Dose prediction method and device and electronic equipment

Abstract

The invention provides a dose prediction method, which comprises the following steps: generating an initial vector based on initial prescription information and patient structure data; generating an initialization parameter based on the initial vector; generating initial dose distribution according to the initialization parameters; and adjusting the initial dose distribution based on a plurality of agents, and generating a target dose distribution, where a single agent is responsible for at least partial adjustment of the initial dose distribution. The invention further provides a device for executing the method and electronic equipment.

Description

Technical Field

[0001] The present disclosure relates to the field of medical radiation, and more specifically, to a dose prediction method, device, and electronic equipment. Background Art

[0002] In existing tumor radiotherapy planning, dose optimization methods primarily rely on rule-driven planning systems and manual adjustments based on expert experience. These traditional methods face significant drawbacks. The first is the heavy workload, requiring clinicians to spend considerable time and effort manually adjusting dose distribution to ensure a balance between target dose coverage and protection of organs at risk. This manual operation is not only inefficient but also prone to fluctuations in the quality of the results, limited by individual experience.

[0003] In recent years, the application of deep learning technology in the field of dose prediction has gradually increased. However, these methods are mostly limited to supervised learning frameworks, cannot dynamically adjust dose distribution, and lack confidence assessment of the output results. In addition, although existing reinforcement learning techniques are beginning to be introduced into radiotherapy plan optimization, most methods use a single-agent architecture, which makes it difficult to solve the collaborative optimization problem of multi-structure, multi-conflict, and multi-target dose distribution. When dealing with complex anatomical structures, single agents often cannot fully consider the personalized needs of different regions, resulting in less than ideal optimization results, especially when multiple target organs need to be taken into account simultaneously. Summary of the Invention

[0004] In view of this, the present disclosure provides a dose prediction method, device, and electronic device.

[0005] One aspect of the present disclosure provides a dose prediction method, comprising: generating an initial vector based on initial prescription information and patient structural data; generating initialization parameters based on the initial vector; generating an initial dose distribution according to the initialization parameters; and adjusting the initial dose distribution based on multiple intelligent agents to generate a target dose distribution, wherein a single intelligent agent is responsible for at least part of the adjustment of the initial dose distribution.

[0006] According to an embodiment of the present disclosure, the intelligent agent is an intelligent agent based on reinforcement learning, and adjusts the initial dose distribution, including: performing the following operations until a stop condition is met: obtaining a dose distribution input, where the dose distribution input is the dose distribution result generated by the previous adjustment, or the initial dose distribution; adjusting the dose distribution input based on multiple intelligent agents to generate an intermediate dose distribution result corresponding to the adjustment; in response to meeting the stop condition, using the intermediate dose distribution result corresponding to the last adjustment as the target dose distribution.

[0007] According to an embodiment of the present disclosure, the stopping condition includes at least one of the following: in N consecutive adjustments, the reward improvement of each intelligent agent in reinforcement learning is less than a preset reward threshold, where N is a positive integer greater than or equal to 5; the number of adjustments is greater than or equal to a preset number threshold; the intermediate dose distribution result obtained after the adjustment meets the preset clinical constraints.

[0008] According to an embodiment of the present disclosure, a process of generating an intermediate dose distribution result at one time includes: obtaining a dose distribution input; adjusting at least one first intelligent agent among multiple intelligent agents with respect to the dose distribution input to obtain a first dose distribution; adjusting at least one second intelligent agent among multiple intelligent agents based on the first dose distribution to obtain a second dose distribution; obtaining an intermediate dose distribution result generated by this adjustment based on the first dose distribution and the second dose distribution; wherein the first intelligent agent and the second intelligent agent are different.

[0009] According to an embodiment of the present disclosure, a process of generating an intermediate dose distribution result at one time includes:

[0010] Each intelligent agent generates its own corresponding third dose distribution based on the dose distribution input, and the third dose distribution represents the adjustment result of the intelligent agent for the dose distribution input; based on the degree of influence of each third dose distribution on the patient's structural data, each third dose distribution is adjusted; based on each adjusted third dose distribution, an intermediate dose distribution result is generated; wherein, the degree of influence at least represents the dose degree of the third dose for the target organ in the patient's structural data; wherein, the target organs targeted by different intelligent agents at least partially do not overlap.

[0011] According to an embodiment of the present disclosure, the intermediate dose distribution result includes a target distribution value floating range, and the process of generating the intermediate dose distribution result once includes: each intelligent agent generates a fourth dose distribution multiple times based on the dose distribution input; generates a target distribution value floating range based on each fourth dose distribution; wherein, the process of generating the fourth dose distribution once includes one of the following operations: randomly obtaining the action value corresponding to the reinforcement learning from the preset first probability distribution, and generating the fourth dose distribution based on the action value; randomly obtaining the preset weight distribution from the preset second probability distribution, adjusting the weights of each network in the intelligent agent according to the preset weight distribution, and generating the fourth dose distribution based on the adjusted intelligent agent; randomly selecting the target neuron connection in the intelligent agent, adjusting the intelligent agent according to the target neuron connection, and generating the fourth dose distribution based on the adjusted intelligent agent, and the target neuron connection does not participate in the generation process of the fourth dose distribution.

[0012] According to an embodiment of the present disclosure, the method further includes: generating characteristic information based on the patient's structural data, the characteristic information at least characterizing the spatial relationship between each target area in the patient, and / or the medical characteristics of each target area, the target area including each of the patient's organ-at-risk areas, the target area targeted by the dose distribution, the dose gradient transition area, and at least one of the patient's normal tissue areas; determining the target calculation area targeted by each intelligent agent based on the characteristic information, the target calculation area containing at least one target organ; adjusting the initial dose distribution, including: each intelligent agent adjusting the dose distribution of the target calculation area targeted by each intelligent agent in the initial dose distribution to generate a target dose distribution.

[0013] According to an embodiment of the present disclosure, structural features are generated based on patient structural data, including: generating feature information based on the spatial overlap between the target area targeted by the dose distribution input and the organs at risk, and / or the distribution location of the organs at risk, and / or the number of organs at risk, and / or the difficulty of predicting the dose distribution of each target area.

[0014] Another aspect of the present disclosure provides a dose prediction device, including: a first generation module for generating an initial vector based on initial prescription information and patient structure data; a second generation module for generating initialization parameters based on the initial vector; a third generation module for generating an initial dose distribution according to the initialization parameters; and a fourth generation module for adjusting the initial dose distribution based on multiple intelligent agents to generate a target dose distribution, wherein a single intelligent agent is responsible for at least part of the adjustment of the initial dose distribution.

[0015] Another aspect of the present disclosure provides an electronic device comprising: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the dose prediction method of any one of the aforementioned embodiments.

[0016] Another aspect of the present disclosure provides a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the dose prediction method according to any one of the aforementioned embodiments.

[0017] Another aspect of the present disclosure provides a computer program product, including a computer program / instruction, characterized in that the computer program / instruction implements the operation of the dose prediction method of any of the aforementioned embodiments when executed by a processor.

[0018] According to the embodiments of the present disclosure, the dose prediction scheme provided by the present disclosure has at least one of the following beneficial effects: First, by introducing a multi-agent architecture, the system can adaptively divide tasks according to the anatomical characteristics of different patients, ensuring that each agent focuses on a specific target calculation area. This approach significantly improves the ability to fine-tune the dose distribution, especially when dealing with complex structures and multi-objective conflicts, and can achieve more efficient dose optimization. The present disclosure adopts a feature information-driven agent division and adjustment mechanism, which enables the system to have good medical interpretability and engineering operability. Feature information includes spatial overlap, the distribution location and number of organs at risk, and the difficulty of dose distribution prediction, which helps the system to have higher flexibility and accuracy when dealing with high-risk areas. In addition, through uncertainty modeling, the confidence interval of the dose map is output, which improves the credibility of clinical decision-making. Doctors can identify high-risk areas and perform manual intervention based on uncertainty prompts to reduce clinical risks and enhance the acceptance of AI systems in clinical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0020] Figure 1 The flowchart of the dose prediction method according to the embodiment of the present disclosure is schematically shown;

[0021] Figure 2 Schematically shows a flow chart for adjusting the initial dose distribution in the dose prediction method according to an embodiment of the present disclosure;

[0022] Figure 3 A flowchart for generating an intermediate dose distribution result in one step in a dose prediction method according to an embodiment of the present disclosure is schematically shown;

[0023] Figure 4 Another flow chart schematically illustrates a method for generating an intermediate dose distribution result in one step according to an embodiment of the present disclosure;

[0024] Figure 5 Another flow chart schematically illustrates a method for generating an intermediate dose distribution result in one step according to an embodiment of the present disclosure;

[0025] Figure 6 Another flow chart of the dose prediction method according to an embodiment of the present disclosure is schematically shown;

[0026] Figure 7 A block diagram schematically shows a dose prediction device according to an embodiment of the present disclosure; and

[0027] Figure 8A block diagram of an electronic device suitable for implementing the above-described method according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0028] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the detailed description below, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0029] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0030] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0031] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0032] In the embodiments of this disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of all data involved (including, but not limited to, user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to safeguard the security of user personal information, network security, and national security.

[0033] An embodiment of the present disclosure provides a dose prediction method, comprising: generating an initial vector based on initial prescription information and patient structural data; generating initialization parameters based on the initial vector; generating an initial dose distribution according to the initialization parameters; and adjusting the initial dose distribution based on multiple intelligent agents to generate a target dose distribution, wherein a single intelligent agent is responsible for at least part of the adjustment of the initial dose distribution.

[0034] Figure 1 The flowchart of the dose prediction method according to an embodiment of the present disclosure is schematically shown.

[0035] like Figure 1 As shown, the dose prediction method may at least include operations S110 to S140.

[0036] In operation S110 , an initial vector is generated based on initial prescription information and patient structure data.

[0037] Initial prescription information can be the constraints of the radiotherapy plan developed by the clinician based on the patient's condition and treatment goals. These constraints typically include minimum dose coverage requirements for the target volume, maximum dose limits for organs at risk, and treatment fraction and total dose requirements. For example, the initial prescription information may include a target D95 (the dose received by 95% of the target volume) of 70 Gy and a maximum dose of no more than 45 Gy to organs at risk, such as the spinal cord.

[0038] Patient structural data can be information about a patient's anatomical structure obtained through medical imaging technology, typically including computed tomography (CT) images, outlines of target volumes and organs at risk, voxel masks, and related spatial geometry information. For example, patient structural data may include a CT image of a lung cancer patient, where the target volume is the tumor area, and organs at risk include the outlines of the lungs, esophagus, and heart, as well as voxel representations of these structures.

[0039] The initial vector can be a unified mathematical representation of the state input for the reinforcement learning environment, formed by feature extraction and encoding of the initial prescription information and patient structural data. The initial vector typically includes the spatial coordinates of the target volume and the organ at risk, the degree of overlap, the difficulty of dose distribution prediction, and historical dose-volume histogram (DVH) features. For example, the initial vector can be a high-dimensional feature vector containing the three-dimensional coordinates of the target center point, the spatial overlap between the target volume and the heart (expressed as a percentage of voxel overlap), and the DVH prediction value based on historical cases.

[0040] In operation S120, initialization parameters are generated based on the initial vector.

[0041] Initialization parameters can be generated based on the initial vector by referring to a historical case database or a pre-trained model, and are used to initialize the parameter set of the reinforcement learning system. These parameters include the weights of the generative model for the initial dose distribution, the initial weights of the reinforcement learning strategy network, or the control parameters of the heuristic algorithm. For example, the initialization parameters may include a dose distribution template similar to the current patient's structure extracted from a historical lung cancer case database, or the weights of a pre-trained convolutional neural network model, used to generate reference values for the initial dose distribution.

[0042] In operation S130 , an initial dose distribution is generated according to the initialization parameters.

[0043] The initial dose distribution can be a preliminary dose distribution map generated based on initialization parameters using a pre-trained model, heuristic algorithm, or historical case template, and is used as the starting point for multi-agent reinforcement learning optimization. The initial dose distribution is usually represented as a three-dimensional voxel grid, with each voxel corresponding to a dose value, which initially meets the prescription requirements but is not optimal. For example, the initial dose distribution can be a three-dimensional dose map, in which the dose values of the target voxels initially meet the requirement of D95 = 70Gy, but the dose distribution of organs at risk, such as the lungs, still needs further optimization to reduce side effects.

[0044] In operation S140 , the initial dose distribution is adjusted based on the multiple agents to generate a target dose distribution, wherein a single agent is responsible for at least a portion of the adjustment of the initial dose distribution.

[0045] An agent is an independent decision-making unit built using a reinforcement learning algorithm. Each agent learns optimization strategies through interaction with the environment, responsible for adjusting the dose to specific areas or targets within the initial dose distribution. Deep reinforcement learning models, such as the Dual-Dual Deep Q-Network (D3QN), are typically used within the agent to achieve efficient action selection and strategy optimization.

[0046] Adjusting the initial dose distribution involves multiple agents dynamically adjusting the dose values within the voxel grid based on the current dose distribution and environmental feedback through an iterative reinforcement learning process to meet clinical prescription requirements and optimize dose distribution quality. The adjustment process involves agents selecting actions (such as increasing or decreasing the dose value of a voxel) based on their respective target areas (such as the target volume or specific organs at risk). The effectiveness of these adjustments is evaluated using a reward function, ultimately resulting in a globally optimized target dose distribution.

[0047] The disclosed embodiment adopts a collaborative optimization mechanism of multiple intelligent agents based on reinforcement learning, breaking through the limitations of "local modification and global uncontrollability" in traditional manual parameter adjustment methods. Traditional manual planning is limited by the perception of multi-variable linkage effects of human decision-making ability, and is prone to falling into the thinking mode of "the current qualified dose is the end point", thereby ignoring the potential for subsequent optimization. However, the reinforcement learning intelligent agent does not take immediate improvement as the only goal, but adjusts its strategy based on long-term accumulated rewards, so that each dose modification can be embedded in the global optimization judgment of the future sequence.

[0048] Furthermore, the multi-agent architecture enables each agent to focus on dose adjustments for specific structures or regions, simulating the division of labor and collaboration among multiple experts. This strengthens the modeling of coordination between conflicting structures, effectively avoiding the problem of a single agent falling into local optimality and improving the global consistency and robustness of dose distribution in spatial and structural dimensions. Therefore, the dose prediction method provided by the disclosed embodiments can generate target dose maps of higher quality, greater coordination, and greater stability, while ensuring clinical dose constraints, and possesses greater individual adaptability and clinical practicality.

[0049] Figure 2 The flowchart of adjusting the initial dose distribution in the dose prediction method according to an embodiment of the present disclosure is schematically shown.

[0050] According to the embodiment of the present disclosure, the intelligent agent is an intelligent agent based on reinforcement learning, such as Figure 2 As shown, based on the above embodiment, S140 may include looping operations S210 to S220 until a stop condition is met, and operation S230.

[0051] In operation S210 , a dose distribution input is acquired, where the dose distribution input is a dose distribution result generated by a previous adjustment, or an initial dose distribution.

[0052] The dose distribution input can be considered as the dose map data for the current state in the reinforcement learning loop. This input can be the initial dose distribution initially generated by the system or an intermediate dose map generated by multiple agents during a previous adjustment operation. This distribution data is expressed as a three-dimensional voxel matrix, with each voxel containing a dose value, which each agent uses to select the next action based on the target structure region it is responsible for. For example, in the first round, the dose distribution input is a dose map generated based on the initialization parameters, while in the second round, it is the intermediate dose map obtained after the first round of collaborative adjustment by the agents.

[0053] In operation S220 , the dose distribution input is adjusted based on the multiple agents, and an intermediate dose distribution result corresponding to the adjustment is generated.

[0054] Each reinforcement learning agent fine-tunes the local or cross-region dose values of the current dose map based on the calculation area it is responsible for (such as a specific target area or organ at risk). Each agent makes an action selection for the current state based on its internal reinforcement learning model (such as D3QN). The selected action may be to increase, decrease, or keep the dose of a certain area unchanged. The outputs of multiple agents can be fused in series or in parallel to finally synthesize a new intermediate dose distribution. The fusion process considers the degree of influence of each region on the clinical goal and weights it. For example, regions with higher target importance are given greater weight during fusion. This intermediate dose distribution is used to subsequently determine whether the stopping condition has been met.

[0055] In operation S230 , in response to a stop condition being met, an intermediate dose distribution result corresponding to the last adjustment is used as a target dose distribution.

[0056] The system determines whether the current optimization is complete based on the stopping criteria set during the reinforcement learning optimization process. If the stopping criteria are met, the current intermediate dose map is used as the final output target dose map for clinical use or for subsequent verification modules.

[0057] According to an embodiment of the present disclosure, the stopping condition includes at least one of the following: in N consecutive adjustments, the reward improvement of each intelligent agent in reinforcement learning is less than a preset reward threshold, where N is a positive integer greater than or equal to 5; the number of adjustments is greater than or equal to a preset number threshold; the intermediate dose distribution result obtained after the adjustment meets the preset clinical constraints.

[0058] Specifically, the stopping condition can be: if the increase in the reward value obtained by each agent in N consecutive rounds (such as N=10) is less than a set threshold (such as 0.01), it means that the system has entered the learning saturation state and is judged to be converged.

[0059] Specifically, the stopping condition may be: if the number of iterations reaches the maximum round limit (e.g., 100 rounds), to prevent infinite loops or model overfitting, the system may also force the output of the current dose map as a result.

[0060] Specifically, the stopping condition can be: if the dose-volume histogram (DVH) indicators of the current intermediate dose map have all met the prescription limits (such as the target area D95 reaches 70Gy and the Dmax of the organ at risk does not exceed the limit value), it can also be regarded as reaching the optimization endpoint.

[0061] Through this optimization mechanism, this paper implements a closed-loop design for the entire "state-action-reward-update" process in a reinforcement learning system, ensuring the traceability and progressiveness of optimization results. Compared to traditional one-time adjustment methods based on heuristic rules, this solution introduces multiple rounds of dynamic updates, ensuring that each dose modification action produces a coherent response based on historical context, supporting the exploration of long-term optimal solutions.

[0062] Figure 3 The flowchart of generating an intermediate dose distribution result in one step in the dose prediction method according to an embodiment of the present disclosure is schematically shown.

[0063] like Figure 3 As shown, based on the above embodiment, S220 may include operations S310 to S340.

[0064] In operation S310 , a dose distribution input is acquired.

[0065] In operation S320 , at least one first intelligent agent among the plurality of intelligent agents makes adjustments based on the dose distribution input to obtain a first dose distribution.

[0066] In this step, the first agent adjusts the dose distribution input based on the structural features within its control range (e.g., a target volume or organ at risk). Based on a reinforcement learning policy model, the agent selects actions, such as increasing or decreasing the dose to a specific voxel region, to generate an adjusted first dose distribution. This adjustment process is based on the "environmental state" of the dose distribution input. The agent takes corresponding actions based on the state value and the policy network output, and updates the dose value of the controlled region. For example, if the first agent controls a portion of the tumor target, it may prioritize increasing the central dose if the target volume D95 is insufficient, thus forming the first dose distribution.

[0067] In operation S330 , at least one second intelligent agent among the plurality of intelligent agents adjusts the dose distribution input based on the first dose distribution to obtain a second dose distribution.

[0068] In this step, the second agent receives the first dose distribution and the original dose distribution as input and determines its action strategy based on both. Unlike the direct adjustments made by the first agent, the second agent can compensate or fine-tune its strategy based on the changing trend of the first dose distribution. For example, if the first agent increases the dose to the target area, the second agent may be responsible for regulating the dose to the organs at risk area spatially adjacent to the target area to prevent it from exceeding the safety limit. This serial mechanism ensures the sequential dependencies between agents and the causal consistency of the adjustment logic, helping to improve the overall stability of the system.

[0069] In operation S340, an intermediate dose distribution result generated by the current adjustment is obtained based on the first dose distribution and the second dose distribution, wherein the first agent and the second agent are different.

[0070] The first agent and the second agent are different. On the one hand, this limitation ensures that each agent in the system has independent division of responsibilities and structural control rights, preventing repeated adjustments by the same agent from causing dose fluctuations or strategy convergence failures. On the other hand, the differentiated agent role settings allow the system to simulate the clinical control ideas of "first target, then avoidance" and "first main, then auxiliary", and realize sequential modeling of phased optimization goals. Through the serial control structure, the system can introduce the posterior compensation mechanism of the second agent based on the adjustment results of the first agent, strengthening local risk control capabilities and global dose coordination.

[0071] Through this serial agent collaboration mechanism, this embodiment effectively avoids conflicts, overlaps, or mutual cancellations that might arise from multiple agents operating in parallel, enhancing causality and coordination among the decision-making units within the system. The clear structural responsibilities of the first and second agents, coupled with sequential actions, make system behavior more aligned with human planning, while offering a higher-dimensional, coordinated, and strategically coordinated capability.

[0072] Figure 4 Another flowchart for generating an intermediate dose distribution result at one time in the dose prediction method according to an embodiment of the present disclosure is schematically shown.

[0073] like Figure 4 As shown, based on the above embodiment, S220 may include operations S410 to S430.

[0074] In operation S410 , each intelligent agent generates a third dose distribution corresponding to itself according to the dose distribution input, where the third dose distribution represents an adjustment result of the intelligent agent with respect to the dose distribution input.

[0075] In this step, multiple reinforcement learning agents receive the same dose distribution input in parallel and independently perform policy reasoning and dose adjustment within their respective target computational regions. Based on its internal policy network, each agent modifies the dose field of its control structure and generates a "third dose distribution" within the complete dose map. This distribution reflects the optimal dose adjustment behavior that the agent should adopt for the current state. For example, the agent responsible for controlling the lung region will attempt to reduce high-dose areas within its region and limit the volume that exceeds the standard; while the agent responsible for the target region may focus on improving dose coverage.

[0076] In combination with the aforementioned embodiments, the third dose distribution generated by at least part of the second intelligent agents among the multiple intelligent agents is generated based on the third dose distribution generated by at least part of the first intelligent agents among the multiple intelligent agents. For details, please refer to the aforementioned embodiments and will not be repeated here.

[0077] In operation S420 , each of the third dose distributions is adjusted based on the degree of influence of each of the third dose distributions on the patient structure data.

[0078] The impact level at least characterizes the dose level of the third dose to the target organ in the patient's structural data. Specifically, the impact level reflects the sensitivity, contribution, and optimization importance of the third dose distribution generated by the intelligent agent to the dose response of key anatomical structures (such as target volumes or organs at risk). The impact level can be calculated based on the following dimensions:

[0079] The rate of change of the dose distribution to the target organ's mean dose, maximum dose or D95 and other key DVH indicators. And / or,

[0080] The degree of improvement in the dose gradient in the area, such as reducing dose spillover or increasing edge dose concentration. And / or,

[0081] The degree of dose interference on adjacent structures, that is, whether it causes unwanted collateral damage.

[0082] For example, if an agent generates a third dose distribution that reduces the maximum dose to the heart region it is responsible for from 46 Gy to 44 Gy, while the clinical limit is 45 Gy, the agent's adjustment behavior has a higher "influence" weight.

[0083] Among them, the target organs targeted by different agents do not overlap. This "non-overlap" does not require that the structures involved by each agent are completely non-overlapping, but means that their task objectives have at least some structural differences. Even if multiple agents have control intersections in the boundary area of the same organ, as long as there are significant differences in their key areas of action, target calculation sub-areas, dose adjustment strategies or reward targets, they can be considered as incomplete overlap. For example, one agent may be responsible for the dose protection of the upper lobe of the lung organ, while another agent focuses on the volume control of the lower lobe. Although both belong to the "lung" structure, their target areas and dose strategies are different.

[0084] In operation S430 , an intermediate dose distribution result is generated based on the adjusted third dose distributions.

[0085] The system fuses the adjusted third dose distributions generated by each agent as candidate results to produce a unified intermediate dose distribution. This fusion process uses a structure-guided weighting approach. Rather than simply averaging or clipping individual dose maps, it assigns fusion weights based on their actual impact on the patient's critical structures.

[0086] Specifically, for each voxel, the system first determines whether the voxel is within the target structure area controlled by an agent, such as the target volume, high-risk organ, or transition zone. If so, the agent's optimization contribution to that area is prioritized, that is, whether the dose adjustment it generates effectively improves target coverage or reduces the risk of excess dose to organs at risk.

[0087] Based on this, the system dynamically assigns fusion priority weights to different agents in different regions. For example, if an agent successfully reduces the dose in the heart region from 47Gy to 44Gy (where the maximum dose requirement is 45Gy), its output in the heart region will be given a higher weight. If another agent improves the D95 value at the edge of the target volume, its output in that region will be prioritized.

[0088] Through the above-mentioned structure-guided weighting strategy, the system achieves local optimal output integration within the entire structure, so that the final generated intermediate dose map is not only numerically continuous and coordinated, but also can reflect the optimal control and protection effect on the key structure.

[0089] Figure 5 Another flowchart for generating an intermediate dose distribution result at one time in the dose prediction method according to an embodiment of the present disclosure is schematically shown.

[0090] According to an embodiment of the present disclosure, the intermediate dose distribution result may include a target distribution value floating range such as Figure 5 As shown, based on the above embodiment, S220 may include operations S510 to S520.

[0091] In operation S510 , each agent generates a fourth dose distribution multiple times based on the dose distribution input.

[0092] In this step, to estimate the uncertainty in the dose prediction process, each reinforcement learning agent, given the same dose distribution input, uses different perturbation mechanisms to perturb its internal decision-making process multiple times. Each perturbation generates a complete dose map, the "fourth dose distribution." These distributions constitute the uncertainty sample set for the dose prediction, which is used for subsequent statistical inference of the target dose range. For example, an agent targeting lung structure can generate 20 dose maps based on the same input state, with each image varying due to the internal policy perturbation, thus characterizing the output variability.

[0093] The process of generating the fourth dose distribution includes one of the following operations:

[0094] The agent randomly extracts action values corresponding to reinforcement learning from a preset first probability distribution and generates a fourth dose distribution based on these action values. Specifically, given the current state input, the agent no longer takes greedy actions or a maximum Q-value strategy, but instead samples actions from a preset policy distribution (such as a Softmax strategy or a Boltzmann distribution). The sampled actions are used to adjust the dose value (e.g., increase, decrease, maintain), and a dose map is output. Multiple random samplings can account for fluctuations in the policy space and reflect output differences caused by policy uncertainty.

[0095] A preset weight distribution is randomly obtained from a preset second probability distribution. The weights of each network in the agent are adjusted according to the preset weight distribution. A fourth dose distribution is generated based on the adjusted agent. Specifically, this simulates the sources of uncertainty at the model structure level. This is done by sampling network parameters (such as the weight matrix in a neural network) from a Bayesian prior distribution or a known Gaussian perturbation, while keeping the input unchanged. This is then applied to the policy network, and a new dose map is generated based on the perturbed network. This approach can be used to evaluate the impact of changes in model structure on the output dose map and constitutes an estimation of structural uncertainty.

[0096] A target neuron connection in the agent is randomly selected, and the agent is adjusted based on the target neuron connection. A fourth dose distribution is generated based on the adjusted agent. The target neuron connection does not participate in the generation of the fourth dose distribution. Specifically, the uncertainty of the neural network activation path is modeled by randomly selecting some neuron connections (such as those in the intermediate hidden layer or output layer) and temporarily blocking the activation paths of these connections. This allows the agent to reason using different sub-network structures for this decision. Each perturbation outputs a new dose map, thereby constructing a set of dose distributions that account for the uncertainty of the activation paths.

[0097] In operation S520 , a target distribution value floating range is generated based on each fourth dose distribution.

[0098] The system regards the fourth dose distribution generated multiple times under the same input state as a sample set of dose values, and statistically analyzes the range of variation at each voxel position. Specifically, the system calculates the mean and variance of all sample maps at the voxels at the same position, and then forms the expected dose value and its confidence interval. For example, if the dose value distribution of a certain voxel in 20 dose maps has a mean of 70Gy and a standard deviation of 1.5Gy, the system can calculate the dose fluctuation range of the voxel at 95% confidence level to be 67Gy~73Gy, and use this as part of the target dose floating range map. The final output includes: central predicted dose map + per-voxel dose confidence interval map, which makes it easier for clinical personnel to identify areas with higher uncertainty, so as to perform manual intervention or risk reassessment

[0099] Traditional systems often struggle to generate representative and stable predictions due to the limited number of training data samples and the high individual variability of tumor types and structural layouts, particularly for rare structural layouts (such as tumors near the heart). If the system only outputs a single "dose map prediction," clinicians will be unable to determine its reliability, thus creating the risk of blindly trusting the AI. This disclosed embodiment introduces uncertainty into the dose prediction process by introducing a mechanism that generates a "fourth dose distribution" through multiple perturbations. The mean and fluctuation range of each voxel dose are statistically inferred. Ultimately, the system output includes a central dose prediction and a replica floating map, clearly indicating which regions have stable and reliable predictions and which regions have high fluctuations or high uncertainty. Physicians can identify areas of potential risk in the system's predictions by reviewing high-uncertainty areas. In complex structural situations, such as tumors near high-risk organs, the uncertainty bandwidth indicator can guide the treatment team to adjust segmentation strategies or adopt a more conservative approach, thereby enhancing the credibility, transparency, and clinical acceptability of AI radiotherapy systems.

[0100] Figure 6 Another flowchart of the dose prediction method according to an embodiment of the present disclosure is schematically shown.

[0101] like Figure 6 As shown, based on the above embodiment, the dose prediction method may include operations S610 to S620.

[0102] In operation S610, characteristic information is generated based on the patient's structural data. The characteristic information at least characterizes the spatial relationship between various target areas in the patient, and / or the medical characteristics of each target area. The target area includes at least one of the patient's organ-at-risk areas, the target area targeted by the dose distribution, the dose gradient transition area, and the patient's normal tissue area.

[0103] In this step, the system extracts information describing the spatial relationships and medical characteristics of target regions from the patient's CT images, MRI data, or other anatomical images. This information includes not only the target region's geometry (such as volume, boundary curvature, and spatial location), but also its functional properties during treatment (such as tissue type, sensitivity, and known dose limits).

[0104] The system identifies the patient's at-risk organ regions, such as the heart, spinal cord, and esophagus, which are sensitive organs requiring special attention during radiotherapy. By identifying the spatial location, proximity, and historical dose constraints of these regions, the system determines their priority and influencing factors within the feature information.

[0105] The target volume (target area) for dose distribution, the primary target area for tumor treatment, typically includes the Gross Tumor Volume (GTV), Clinical Target Volume (CTV), and Planning Target Volume (PTV). The system meticulously delineates the target volume and annotates its spatial relationship with surrounding OARs using feature information to optimize dose coverage.

[0106] The dose gradient transition zone, typically 1-2 cm wide, is the dose buffer zone between the target and normal tissue. The system uses signature information to identify the rate of dose change in this region to ensure that the dose gradient is properly controlled during treatment planning.

[0107] The patient's normal tissue area can include background areas such as lung parenchyma, soft tissue, and bone. The system uses feature information to describe its impact on the entire dose distribution to avoid over-irradiation.

[0108] In operation S620, a target calculation region targeted by each agent is determined based on the feature information, where the target calculation region includes at least one target organ.

[0109] For example, the target calculation area for each agent is determined based on the spatial overlap of target regions within the patient. By calculating the voxel overlap, contact area, and relative positional relationships between different target regions, the complexity of the region is determined and agent responsibilities are assigned accordingly. For example, if the target region has a high degree of overlap with the heart, the system will prioritize the dedicated agent for refined dose control.

[0110] For example, based on the distribution location of each organ at risk and / or the number of organs at risk, the target calculation area targeted by each intelligent agent is determined. By identifying the spatial density and number of OARs, the system automatically adjusts the distribution strategy of the intelligent agent to ensure that high-risk areas are managed by intelligent agents with multi-target collaboration capabilities.

[0111] For example, based on the difficulty of predicting the dose distribution in each target area, the target calculation area targeted by each intelligent agent is determined. The system analyzes the prediction error and dose fluctuation rate of different areas based on historical data, and configures intelligent agent resources to deal with high-difficulty areas to ensure the stability and accuracy of the optimization process.

[0112] S140 may include operation S630.

[0113] In operation S630 , each agent adjusts the dose distribution of the target calculation region targeted by each agent in the initial dose distribution to generate a target dose distribution.

[0114] By introducing a feature-information-driven agent division and adjustment mechanism, the present disclosure realizes a technical shift from static task configuration to dynamic structure-aware division, enabling the system to adaptively determine the area and strategy parameters that each agent should control based on the anatomical characteristics of each patient, thereby more efficiently utilizing computing resources and avoiding unnecessary agent redundancy.

[0115] At the same time, the use of feature information makes the segmentation process medically interpretable and engineering computable, making it easier for doctors to understand the AI segmentation logic and making the subsequent dose map generation more clinically reliable. This significantly enhances optimization effectiveness and risk control capabilities, especially for cases with high structural heterogeneity and ambiguous treatment boundaries, such as those in the head and neck, mediastinum, or pelvic regions with multiple organs coexisting.

[0116] According to an embodiment of the present disclosure, generating characteristic information based on patient structural data may include: generating characteristic information based on the spatial overlap between the target area targeted by the dose distribution input and the organs at risk, and / or the distribution position of the organs at risk, and / or the number of organs at risk, and / or the difficulty of predicting the dose distribution of each target area.

[0117] Specifically, to calculate spatial overlap, the system first extracts 3D structural information of the target volume and organs at risk from the patient's CT or MRI images. By calculating the voxel overlap, contact area, and relative positional relationship between the target volume and each organ at risk, the system generates spatial overlap metrics. These metrics are used to assess conflict risk and complexity during dose adjustment.

[0118] Specifically, the system identifies the location and number of organs at risk within the patient's body using anatomical labeling or image segmentation techniques and records their spatial coordinates. The system further calculates the number and density of organs at risk within a unit volume to determine the concentration of high-risk areas, guiding the division of agents and resource allocation.

[0119] Specifically, the system assesses the difficulty of dose distribution prediction: Based on historical data and model performance analysis, it calculates the dose prediction difficulty for each target region. This includes evaluating the model's prediction error, dose volatility, and gradient stability on similar structures. The results of this assessment are used to dynamically adjust the agent's learning strategy and optimize resource allocation to ensure dose control accuracy in challenging regions.

[0120] By leveraging this feature information generation mechanism, the present disclosure enables refined description and quantitative analysis of patient anatomy, ensuring both medical interpretability and engineering feasibility for agent segmentation and dose optimization strategies. The system not only dynamically adapts to individual patient differences but also enables precise dose control within complex anatomic configurations, significantly improving the quality and safety of treatment plans.

[0121] In addition, the use of feature information enables the system to have greater flexibility and accuracy when dealing with high-risk areas, especially when the target area and the organs at risk highly overlap or the structure is dense. The system can optimize the dose distribution and reduce clinical risks through refined feature analysis and intelligent configuration.

[0122] Figure 7 The figure schematically shows a block diagram of a dose prediction device according to an embodiment of the present disclosure.

[0123] like Figure 7 As shown, the dose prediction device 700 may include a first generation module 710 , a second generation module 720 , a third generation module 730 , and a fourth generation module 740 .

[0124] The first generating module 710 is used to generate an initial vector based on the initial prescription information and the patient structure data. In some embodiments, the first generating module 710 can be used to perform operation S110 in the above-mentioned dose prediction method, which will not be described in detail here.

[0125] The second generation module 720 is used to generate initialization parameters based on the initial vector. In some embodiments, the second generation module 720 can be used to perform operation S120 in the above-mentioned dose prediction method, which will not be described in detail here.

[0126] The third generating module 720 is used to generate an initial dose distribution according to the initialization parameters. In some embodiments, the third generating module 720 can be used to perform operation S130 in the above-mentioned dose prediction method, which will not be described in detail here.

[0127] The fourth generation module 720 is configured to adjust the initial dose distribution based on the multiple agents to generate a target dose distribution, wherein a single agent is responsible for at least a portion of the adjustment of the initial dose distribution. In some embodiments, the fourth generation module 720 can be configured to perform operation S140 in the aforementioned dose prediction method, which is not further described here.

[0128] According to an embodiment of the present disclosure, the fourth generating module may include a first acquiring module, a first adjusting module, and a first determining module.

[0129] The first acquisition module is used to obtain a dose distribution input, which is a dose distribution result generated by a previous adjustment or an initial dose distribution. In some embodiments, the first acquisition module can be used to perform operation S210 in the above-mentioned dose prediction method, which will not be described in detail here.

[0130] The first adjustment module is used to adjust the dose distribution input based on multiple agents and generate an intermediate dose distribution result corresponding to the adjustment. In some embodiments, the first adjustment module can be used to perform operation S220 in the above-mentioned dose prediction method, which will not be described in detail here.

[0131] The first determination module is used to respond to the stop condition and use the intermediate dose distribution result corresponding to the last adjustment as the target dose distribution. In some embodiments, the first determination module can be used to perform operation S230 in the above-mentioned dose prediction method, which will not be described in detail here.

[0132] According to an embodiment of the present disclosure, the first adjustment module may include a second acquisition module, a second adjustment module, a third adjustment module, and a first integration module.

[0133] The second acquisition module is used to obtain a dose distribution input. In some embodiments, the second acquisition module can be used to perform operation S310 in the above-mentioned dose prediction method, which will not be described in detail here.

[0134] The second adjustment module is used to adjust at least one first agent among the multiple agents based on the dose distribution input to obtain a first dose distribution. In some embodiments, the second adjustment module can be used to perform operation S320 in the above-mentioned dose prediction method, which will not be described in detail here.

[0135] The third adjustment module is configured to adjust the dose distribution input based on the first dose distribution to obtain a second dose distribution. In some embodiments, the third adjustment module can be configured to perform operation S330 in the above-described dose prediction method, which is not described in detail here.

[0136] The first integration module is used to obtain an intermediate dose distribution result generated by this adjustment based on the first dose distribution and the second dose distribution. In some embodiments, the first integration module can be used to perform operation S340 in the above-mentioned dose prediction method, which will not be described in detail here.

[0137] Wherein, the first agent and the second agent are different.

[0138] According to an embodiment of the present disclosure, the first adjustment module may include a fifth generation module, a fourth adjustment module, and a third integration module.

[0139] The fifth generation module is used by each agent to generate a corresponding third dose distribution based on the dose distribution input. The third dose distribution represents the adjustment result of the agent based on the dose distribution input. In some embodiments, the fifth generation module can be used to perform operation S410 in the above-mentioned dose prediction method, which is not further described here.

[0140] The fourth adjustment module is used to adjust each third dose distribution based on the degree of influence of each third dose distribution on the patient structure data. In some embodiments, the fourth adjustment module can be used to perform operation S420 in the above-mentioned dose prediction method, which will not be described in detail here.

[0141] The third integration module is used to generate an intermediate dose distribution result based on the adjusted third dose distributions. In some embodiments, the third integration module can be used to perform operation S430 in the above-mentioned dose prediction method, which will not be described in detail here.

[0142] The impact degree at least represents the dose degree of the third dose to the target organ in the patient structure data;

[0143] Among them, the target organs targeted by different intelligent agents do not overlap.

[0144] According to an embodiment of the present disclosure, the intermediate dose distribution result includes a target distribution value floating range, and the first adjustment module may include a sixth generation module, a seventh generation module, a preset 23 module, a preset 24 module, and a preset 25 module.

[0145] The sixth generation module is used for each intelligent agent to generate the fourth dose distribution multiple times based on the dose distribution input. In some embodiments, the sixth generation module can be used to perform operation S510 in the above-mentioned dose prediction method, which will not be described in detail here.

[0146] The seventh generating module is used to generate a target distribution value floating range based on each fourth dose distribution. In some embodiments, the seventh generating module can be used to perform operation S520 in the above-mentioned dose prediction method, which will not be described in detail here.

[0147] The process of generating the fourth dose distribution includes one of the following operations:

[0148] Randomly obtaining an action value corresponding to the reinforcement learning from a preset first probability distribution, and generating a fourth dose distribution according to the action value;

[0149] randomly obtaining a preset weight distribution from a preset second probability distribution, adjusting the weights of each network in the intelligent agent according to the preset weight distribution, and generating a fourth dose distribution based on the adjusted intelligent agent;

[0150] A target neuron connection in the agent is randomly selected, the agent is adjusted according to the target neuron connection, and a fourth dose distribution is generated based on the adjusted agent. The target neuron connection does not participate in the generation process of the fourth dose distribution.

[0151] According to an embodiment of the present disclosure, the dose prediction device may include an eighth generation module and a third determination module.

[0152] The eighth generation module is configured to generate characteristic information based on the patient's structural data. The characteristic information at least characterizes the spatial relationship between target regions within the patient and / or the medical characteristics of each target region. The target region includes at least one of the patient's organ-at-risk regions, the target volume of the dose distribution, the dose gradient transition region, and the patient's normal tissue regions. In some embodiments, the eighth generation module can be configured to perform operation S610 in the aforementioned dose prediction method, which is not further described here.

[0153] The third determination module is used to determine the target calculation region targeted by each agent based on the feature information, where the target calculation region includes at least one target organ. In some embodiments, the third determination module can be used to perform operation S620 in the above-mentioned dose prediction method, which will not be described in detail here.

[0154] The fourth generation module further includes a ninth generation module configured to adjust the dose distribution of the target calculation region targeted by each agent in the initial dose distribution to generate a target dose distribution. In some embodiments, the ninth generation module may be configured to perform operation S630 of the aforementioned dose prediction method, which is not further described herein.

[0155] According to the embodiments of the present invention, any number of modules, sub-modules, units, and sub-units, or at least part of the functions of any number of them, can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, sub-modules, units, and sub-units can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, sub-modules, units, and sub-units can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging the circuit, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, sub-modules, units, and sub-units can be at least partially implemented as a computer program module, which can perform the corresponding functions when the computer program module is executed.

[0156] For example, any number of the first generation module 710, the second generation module 720, the third generation module 730, the fourth generation module 750, and the preset 5 module 750 can be combined into a single module / unit / sub-unit, or any one of these modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functionality of one or more of these modules / units / sub-units can be combined with at least part of the functionality of other modules / units / sub-units and implemented in a single module / unit / sub-unit. According to an embodiment of the present disclosure, at least one of the first generation module 710, the second generation module 720, the third generation module 730, the fourth generation module 750, and the preset 5 module 750 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented in hardware or firmware through any other reasonable means of circuit integration or packaging, or can be implemented in any one of software, hardware, and firmware, or any suitable combination of any of these. Alternatively, at least one of the first generation module 710, the second generation module 720, the third generation module 730, the fourth generation module 750 and the preset 5 module 750 can be at least partially implemented as a computer program module, which can perform corresponding functions when executed.

[0157] It should be noted that the data processing system part in the embodiments of the present disclosure corresponds to the data processing method part in the embodiments of the present disclosure. The description of the data processing system part specifically refers to the data processing method part and will not be repeated here.

[0158] Figure 8 A block diagram of an electronic device suitable for implementing the above-described method according to an embodiment of the present disclosure is schematically shown. Figure 8 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0159] like Figure 8As shown, the electronic device 800 according to an embodiment of the present disclosure includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage unit 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0160] Various programs and data required for the operation of the electronic device 800 are stored in the RAM 803. The processor 801, ROM 802, and RAM 803 are connected to each other via a bus 804. The processor 801 executes the various operations of the method flow according to the embodiment of the present disclosure by executing the programs in the ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than the ROM 802 and RAM 803. The processor 801 may also execute the various operations of the method flow according to the embodiment of the present disclosure by executing the programs stored in the one or more memories.

[0161] According to an embodiment of the present disclosure, electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to bus 804. Electronic device 800 may also include one or more of the following components connected to I / O interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including devices such as a cathode ray tube (CRT), liquid crystal display (LCD), and speakers; a storage section 808 including a hard disk; and a communication section 809 including a network interface card such as a LAN card or modem. Communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to I / O interface 805 as needed. Removable media 811, such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed in drive 810 as needed, so that computer programs read from the removable media can be installed into storage section 808 as needed.

[0162] According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the processor 801, the above-mentioned functions defined in the system of the embodiment of the present disclosure are executed. According to an embodiment of the present disclosure, the system, device, apparatus, module, unit, etc. described above can be implemented by a computer program module.

[0163] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not be incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, and when executed, implements the method according to the embodiments of the present disclosure.

[0164] According to embodiments of the present disclosure, a computer-readable storage medium may be a non-volatile computer-readable storage medium. Examples include, but are not limited to, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0165] For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 802 and / or the RAM 803 described above and / or one or more memories other than the ROM 802 and the RAM 803 .

[0166] An embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method provided by the embodiment of the present disclosure. When the computer program product runs on an electronic device, the program code is used to enable the electronic device to implement the control method provided by the embodiment of the present disclosure.

[0167] When the computer program is executed by the processor 801, the above functions defined in the system / device of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0168] In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and downloaded and installed through the communication part 809, and / or installed from a removable medium 811. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above. According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure may be written in any combination of one or more programming languages. Specifically, these computer programs may be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages include but are not limited to languages such as Java, C++, Python, "C" or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. Where a remote computing device is involved, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0169] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of the systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or part of the code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two boxes shown in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, as well as the combination of boxes in the block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or operations, or can be implemented using a combination of dedicated hardware and computer instructions. It will be understood by those skilled in the art that the features described in the various embodiments of the present disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or coupled in various ways, and all of these combinations and / or couplings fall within the scope of the present disclosure.

[0170] The above describes the embodiments of the present disclosure. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although each embodiment has been described separately above, this does not mean that the measures in each embodiment cannot be advantageously used in combination. Without departing from the scope of the present disclosure, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A dose prediction method, comprising: Generate an initial vector based on initial prescription information and patient structure data; Based on the initial vector, generating initialization parameters; generating an initial dose distribution according to the initialization parameters; The initial dose distribution is adjusted based on a plurality of intelligent agents to generate a target dose distribution, wherein a single intelligent agent is responsible for at least part of the adjustment of the initial dose distribution.

2. The method according to claim 1, characterized in that The intelligent agent is an intelligent agent based on reinforcement learning, and the adjusting the initial dose distribution includes: Perform the following actions until the stopping condition is met: Acquiring a dose distribution input, where the dose distribution input is a dose distribution result generated by a previous adjustment, or the initial dose distribution; Adjusting the dose distribution input based on multiple intelligent agents to generate an intermediate dose distribution result corresponding to the adjustment; In response to the stopping condition being met, the intermediate dose distribution result corresponding to the last adjustment is used as the target dose distribution.

3. The method according to claim 2, characterized in that The stop condition includes at least one of the following: During N consecutive adjustments, the reward increase of each of the intelligent agents in the reinforcement learning is less than a preset reward threshold, where N is a positive integer greater than or equal to 5; The number of adjustments is greater than or equal to the preset number threshold; The intermediate dose distribution result obtained after adjustment meets the preset clinical constraints.

4. The method according to claim 2, characterized in that The process of generating an intermediate dose distribution result includes: obtaining the dose distribution input; At least one first intelligent agent among the plurality of intelligent agents adjusts according to the dose distribution input to obtain a first dose distribution; At least one second intelligent agent among the plurality of intelligent agents adjusts the dose distribution input based on the first dose distribution to obtain a second dose distribution; Obtaining an intermediate dose distribution result generated by this adjustment based on the first dose distribution and the second dose distribution; Wherein, the first agent and the second agent are different.

5. The method according to claim 2, characterized in that The process of generating an intermediate dose distribution result includes: Each of the intelligent agents generates a third dose distribution corresponding to itself according to the dose distribution input, wherein the third dose distribution represents an adjustment result of the intelligent agent with respect to the dose distribution input; adjusting each of the third dose distributions based on the degree of influence of each of the third dose distributions on the patient structure data; generating the intermediate dose distribution result based on each of the adjusted third dose distributions; The impact degree at least represents the dose degree of the third dose to the target organ in the patient structure data; Among them, the target organs targeted by different intelligent agents at least partially do not overlap.

6. The method according to claim 2, characterized in that The intermediate dose distribution result includes a target distribution value floating range, and the process of generating the intermediate dose distribution result once includes: Each of the intelligent agents generates a fourth dose distribution multiple times based on the dose distribution input; generating a target distribution value floating range based on each fourth dose distribution; The process of generating the fourth dose distribution includes one of the following operations: randomly obtaining an action value corresponding to the reinforcement learning from a preset first probability distribution, and generating a fourth dose distribution according to the action value; randomly obtaining a preset weight distribution from a preset second probability distribution, adjusting the weights of each network in the intelligent agent according to the preset weight distribution, and generating a fourth dose distribution based on the adjusted intelligent agent; A target neuron connection in the agent is randomly selected, the agent is adjusted according to the target neuron connection, and the fourth dose distribution is generated according to the adjusted agent, wherein the target neuron connection does not participate in the generation process of the fourth dose distribution.

7. The method according to claim 1, characterized in that The method further comprises: generating characteristic information based on the patient structural data, the characteristic information at least characterizing a spatial relationship between target regions in the patient and / or a medical characteristic of each target region, the target region including at least one of each organ at risk region of the patient, a target region targeted by a dose distribution, a dose gradient transition region, and a normal tissue region of the patient; Determining a target calculation area targeted by each of the intelligent agents based on the characteristic information, wherein the target calculation area includes at least one target organ; The adjusting the initial dose distribution comprises: Each of the intelligent agents adjusts the dose distribution of the target calculation area targeted by each of the intelligent agents in the initial dose distribution to generate the target dose distribution.

8. The method according to claim 7, characterized in that Generating structural features according to the patient structural data includes: the spatial overlap between the target volume and the organ at risk according to the dose distribution input, and / or Location of organs at risk, and / or number of organs at risk, and / or, The difficulty of predicting the dose distribution of each target area is used to generate the characteristic information.

9. A dose prediction device, comprising: A first generating module is used to generate an initial vector based on initial prescription information and patient structure data; A second generating module, configured to generate initialization parameters based on the initial vector; A third generating module is used to generate an initial dose distribution according to the initialization parameters; as well as The fourth generating module is configured to adjust the initial dose distribution based on a plurality of intelligent agents to generate a target dose distribution, wherein a single intelligent agent is responsible for at least a portion of the adjustment of the initial dose distribution.

10. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, It is characterized in that the one or more processors execute the one or more computer programs to implement the method according to any one of claims 1 to 8.

Images