A smart adaptive dose verification calculation method and system

By optimizing Monte Carlo calculations through a deep learning-based dose prediction model and intelligent adaptive simulation weight allocation, the problem of low efficiency in Monte Carlo dose calculations is solved, achieving efficient and accurate dose verification. This is applicable to different patients and treatment plans, and improves the quality control level of radiotherapy.

CN122075941APending Publication Date: 2026-05-26SUPERACCURACY SCIENCE & TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202512031915.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-30
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

Monte Carlo dosing calculation methods are inefficient, limiting their application in routine clinical validation. Furthermore, traditional variance reduction techniques rely on human experience and are difficult to adapt to different patient anatomy and treatment plans, resulting in limited acceleration effects or the introduction of biases.

Method used

A deep learning-based dose prediction model is used to obtain preliminary dose distribution data. The variance reduction parameter in the Monte Carlo simulation is dynamically allocated, and the calculation process is optimized through intelligent adaptive simulation weight allocation. This accelerates the convergence in the low-dose region and achieves intelligent rebalancing of resources.

Benefits of technology

It significantly improves the computational efficiency of Monte Carlo simulations, ensures high-precision dose verification results, achieves optimal global efficiency, requires no manual intervention, is applicable to different clinical cases, and maintains the accuracy and universality of the Monte Carlo method.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122075941A_ABST
    Figure CN122075941A_ABST
Patent Text Reader

Abstract

This invention discloses an intelligent adaptive dose verification calculation method and system. The intelligent adaptive dose verification calculation method includes: obtaining preliminary dose distribution data for a patient based on a deep learning-based dose prediction model; dynamically assigning variance reduction parameters to particles in a Monte Carlo simulation based on the high and low dose distribution information in the preliminary dose distribution data, wherein the variance reduction parameters include simulation weights, with simulation weights assigned to particles predicted to be in high-dose regions being less than those assigned to particles predicted to be in low-dose regions; and using a Monte Carlo dose calculation engine, simulating the planned beam based on the dynamically assigned variance reduction parameters to obtain high-precision verification dose distribution data. This invention's intelligent adaptive dose verification calculation method ensures the physical authenticity of the final verification dose distribution and the credibility of the "gold standard."
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of radiotherapy planning technology, specifically to an intelligent adaptive dose verification calculation method and system. Background Technology

[0002] Radiation therapy is one of the main methods of cancer treatment, aiming to kill tumor cells to the maximum extent while protecting surrounding normal tissues and organs at risk. To achieve this goal, a precise treatment plan must be developed before treatment begins. The core of this plan is accurately calculating the radiation dose distribution deposited in the patient's body, i.e., dose calculation.

[0003] Currently, the clinical dose calculation and validation process mainly comprises two levels: First, the Treatment Planning System (TPS) uses deterministic algorithms such as pencil-beam convolution and stacking algorithms to perform rapid dose calculations to generate preliminary treatment plans. These algorithms are efficient and can meet the needs of clinical interactive design, but their physical models have been simplified to varying degrees, resulting in inherent limitations in calculation accuracy in areas with complex tissue interfaces (such as the lung-soft tissue interface) or dose build-up regions.

[0004] To ensure the safety, effectiveness, and accuracy of the treatment plan, a dose calculation method independent of TPS and with a more accurate physical model must be used for verification before final approval. The Monte Carlo (MC) method calculates the dose distribution by simulating the random transport of a large number of particles (such as photons and electrons) in a medium. Because it can most realistically simulate the interaction between radiation and matter, it is recognized as the "gold standard" for dose calculation, and is especially suitable for verification processes with extremely high accuracy requirements.

[0005] However, the Monte Carlo method has a significant drawback: extremely low computational efficiency. To achieve an acceptable level of statistical noise, it is necessary to simulate massive amounts of particle history (hundreds of millions or even more), resulting in single calculations taking hours or even tens of hours, which severely hinders its application in routine clinical validation.

[0006] To address the computational inefficiency of Monte Carlo simulations, various acceleration techniques have been proposed, one important approach being "variance reduction." Traditional variance reduction techniques (such as geometric splitting and roulette wheel, forced collision, source bias, etc.) typically rely on the prior knowledge and manual experience of physicists or engineers to set global or regional parameters. The drawbacks of this approach are: 1) It is highly dependent on human experience, lacking universality and adaptability, requiring repeated parameter adjustments for different patient anatomy and treatment plans; 2) The parameter setting strategy is relatively fixed and coarse, making it difficult to precisely allocate resources based on the actual importance of dose distribution, potentially leading to limited acceleration or unexpectedly introducing biases in critical areas.

[0007] Therefore, there is an urgent need in this field for an innovative technical solution that can maintain the high-precision physical core of the Monte Carlo method to ensure the reliability of the verification results, while also intelligently and adaptively optimizing the calculation process to break through its computational efficiency bottleneck. This would enable high-precision Monte Carlo dose verification to be truly applied to routine clinical workflows, thereby improving the quality control level and safety of radiotherapy.

[0008] In view of this, this invention patent is hereby proposed. Summary of the Invention

[0009] To address the aforementioned problems, this invention provides an intelligent adaptive dose verification calculation method and system, specifically employing the following technical solution: A smart adaptive dose verification calculation method, comprising: The patient's medical imaging data, target area structure data, prescription dose data, and planned beam parameters are acquired and input into a pre-trained deep learning-based dose prediction model to obtain the patient's preliminary dose distribution data. Based on the dose distribution information in the preliminary dose distribution data, a variance reduction parameter is dynamically assigned to the particles in the Monte Carlo simulation. The variance reduction parameter includes simulation weights, and the simulation weights assigned to particles predicted to be in high-dose regions are less than the simulation weights assigned to particles predicted to be in low-dose regions. Using a Monte Carlo dose calculation engine, the planned beam is simulated and calculated based on the dynamically allocated variance reduction parameters to obtain high-precision verification dose distribution data.

[0010] As an optional embodiment of the present invention, in an intelligent adaptive dose verification calculation method of the present invention, based on the dose high and low distribution information in the preliminary dose distribution data, a variance reduction parameter is dynamically assigned to the particles in the Monte Carlo simulation, including: The axial bounding box (BBox) between the tumor target area and the organs at risk is calculated based on the target area structure data. Based on the preset high dose value DOSEmax, the isodose lines of the preliminary dose distribution data are calculated, and the area with a dose value greater than or equal to the high dose value DOSEmax is defined as the high dose area Area-hi. In the Monte Carlo simulation, the initial simulation weights for each particle are set to W0; For beam i in the planned beam, determine whether the particle P(m, n, o, θ, φ) in beam i intersects with the axial bounding box BBox, where (m, n, o) are the position coordinates of the particle in the dose distribution space, and (θ, φ) are the angular directions of the particle; If they do not intersect, the simulated weights of the particles P(m, n, o, θ, φ) are set to 0 for truncation. If they intersect, based on the preliminary dose distribution data, it is further determined whether the particle P(m, n, o, θ, φ) belongs to a high-dose region or a low-dose region, and the corresponding simulation weight allocation operation is performed, wherein the simulation weight allocated to particles in the high-dose region is less than the simulation weight allocated to particles in the low-dose region.

[0011] As an optional embodiment of the present invention, in an intelligent adaptive dose verification calculation method of the present invention, the step of performing the corresponding simulation weight allocation operation includes: Determine whether the particle P(m, n, o, θ, φ) is located in the high-dose region Area-hi; If the particle is located in the high-dose region Area-hi, the simulated weight remains unchanged; if it is located outside the high-dose region Area-hi, the simulated weight is determined according to the predefined dose-weight mapping relationship.

[0012] As an optional embodiment of the present invention, in an intelligent adaptive dose verification calculation method of the present invention, the preliminary dose distribution data of the patient includes: The sub-dose distribution data DS-AIi corresponding to each planned beam is predicted using a pre-trained deep learning-based dose prediction model. The sub-dose distribution data of all planned beams are then summed to obtain the total preliminary dose distribution data DT-AI.

[0013] As an optional embodiment of the present invention, in an intelligent adaptive dose verification calculation method of the present invention, determining the simulation weights according to a predefined dose-weight mapping relationship includes: For any particle, obtain the dose prediction value corresponding to its spatial position in the sub-dose distribution data DS-AIi; Based on the ratio of the predicted dose value to the maximum dose value in the sub-dose distribution data DS-AIi corresponding to the planned beam to which the particle belongs, the simulated weight parameter of the particle is determined through a predefined dose-weight mapping relationship; wherein, the larger the ratio, the lower the determined simulated weight parameter.

[0014] As an optional embodiment of the present invention, an intelligent adaptive dose verification calculation method of the present invention includes: The sub-dose distribution data DS-AIi corresponding to the beam i to which the particle belongs is obtained from the preliminary dose distribution data; In the sub-dose distribution data DS-AIi, query the dose prediction value DS-AIi-P(m,n,o) for the particle position; Query the maximum dose value DS-AIi-MAX in the sub-dose distribution data DS-AIi; Calculate the dose ratio Ratio-P = DS-AIi-P(m,n,o) / DS-AIi-MAX; Based on the preset segmented interval into which the dose ratio Ratio-P falls, and according to the preset dose ratio-weight mapping relationship, the simulated weight parameters of the particle are determined.

[0015] As an optional embodiment of the present invention, in an intelligent adaptive dose verification calculation method of the present invention, the preset dose ratio-weight mapping relationship is as follows: WP = Wj, Ratio-P belongs to (Rj, Rj+1], where j=1,..,m, and R1 takes the value 0; The dose ratio Ratio-P ranges from (0 to 1). The dose ratio Ratio-P is set as a segmented interval {(R1,R2], …, (Rj, Rj+1], …, (Rm, 1]}. The weight value sequence {Wj} corresponding to the segmented interval of the dose ratio Ratio-P is set, j=1,..,m.

[0016] As an optional embodiment of the present invention, an intelligent adaptive dose verification calculation method of the present invention includes: The high-precision verified dose distribution data is compared with the preliminary dose distribution data generated by the treatment planning system to generate a dose verification result report.

[0017] This invention also provides an intelligent adaptive dose verification calculation system for implementing the method, comprising: The dose prediction module acquires the patient's medical imaging data, target area structure data, prescription dose data, and planned beam parameters, and inputs them into a pre-trained deep learning-based dose prediction model to obtain the patient's preliminary dose distribution data. The adaptive variance reduction generation module dynamically assigns variance reduction parameters to particles in the Monte Carlo simulation based on the high and low dose distribution information in the preliminary dose distribution data. The variance reduction parameters include simulation weights, with the simulation weights assigned to particles predicted to be in high-dose regions being less than those assigned to particles predicted to be in low-dose regions. The Monte Carlo dose calculation module uses a Monte Carlo dose calculation engine to simulate and calculate the planned beam based on the dynamically allocated variance reduction parameters, thereby obtaining high-precision verification dose distribution data.

[0018] As an optional embodiment of the present invention, the adaptive variance reduction generation module includes: The region analysis unit calculates the axial bounding box (BBox) between the tumor target region and the organs at risk based on the target region structure data. Based on the preset high dose value DOSEmax, the isodose lines of the preliminary dose distribution data are calculated, and the area with a dose value greater than or equal to the high dose value DOSEmax is defined as the high dose area Area-hi. The particle weight decision unit sets the initial simulation weight of each particle in the Monte Carlo simulation to W0; For beam i in the planned beam, determine whether the particle P(m, n, o, θ, φ) in beam i intersects with the axial bounding box BBox, where (m, n, o) are the position coordinates of the particle in the dose distribution space, and (θ, φ) are the angular directions of the particle; If they do not intersect, the simulated weights of the particles P(m, n, o, θ, φ) are set to 0 for truncation. If they intersect, based on the preliminary dose distribution data, it is further determined whether the particle P(m, n, o, θ, φ) belongs to a high-dose region or a low-dose region, and the corresponding simulation weight allocation operation is performed, wherein the simulation weight allocated to particles in the high-dose region is less than the simulation weight allocated to particles in the low-dose region.

[0019] By adopting the above technical solution, the present invention has the following technical effects: 1. It specifically accelerates convergence in the low-dose region, solving the efficiency bottleneck in Monte Carlo simulations.

[0020] In traditional Monte Carlo simulations, the low-dose region, due to its low particle flux and large statistical fluctuations, is often the most time-consuming and difficult to converge to the predetermined statistical accuracy, becoming a bottleneck for overall computational efficiency. This embodiment presents an intelligent adaptive dose verification calculation method that creatively assigns higher simulation weights to particles in these regions, amplifying the contribution of each particle's historical data to the final dose statistics. This is equivalent to significantly enhancing the signal strength in the low-dose region without increasing the number of particle emissions, thereby enabling the dose calculation results in this region to converge quickly to a stable value with fewer simulation samples and shorter simulation time, directly overcoming the key challenge restricting overall computational speed.

[0021] 2. It achieves intelligent rebalancing of computing resources, achieving optimal global efficiency while ensuring accuracy.

[0022] This invention presents an intelligent adaptive dose verification calculation method that embodies highly intelligent resource scheduling logic: maintaining the original strategy for parts with good calculation status (high-dose region), and specifically strengthening parts with calculation status needing improvement (low-dose region). This avoids the complexity and uncertainty that may arise from adjusting weights across the entire domain. By precisely allocating limited additional computing resources (represented by weights) to the most needed low-dose regions, the system achieves an effective balance of convergence speeds between different regions with minimal strategy adjustments and computational overhead, thereby driving the entire dose distribution to reach the preset statistical accuracy requirements in a shorter time.

[0023] 3. It embodies true "adaptive" optimization, with a high degree of intelligence and no need for human intervention.

[0024] The variance reduction strategy is automatically and in real-time generated by the system based on AI-predicted dose distribution. The system intelligently identifies the dose distribution under each patient's unique anatomy and treatment plan, and dynamically determines the weight of each particle accordingly. The entire process eliminates the need for physicists to manually set any global or regional variance reduction parameters based on experience, completely eliminating the dependence on human experience and the resulting inaccuracies and instabilities in traditional methods. This achieves full automation and intelligence in the variance reduction process, ensuring the method's high universality and robustness across different clinical cases.

[0025] 4. It adheres to the "gold standard" precision core of the Monte Carlo method, ensuring the credibility of the verification results.

[0026] The core of this invention lies in optimizing the Monte Carlo simulation sampling strategy, rather than altering its underlying physical model. The final dose calculation results still originate entirely from simulations performed by a rigorously physically modeled Monte Carlo engine. Weight adjustment, as an unbiased variance reduction technique, mathematically guarantees that the expected value of the final dose statistics remains unchanged, i.e., it does not affect the calculation accuracy. Therefore, this method achieves significant efficiency gains while fully preserving the physical rigor and reliability of Monte Carlo dose calculation as the "gold standard," perfectly meeting the core requirement of authoritative results for independent clinical dose validation.

[0027] In summary, the intelligent adaptive dose verification calculation method of this invention, through reverse thinking, focuses on improving the statistical contribution of particles in the low-dose region, cleverly solving the convergence efficiency bottleneck of Monte Carlo simulations. This scheme combines significant acceleration, fully intelligent adaptive characteristics, and uncompromising computational accuracy, providing a practical and feasible key technical path for the large-scale application of high-precision Monte Carlo dose verification in routine clinical workflows. Attached Figure Description

[0028] Figure 1A flowchart of an intelligent adaptive dose verification calculation method according to an embodiment of the present invention Figure 1 ; Figure 2 A flowchart of an intelligent adaptive dose verification calculation method according to an embodiment of the present invention Figure 2 . Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0030] Therefore, the following detailed description of embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely illustrates some embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0031] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the present invention can be combined with each other.

[0032] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0033] In the description of this invention, it should be noted that the terms "upper," "lower," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. These terms are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0034] See Figure 1 As shown, an intelligent adaptive dose verification calculation method in this embodiment includes: The patient's medical imaging data, target area structure data, prescription dose data, and planned beam parameters are acquired and input into a pre-trained deep learning-based dose prediction model to obtain the patient's preliminary dose distribution data. Based on the dose distribution information in the preliminary dose distribution data, a variance reduction parameter is dynamically assigned to the particles in the Monte Carlo simulation. The variance reduction parameter includes simulation weights, and the simulation weights assigned to particles predicted to be in high-dose regions are less than the simulation weights assigned to particles predicted to be in low-dose regions. Using a Monte Carlo dose calculation engine, the planned beam is simulated and calculated based on the dynamically allocated variance reduction parameters to obtain high-precision verification dose distribution data.

[0035] The intelligent adaptive dose verification calculation method of this embodiment produces the following significant and beneficial technical effects: 1. It specifically accelerates convergence in the low-dose region, solving the efficiency bottleneck in Monte Carlo simulations.

[0036] In traditional Monte Carlo simulations, the low-dose region, due to its low particle flux and large statistical fluctuations, is often the most time-consuming and difficult to converge to the predetermined statistical accuracy, becoming a bottleneck for overall computational efficiency. This embodiment presents an intelligent adaptive dose verification calculation method that creatively assigns higher simulation weights to particles in these regions, amplifying the contribution of each particle's historical data to the final dose statistics. This is equivalent to significantly enhancing the signal strength in the low-dose region without increasing the number of particle emissions, thereby enabling the dose calculation results in this region to converge quickly to a stable value with fewer simulation samples and shorter simulation time, directly overcoming the key challenge restricting overall computational speed.

[0037] 2. It achieves intelligent rebalancing of computing resources, achieving optimal global efficiency while ensuring accuracy.

[0038] This embodiment presents an intelligent adaptive dose verification calculation method that embodies highly intelligent resource scheduling logic: maintaining the original strategy for parts with good calculation status (high-dose region), and specifically strengthening parts with calculation status needing improvement (low-dose region). This avoids the complexity and uncertainty that may arise from adjusting weights across the entire domain. By precisely allocating limited additional computing resources (represented by weights) to the most needed low-dose regions, the system achieves an effective balance of convergence speeds between different regions with minimal strategy adjustments and computational overhead, thereby driving the entire dose distribution to reach the preset statistical accuracy requirements in a shorter time.

[0039] 3. It embodies true "adaptive" optimization, with a high degree of intelligence and no need for human intervention.

[0040] The variance reduction strategy is automatically and in real-time generated by the system based on AI-predicted dose distribution. The system intelligently identifies the dose distribution under each patient's unique anatomy and treatment plan, and dynamically determines the weight of each particle accordingly. The entire process eliminates the need for physicists to manually set any global or regional variance reduction parameters based on experience, completely eliminating the dependence on human experience and the resulting inaccuracies and instabilities in traditional methods. This achieves full automation and intelligence in the variance reduction process, ensuring the method's high universality and robustness across different clinical cases.

[0041] 4. It adheres to the "gold standard" precision core of the Monte Carlo method, ensuring the credibility of the verification results.

[0042] The core of this approach lies in optimizing the Monte Carlo simulation sampling strategy, rather than altering its underlying physical model. The final dose calculation results still originate entirely from simulations performed by a rigorously physically modeled Monte Carlo engine. Weight adjustment, as an unbiased variance reduction technique, mathematically guarantees that the expected value of the final dose statistics remains unchanged, i.e., it does not affect the calculation accuracy. Therefore, this method achieves significant efficiency gains while fully preserving the physical rigor and reliability of Monte Carlo dose calculation as the "gold standard," perfectly meeting the core requirement of authoritative results for independent clinical dose validation.

[0043] In summary, the intelligent adaptive dose verification calculation method of this embodiment, through reverse thinking, focuses on improving the statistical contribution of particles in the low-dose region, cleverly solving the convergence efficiency bottleneck of Monte Carlo simulations. This scheme combines significant acceleration, fully intelligent adaptive characteristics, and uncompromising computational accuracy, providing a practical and feasible key technical path for the large-scale application of high-precision Monte Carlo dose verification in routine clinical workflows. For an optional implementation method of this embodiment, see [link to implementation details]. Figure 2 As shown, in an intelligent adaptive dose verification calculation method of this embodiment, based on the dose high and low distribution information in the preliminary dose distribution data, a variance reduction parameter is dynamically assigned to the particles in the Monte Carlo simulation, including: The axial bounding box (BBox) between the tumor target area and the organs at risk is calculated based on the target area structure data. Based on the preset high dose value DOSEmax, the isodose lines of the preliminary dose distribution data are calculated, and the area with a dose value greater than or equal to the high dose value DOSEmax is defined as the high dose area Area-hi. In the Monte Carlo simulation, the initial simulation weights for each particle are set to W0; For beam i in the planned beam, determine whether the particle P(m, n, o, θ, φ) in beam i intersects with the axial bounding box BBox, where (m, n, o) are the position coordinates of the particle in the dose distribution space, and (θ, φ) are the angular directions of the particle; If they do not intersect, the simulated weights of the particles P(m, n, o, θ, φ) are set to 0 for truncation. If they intersect, based on the preliminary dose distribution data, it is further determined whether the particle P(m, n, o, θ, φ) belongs to a high-dose region or a low-dose region, and the corresponding simulation weight allocation operation is performed, wherein the simulation weight allocated to particles in the high-dose region is less than the simulation weight allocated to particles in the low-dose region.

[0044] The above-described optional implementation provides a specific, reliable, and efficient technical path for the dynamic allocation of variance reduction parameters by introducing a dual fine-grained judgment mechanism of "axial bounding box" and "quantification of high-dose region," thereby generating the following significant beneficial effects on the basis of general methods: (1) It realizes “two-level precise screening” of computing resources, which greatly improves the acceleration efficiency.

[0045] This implementation innovatively employs a two-stage progressive judgment process. In the first stage, utilizing the concise and efficient spatial model of the axial bounding box (BBox), particles whose trajectories will never enter the target area or critical regions of the organ are quickly filtered out, and their weights are directly reset to zero (truncated). This step significantly eliminates completely unnecessary computational overhead in the earliest stage of simulation. In the second stage, for particles that enter the critical region, differentiated weights are assigned based on their position within the quantized high-dose region Area-hi. This two-stage "space-first, dose-later" screening mechanism, compared to weight adjustments solely based on dose information, can reduce unnecessary particle simulations earlier and more thoroughly, thus achieving a greater computational acceleration at the source.

[0046] (2) An objective and repeatable decision-making standard was established, which ensured the robustness and consistency of the method.

[0047] This implementation uses a well-defined geometric definition (axial bounding box) and a quantization threshold (preset DOSEmax) as the basis for all decision-making. The generation of the axial bounding box and the calculation of isodose lines are both deterministic algorithmic processes, completely unaffected by subjective human experience. This ensures that the system can execute variance reduction strategies according to uniform and objective standards for different patients and different treatment plans, completely eliminating inconsistencies in the validation process caused by differences in human experience, and significantly improving the reliability and comparability (robustness) of this method in large-scale clinical applications.

[0048] Furthermore, in the intelligent adaptive dose verification calculation method of this embodiment, the execution of the corresponding simulation weight allocation operation includes: Determine whether the particle P(m, n, o, θ, φ) is located in the high-dose region Area-hi; If the particle is located in the high-dose region Area-hi, the simulated weight remains unchanged; if it is located outside the high-dose region Area-hi, the simulated weight is determined according to the predefined dose-weight mapping relationship.

[0049] Specifically, for a particle P(m, n, o, θ, φ) and its direction DIR(θ, φ), calculate whether the particle intersects the axial bounding box BBox from its current position Position(m, n, o) along the direction DIR(θ, φ).

[0050] In the intelligent adaptive dose verification calculation method of this embodiment, the patient's preliminary dose distribution data includes: The sub-dose distribution data DS-AIi corresponding to each planned beam is predicted using a pre-trained deep learning-based dose prediction model. The sub-dose distribution data of all planned beams are then summed to obtain the total preliminary dose distribution data DT-AI.

[0051] Specifically, the calculation process for the overall preliminary dose distribution data DT-AI includes: (1) The deep learning dose prediction model predicts and generates dose distribution data DS-AIi (i=1,..,n) generated by each beam in the patient’s body based on the patient’s CT images, target area delineation data, organ at risk delineation data, the target area prescription dose given by the doctor, and the number and direction of the beam given by the doctor. (2) The dose distribution data of each beam are summed to calculate the total dose distribution data DT-AI: .

[0052] The dose prediction module of deep learning in this embodiment uses a prediction network including the UNet network model or its improved model.

[0053] In the intelligent adaptive dose verification calculation method of this embodiment, determining the simulation weights according to the predefined dose-weight mapping relationship includes: For any particle, obtain the dose prediction value corresponding to its spatial position in the sub-dose distribution data DS-AIi; Based on the ratio of the predicted dose value to the maximum dose value in the sub-dose distribution data DS-AIi corresponding to the planned beam to which the particle belongs, the simulated weight parameter of the particle is determined through a predefined dose-weight mapping relationship; wherein, the larger the ratio, the smaller the determined simulated weight parameter.

[0054] The specific implementation of this embodiment introduces a continuous or segmented mapping relationship based on the normalized dose ratio, providing a scientific, precise, and universally applicable technical means for the dynamic allocation of particle weights, achieving "refinement" and "smoothness" in weight allocation, and optimizing the balance between acceleration and accuracy.

[0055] This implementation abandons the simple binary weight allocation of "high / low dose regions" and instead adopts a continuous or piecewise mapping directly related to the normalized dose ratio. This means that the weight value of a particle can more finely reflect the relative dose importance of its location. In regions with drastic dose gradient changes (such as the edge of the target area), adjacent particles may obtain a more reasonable weight gradient due to slight positional differences. This avoids either-or weight jumps, making the allocation of computational resources smoother and more precise. This fine-grained control helps optimize the acceleration effect at a more microscopic scale, while ensuring that the statistical uncertainty of the dose distribution varies more reasonably in space, thus achieving a better balance between acceleration efficiency and computational accuracy overall.

[0056] Therefore, this dose-weight mapping scheme is the core algorithm for the efficient, accurate, and reliable implementation of the intelligent adaptive strategy of this invention. It transforms clinical experience and physical goals into executable mathematical rules, and through refined weight control, provides crucial technical support for achieving high-precision Monte Carlo validation in routine clinical applications.

[0057] Specifically, the intelligent adaptive dose verification calculation method of this embodiment includes: The sub-dose distribution data DS-AIi corresponding to the beam i to which the particle belongs is obtained from the preliminary dose distribution data; In the sub-dose distribution data DS-AIi, query the dose prediction value DS-AIi-P(m,n,o) for the particle position; Query the maximum dose value DS-AIi-MAX in the sub-dose distribution data DS-AIi; Calculate the ratio Ratio-P = DS-AIi-P(m,n,o) / DS-AIi-MAX; Based on the preset segmented interval into which the ratio Ratio-P falls, and according to the preset dose-weight mapping relationship, the simulated weight parameters of the particle are determined.

[0058] This implementation method introduces a "segmented mapping based on normalized dose ratio" mechanism, providing a precise, quantifiable, and flexible execution path for the aforementioned core strategy of "enhancing particle weight in the low-dose region," thereby producing the following key and significant technical effects: 1. It achieves "refined" and "automated" quantitative control of particle weights, greatly improving the execution efficiency and accuracy of acceleration strategies.

[0059] This scheme transforms the abstract concept of "dose level" into a specific mathematical ratio, Ratio-P, which serves as the sole and objective input parameter for particle weight allocation. Through automatic querying, calculation, and interval judgment, the system can determine the precise simulated weight of each particle in a massive dataset in real time, without any manual intervention. This process is fully programmable and parallelizable, transforming intelligent decision-making into efficient arithmetic and logical operations. This ensures that the entire adaptive variance reduction module can run at extremely high speeds, avoiding becoming a new computational bottleneck, thus translating the theoretical advantages of the acceleration strategy into practical system performance improvements.

[0060] 2. Standardized criteria for measuring “relative importance” have been established to ensure the universality of the strategy across different beams and plans.

[0061] This approach creatively employs an "intra-beam normalized ratio" as the basis for weight allocation by dividing the predicted dose of a particle's location by the maximum dose of its associated beam (DS-AIi-MAX). This method cleverly eliminates the absolute dose level differences caused by variations in energy, weight, or path length among different beams, ensuring that weight decisions are based solely on the particle's relative dose contribution within its beam. This endows the method with strong adaptability: the same set of preset mapping relationships can be automatically applied to multiple beams with varying dose levels in a treatment plan, as well as various personalized plans for different patients, without adjustment, significantly improving the method's versatility and ease of clinical deployment.

[0062] 3. It provides a highly flexible and interpretable "strategy configuration" interface to meet diverse clinical needs for balancing accuracy and efficiency.

[0063] The pre-defined segmentation intervals and mapping relationships constitute a clear and configurable strategy panel. Medical physicists can scientifically design segmentation thresholds and corresponding weight values ​​based on the unique needs of specific clinical scenarios (such as head and neck treatments that are extremely sensitive to dose gradients, or routine validation that pursues extremely high efficiency). For example, the highest weight can be set for extremely low ratio intervals to accelerate their convergence, while for high ratio regions close to 1, the weight can be set to an unchanged initial value W0. This hierarchical and non-linear control capability allows this method to allocate computational resources in a more scientific way that aligns with clinical physics goals, achieving optimal overall computational efficiency while ensuring absolute priority for computational accuracy in critical regions.

[0064] 4. It ensures the "determinism" and "verifiability" of the algorithm process, providing a reliable guarantee for clinical quality control.

[0065] The entire weight determination process consists of a series of deterministic queries, calculations, and logical comparisons, ensuring complete repeatability. The same input (dose distribution, particle position) will inevitably produce the same weight output. This determinism not only guarantees the reliability of the calculation results but also greatly facilitates system quality assurance (QA). Standard test cases can be designed to verify whether the weight output fully conforms to the preset segmented mapping table. This quantifiable and automated verification characteristic makes the core algorithm module easy to verify, calibrate, and audit, fully complying with the stringent regulatory requirements for key algorithm components in medical device software, ensuring its long-term, stable, and reliable operation in a clinical environment. As an optional implementation of this embodiment, in an intelligent adaptive dose verification calculation method, the preset dose-weight mapping relationship is as follows: WP = Wj, Ratio-P belongs to (Rj, Rj+1], where j=1,..,m, and R1 takes the value 0; The dose ratio Ratio-P ranges from (0 to 1). The dose ratio Ratio-P is set as a segmented interval {(R1,R2], …, (Rj, Rj+1], …, (Rm, 1]}. The weight value sequence {Wj} corresponding to the segmented interval of the dose ratio Ratio-P is set, j=1,..,m.

[0066] This embodiment of an intelligent adaptive dose verification calculation method includes: The high-precision verified dose distribution data is compared with the preliminary dose distribution data generated by the treatment planning system to generate a dose verification result report.

[0067] This embodiment presents an intelligent adaptive dose verification calculation method that directly and objectively compares high-precision Monte Carlo calculation results with the original planned dose from the treatment planning system (TPS), thus completing the crucial leap from "independent calculation" to "result verification." It establishes a complete and closed loop throughout the entire technical process, from plan input and intelligent accelerated calculation to final conclusion output, ensuring that the final output of the method is a clear clinical conclusion (pass / fail, or a quantitative report of difference), rather than isolated numerical data. This significantly improves the clinical applicability of the technology and the level of end-to-end automation.

[0068] By comparing the "gold standard" Monte Carlo results with the planned dose, this method provides the most authoritative objective benchmark for assessing the dosimetric accuracy of treatment plans. The generated dose validation reports (typically including quantitative indicators such as Gamma pass rate, dose difference distribution maps, and key point dose deviations) clearly and intuitively reveal potential dosimetric biases in the plan that may be masked by fast algorithms. This provides medical physicists and physicians with a solid and reliable basis for decision-making, helping them determine whether the plan is safe and feasible, or requires further optimization. This ensures the accuracy of radiotherapy and patient safety at key stages, significantly improving the quality control capabilities of the entire treatment process.

[0069] This embodiment also provides an intelligent adaptive dose verification calculation system for implementing the method, including: The dose prediction module acquires the patient's medical imaging data, target area structure data, prescription dose data, and planned beam parameters, and inputs them into a pre-trained deep learning-based dose prediction model to obtain the patient's preliminary dose distribution data. The adaptive variance reduction generation module dynamically assigns variance reduction parameters to particles in the Monte Carlo simulation based on the high and low dose distribution information in the preliminary dose distribution data. The variance reduction parameters include simulation weights, with the simulation weights assigned to particles predicted to be in high-dose regions being less than those assigned to particles predicted to be in low-dose regions. The Monte Carlo dose calculation module uses a Monte Carlo dose calculation engine to simulate and calculate the planned beam based on the dynamically allocated variance reduction parameters, thereby obtaining high-precision verification dose distribution data.

[0070] As an optional implementation of this embodiment, the adaptive variance reduction generation module includes: The region analysis unit calculates the axial bounding box (BBox) between the tumor target region and the organs at risk based on the target region structure data. Based on the preset high dose value DOSEmax, the isodose lines of the preliminary dose distribution data are calculated, and the area with a dose value greater than or equal to the high dose value DOSEmax is defined as the high dose area Area-hi. The particle weight decision unit sets the initial simulation weight of each particle in the Monte Carlo simulation to W0; For beam i in the planned beam, determine whether the particle P(m, n, o, θ, φ) in beam i intersects with the axial bounding box BBox, where (m, n, o) are the position coordinates of the particle in the dose distribution space, and (θ, φ) are the angular directions of the particle; If they do not intersect, the simulated weights of the particles P(m, n, o, θ, φ) are set to 0 for truncation. If they intersect, based on the preliminary dose distribution data, it is further determined whether the particle P(m, n, o, θ, φ) belongs to a high-dose region or a low-dose region, and the corresponding simulation weight allocation operation is performed, wherein the simulation weight allocated to particles in the high-dose region is less than the simulation weight allocated to particles in the low-dose region.

[0071] This embodiment also provides a computer-readable storage medium storing a computer-executable program, which, when executed, implements the intelligent adaptive dose verification calculation method as described above.

[0072] The computer-readable storage medium described in this embodiment may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0073] This embodiment also provides an electronic device, including a processor and a memory, wherein the memory is used to store a computer-executable program, and when the computer program is executed by the processor, the processor executes the intelligent adaptive dose verification calculation method.

[0074] The electronic device is manifested in the form of a general-purpose computing device. It may contain one or more processors that work collaboratively. This invention also does not preclude distributed processing, meaning that processors may be distributed across different physical devices. The electronic device of this invention is not limited to a single entity, but may also be the sum of multiple physical devices.

[0075] The memory stores a computer-executable program, typically machine-readable code. The computer-readable program can be executed by the processor to enable the electronic device to perform the method of the present invention, or at least some steps of the method.

[0076] The memory includes volatile memory, such as random access memory (RAM) and / or cache memory, and may also be non-volatile memory, such as read-only memory (ROM).

[0077] It should be understood that the electronic device of the present invention may also include elements or components not shown in the examples above. For example, some electronic devices also include display units such as a display screen, and some electronic devices also include human-computer interaction elements such as buttons and keyboards. Any electronic device capable of executing a computer-readable program in its memory to implement the method of the present invention or at least some steps of the method can be considered as an electronic device covered by the present invention.

[0078] From the above description of the embodiments, those skilled in the art will readily understand that the present invention can be implemented by hardware capable of executing specific computer programs, such as the system of the present invention, and the electronic processing unit, server, client, mobile phone, control unit, processor, etc. included in the system. The present invention can also be implemented by computer software executing the methods of the present invention, for example, by control software executed by a microprocessor, electronic control unit, client, server, etc. However, it should be noted that the computer software executing the methods of the present invention is not limited to execution in one or a specific set of hardware entities; it can also be implemented in a distributed manner by unspecified hardware. For computer software, the software product can be stored in a computer-readable storage medium (such as a CD-ROM, USB flash drive, portable hard drive, etc.) or distributed across a network, as long as it enables electronic devices to execute the methods according to the present invention.

[0079] The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described herein. Although the present invention has been described in detail with reference to the above embodiments, the present invention is not limited to the specific embodiments described above. Therefore, any modifications or equivalent substitutions to the present invention, as well as all technical solutions and improvements that do not depart from the spirit and scope of the invention, are covered within the scope of the claims of the present invention.

Claims

1. A smart adaptive dose verification calculation method, characterized in that, include: The patient's medical imaging data, target area structure data, prescription dose data, and planned beam parameters are acquired and input into a pre-trained deep learning-based dose prediction model to obtain the patient's preliminary dose distribution data. Based on the dose distribution information in the preliminary dose distribution data, a variance reduction parameter is dynamically assigned to the particles in the Monte Carlo simulation. The variance reduction parameter includes simulation weights, and the simulation weights assigned to particles predicted to be in high-dose regions are less than the simulation weights assigned to particles predicted to be in low-dose regions. Using a Monte Carlo dose calculation engine, the planned beam is simulated and calculated based on the dynamically allocated variance reduction parameters to obtain high-precision verification dose distribution data.

2. The intelligent adaptive dose verification calculation method according to claim 1, characterized in that, Based on the dose distribution information in the preliminary dose distribution data, variance reduction parameters are dynamically assigned to particles in the Monte Carlo simulation, including: The axial bounding box (BBox) between the tumor target area and the organs at risk is calculated based on the target area structure data. Based on the preset high dose value DOSEmax, the isodose lines of the preliminary dose distribution data are calculated, and the area with a dose value greater than or equal to the high dose value DOSEmax is defined as the high dose area Area-hi. In the Monte Carlo simulation, the initial simulation weights for each particle are set to W0; For beam i in the planned beam, determine whether the particle P(m, n, o, θ, φ) in beam i intersects with the axial bounding box BBox, where (m, n, o) are the position coordinates of the particle in the dose distribution space, and (θ, φ) are the angular directions of the particle; If they do not intersect, the simulated weights of the particles P(m, n, o, θ, φ) are set to 0 for truncation. If they intersect, based on the preliminary dose distribution data, it is further determined whether the particle P(m, n, o, θ, φ) belongs to a high-dose region or a low-dose region, and the corresponding simulation weight allocation operation is performed, wherein the simulation weight allocated to particles in the high-dose region is less than the simulation weight allocated to particles in the low-dose region.

3. The intelligent adaptive dose verification calculation method according to claim 2, characterized in that, The execution of the corresponding simulated weight allocation operation includes: Determine whether the particle P(m, n, o, θ, φ) is located in the high-dose region Area-hi; If the particle is located in the high-dose region Area-hi, the simulated weight remains unchanged; if it is located outside the high-dose region Area-hi, the simulated weight is determined according to the predefined dose-weight mapping relationship.

4. The intelligent adaptive dose verification calculation method according to claim 3, characterized in that, The patient's preliminary dose distribution data includes: The sub-dose distribution data DS-AIi corresponding to each planned beam is predicted using a pre-trained deep learning-based dose prediction model. The sub-dose distribution data of all planned beams are then summed to obtain the total preliminary dose distribution data DT-AI.

5. The intelligent adaptive dose verification calculation method according to claim 4, characterized in that, The step of determining the simulation weights based on the predefined dose-weight mapping relationship includes: For any particle, obtain the dose prediction value corresponding to its spatial position in the sub-dose distribution data DS-AIi; Based on the ratio of the predicted dose value to the maximum dose value in the sub-dose distribution data DS-AIi corresponding to the planned beam to which the particle belongs, the simulated weight parameter of the particle is determined through a predefined dose-weight mapping relationship; wherein, the larger the ratio, the lower the determined simulated weight parameter.

6. The intelligent adaptive dose verification calculation method according to claim 5, characterized in that, include: The sub-dose distribution data DS-AIi corresponding to the beam i to which the particle belongs is obtained from the preliminary dose distribution data; In the sub-dose distribution data DS-AIi, query the dose prediction value DS-AIi-P(m,n,o) for the particle position; Query the maximum dose value DS-AIi-MAX in the sub-dose distribution data DS-AIi; Calculate the dose ratio Ratio-P = DS-AIi-P(m,n,o) / DS-AIi-MAX; Based on the preset segmented interval into which the dose ratio Ratio-P falls, and according to the preset dose ratio-weight mapping relationship, the simulated weight parameters of the particle are determined.

7. The intelligent adaptive dose verification calculation method according to claim 6, characterized in that, The preset dose ratio-weight mapping relationship is as follows: WP = W j Ratio-P belongs to (R j ,R j+1 ], where j=1,..,m, and R1 takes the value 0; The dose ratio Ratio-P ranges from (0 to 1). The dose ratio Ratio-P is set as a segmented interval {(R1,R2], …, (Rj, Rj+1], …, (Rm, 1]}. The weight value sequence {Wj} corresponding to the segmented interval of the dose ratio Ratio-P is set, j=1,..,m.

8. The intelligent adaptive dose verification calculation method according to claim 1, characterized in that, include: The high-precision verified dose distribution data is compared with the preliminary dose distribution data generated by the treatment planning system to generate a dose verification result report.

9. An intelligent adaptive dose verification calculation system, used to implement the method as described in any one of claims 1-8, characterized in that, include: The dose prediction module acquires the patient's medical imaging data, target area structure data, prescription dose data, and planned beam parameters, and inputs them into a pre-trained deep learning-based dose prediction model to obtain the patient's preliminary dose distribution data. The adaptive variance reduction generation module dynamically assigns variance reduction parameters to particles in the Monte Carlo simulation based on the high and low dose distribution information in the preliminary dose distribution data. The variance reduction parameters include simulation weights, with the simulation weights assigned to particles predicted to be in high-dose regions being less than those assigned to particles predicted to be in low-dose regions. The Monte Carlo dose calculation module uses a Monte Carlo dose calculation engine to simulate and calculate the planned beam based on the dynamically allocated variance reduction parameters, thereby obtaining high-precision verification dose distribution data.

10. The intelligent adaptive dose verification calculation system according to claim 9, characterized in that, The adaptive variance reduction generation module includes: The region analysis unit calculates the axial bounding box (BBox) between the tumor target region and the organs at risk based on the target region structure data. Based on the preset high dose value DOSEmax, the isodose lines of the preliminary dose distribution data are calculated, and the area with a dose value greater than or equal to the high dose value DOSEmax is defined as the high dose area Area-hi. The particle weight decision unit sets the initial simulation weight of each particle in the Monte Carlo simulation to W0; For beam i in the planned beam, determine whether the particle P(m, n, o, θ, φ) in beam i intersects with the axial bounding box BBox, where (m, n, o) are the position coordinates of the particle in the dose distribution space, and (θ, φ) are the angular directions of the particle; If they do not intersect, the simulated weights of the particles P(m, n, o, θ, φ) are set to 0 for truncation. If they intersect, based on the preliminary dose distribution data, it is further determined whether the particle P(m, n, o, θ, φ) belongs to a high-dose region or a low-dose region, and the corresponding simulation weight allocation operation is performed, wherein the simulation weight allocated to particles in the high-dose region is less than the simulation weight allocated to particles in the low-dose region.