than the second-level deviation threshold, it is marked as a mild deviation area, and a recommended adjustment plan is generated: fine-tune the irradiation direction and readjust the incident angle; increase or decrease the tumor dose at the current location based on the patient's actual situation; If the comprehensive deviation intensity value is greater than or equal to the second-level deviation threshold, it is marked as a severe deviation area, and a recommendation to suspend irradiation is issued, prompting the clinician to re-evaluate the treatment plan; Based on the grading results and recommended strategies, an irradiation control parameter adjustment recommendation table is automatically generated and sent to clinicians; clinicians adjust the irradiation control parameters according to the adjustment recommendation table, and recalculate the comprehensive deviation intensity value, compare the comprehensive deviation intensity value with the deviation threshold, and verify whether the adjustment achieves the expected control effect; if the deviation control target is still not met, it automatically enters the next round of regulation and control, and continues to optimize the parameters until the comprehensive deviation intensity value is in the safe zone.

7. The method for monitoring tumor dose based on radiotherapy according to claim 1, characterized in that: The specific steps of focusing on the statistical deviation distribution characteristics of the deep estimation points of the target area are as follows: After completing the output of the final three-dimensional dose map and the calculation of the comprehensive deviation intensity value, the deep estimation points located deep in the target area are screened out, and their corresponding tumor dose assessment values, tissue CT values, estimation point coordinates, and beam source coordinates are extracted; for these deep estimation points, their comprehensive deviation intensity values ​​are aggregated, and combined with the tissue CT value distribution and spatial depth information, the deviation characteristics under each CT interval and distance level are statistically analyzed to establish a mapping relationship between deviation behavior and tissue structure properties.

8. The method for monitoring tumor dose based on radiotherapy according to claim 1, characterized in that: The specific steps of generating structured correction data for subsequent reconstruction input and risk prompt generation are as follows: By utilizing the mapping relationship between deviation behavior and tissue structure properties, a spatial correction factor is constructed based on the existing tumor dose assessment value and the comprehensive deviation intensity value to compensate for the error of the tumor dose assessment value at the deep estimation point. The spatial correction factor is used as an incremental input in the subsequent reconstruction of the final 3D dose map. The tissue CT values, detection point coordinates, beam source coordinates, photon beam energy, and integrated deviation intensity values ​​in the deep region are reorganized as training samples to further optimize the weight configuration of the spatial dose reconstruction network. The final reconstructed 3D dose map is named a 3D dose map multi-level dose distribution visualization map. The internal space grid is divided with the tumor target area as the boundary. The mean value of the integrated deviation intensity value and the difference between the integrated deviation intensity values ​​before and after correction in each grid are annotated to generate a target area response stability distribution map, which intuitively reflects the response differences of various regions in the target area to changes in irradiation parameters. The spatial correction factor generated in the deep area of ​​the current patient is bound to the patient's preoperative CT image to form structured index data for loading and calling before subsequent treatment; if the correction area is close to the reference surface dose corresponding area and the boundary of sensitive anatomical structure, a restricted enhancement recommendation mark is automatically added when generating the irradiation control parameter adjustment recommendation table to assist doctors in making more informed judgments when making dose enhancement or compensation decisions.

9. A tumor dose monitoring system based on radiotherapy, comprising: Tumor dose data acquisition preprocessing module, dose assessment and 3D mapping module, deviation identification and closed-loop feedback module, deep dose error backtracking module, including: The tumor dose data acquisition and preprocessing module is used to collect tumor dose data using an implantable optical fiber dose detector, and perform data cleaning and normalization processing on the tumor dose data; The dose assessment and 3D mapping module is used to perform preliminary dose estimation and physical modeling compensation on preprocessed tumor dose data, and to construct a 3D convolutional neural network model to refine the 3D dose distribution; The deviation identification and closed-loop feedback module is used to analyze the deviation intensity based on the comparison results of the tumor dose assessment value and the reference surface dose, classify the deviation intensity and generate adjustment suggestions, forming a control process for regulating the closed-loop feedback; The deep dose error backtracking module is used to focus on the statistical deviation distribution characteristics of the deep estimation points in the target area and generate structured correction data for subsequent reconstruction input and risk prompt generation.

Images