Radiotherapy collision avoidance method and system based on three-dimensional scanning and real-time simulation

Abstract

Embodiments of the present description provide a radiotherapy collision avoidance method and system based on three-dimensional scanning and real-time simulation. The method comprises: on the basis of body surface data of a patient, generating an initial three-dimensional model; and on the basis of the initial three-dimensional model, performing collision avoidance simulation during radiotherapy process, and determining a first planned path during radiotherapy.

Description

A Collision Avoidance Method and System for Radiotherapy Based on 3D Scanning and Real-Time Simulation

[0001] This application claims priority to Chinese Patent Application No. 202411777099.6, filed on December 5, 2024, which is incorporated herein by reference in its entirety. Technical Field

[0002] This application relates to the field of collision avoidance in radiotherapy, and for example to a method and system for collision avoidance in radiotherapy based on three-dimensional scanning and real-time simulation. Background Technology

[0003] During radiotherapy, preventing collisions between patients, treatment beds, treatment heads, and equipment within the treatment room is crucial. However, conventional collision avoidance methods have limitations. For example, collision avoidance technologies based solely on CT scans or cameras fail to adequately consider the patient's overall posture, particularly the hands and feet, as well as the impact of supporting structures. Therefore, this application provides a radiotherapy collision avoidance method and system based on three-dimensional scanning and real-time simulation. Summary of the Invention

[0004] This application provides a collision avoidance method for radiotherapy based on three-dimensional scanning and real-time simulation. The method includes: generating an initial three-dimensional model based on the patient's body surface data, wherein the initial three-dimensional model includes at least the patient's three-dimensional information; and performing a collision avoidance simulation during radiotherapy based on the initial three-dimensional model to determine a first planned path during the radiotherapy process.

[0005] This application provides a radiotherapy collision avoidance system based on three-dimensional scanning and real-time simulation. The system includes a model generation module and a simulation module. The model generation module is configured to generate an initial three-dimensional model based on the patient's body surface data. The initial three-dimensional model includes at least the patient's three-dimensional information. The simulation module is configured to perform a simulation based on the initial three-dimensional model to determine a first planned path.

[0006] This application provides a computer-readable storage medium that stores computer instructions. When a computer reads the computer instructions from the storage medium, the computer executes the method described in the above embodiments. Attached Figure Description

[0007] This application will be further described below with reference to the accompanying drawings and specific embodiments.

[0008] Figure 1 is a schematic diagram of a radiotherapy anti-collision system based on three-dimensional scanning and real-time simulation, according to an embodiment of this specification.

[0009] Figure 2 is an exemplary flowchart of a radiotherapy collision avoidance method based on three-dimensional scanning and real-time simulation, according to an embodiment of this specification.

[0010] Figure 3 is an exemplary schematic diagram of a risk prediction model according to an embodiment of this specification;

[0011] Figure 4 is an exemplary flowchart illustrating the process of obtaining the second planned path according to an embodiment of this specification;

[0012] Figure 5 is an exemplary flowchart of determining a first planned path according to an embodiment of this specification. Detailed Implementation

[0013] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. The accompanying drawings do not represent all implementation methods.

[0014] It should be understood that the terms "system," "device," "unit," and / or "module" used herein are a method of distinguishing different components, elements, parts, sections, or assemblies at different levels. If other terms can achieve the same purpose, they may be replaced by other expressions.

[0015] Unless the context clearly indicates an exception, words such as "a," "an," "a kind," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0016] In the embodiments of this specification, the order of the steps described in the step-by-step instructions is interchangeable unless otherwise specified, and steps may be omitted. Other steps may also be included in the operation process.

[0017] Figure 1 is a schematic diagram of a radiotherapy collision avoidance system based on three-dimensional scanning and real-time simulation, according to an embodiment of this specification.

[0018] In this embodiment, the radiotherapy collision avoidance system 100 based on three-dimensional scanning and real-time simulation includes a model generation module and a simulation module.

[0019] The model generation module is configured to generate an initial 3D model based on the patient's body surface data.

[0020] The simulation module is configured to perform collision avoidance simulations during radiotherapy based on the initial 3D model, and to determine the first planned path during the radiotherapy process.

[0021] In one possible implementation, the radiotherapy collision avoidance system 100 based on 3D scanning and real-time simulation also includes a processor and a memory.

[0022] The processor is configured to process data from one or more modules or external data sources of the radiotherapy collision avoidance system 100 based on 3D scanning and real-time simulation. In one possible implementation, the processor includes a central processing unit, an application-specific integrated circuit, an image processing unit, a controller, or any combination thereof.

[0023] The memory is configured to store data, instructions, and / or any other information. In one possible implementation, the memory includes mass storage, removable memory, or any combination thereof.

[0024] For a detailed explanation of the foregoing, please refer to the relevant descriptions in Figures 2 to 5.

[0025] It should be understood that the radiotherapy collision avoidance system and its modules based on 3D scanning and real-time simulation shown in Figure 1 can be implemented in various ways. It should be noted that the above description of the system and its modules is for convenience only and should not limit this specification to the scope of the embodiments described. It is understood that those skilled in the art, after understanding the principle of the system, may arbitrarily combine the various modules or construct subsystems connected to other modules without departing from this principle. In one possible implementation, the model generation module 110 and simulation module 120 disclosed in Figure 1 can be different modules within the same system, or one module can implement the functions of two or more of the above-mentioned modules. For example, the modules can share a storage module, or each module can have its own storage module. Such variations are all within the scope of protection of this specification.

[0026] Figure 2 is an exemplary flowchart illustrating a radiotherapy collision avoidance method based on three-dimensional scanning and real-time simulation according to an embodiment of this specification. In this embodiment, process 200 is executed by a radiotherapy collision avoidance system based on three-dimensional scanning and real-time simulation (hereinafter referred to as the collision avoidance system).

[0027] In one possible implementation, the collision avoidance system generates an initial three-dimensional model 220 based on the patient's body surface data 210, and performs a collision avoidance simulation during radiotherapy based on the initial three-dimensional model 220 to determine the first planned path 230 during radiotherapy.

[0028] For details regarding the collision avoidance system, please refer to the corresponding description in Figure 1.

[0029] Body surface data refers to data related to the patient's body surface. This includes, for example, one or more of the patient's body type, height, dimensions of various body parts, and posture. Patients include those who require radiation therapy.

[0030] In one possible implementation, the collision avoidance system is communicatively connected to a data acquisition device, which acquires surface data.

[0031] In one possible implementation, the data acquisition device can be deployed around the patient or at any feasible location, including one or more of a 3D scanning device, an optical camera, and an ultrasound scanning device. The 3D scanning device includes lasers or photosensors, etc.

[0032] Body surface data acquired by a 3D scanning device is represented as a three-dimensional point cloud. Body surface data acquired by an optical camera is represented as an image. Body surface data acquired by an ultrasonic scanning device is represented as ultrasonic wave reflection data.

[0033] The initial 3D model refers to the 3D model obtained based on body surface data. The initial 3D model includes the patient's 3D information, which refers to the information in the 3D model corresponding to the body surface data. For example, 3D information includes the dimensions of the 3D model corresponding to the patient's body shape.

[0034] In one possible implementation, the collision avoidance system generates an initial 3D model based on body surface data using any feasible modeling method (such as factorization, neural network, etc.) and / or external modeling software.

[0035] In one possible implementation, the collision avoidance system can also acquire structural data of the supporting structure that supports the patient, and generate an initial three-dimensional model based on the body surface data and structural data.

[0036] Support structures are structures used to support or come into contact with the patient's body during radiotherapy. Examples include the treatment bed and equipment connected to the patient's body surface.

[0037] Structural data refers to data related to the physical dimensions of the supporting structure. Examples include one or more of the following: the dimensions of the supporting structure, its relative position to the patient, etc.

[0038] In one possible implementation, the collision avoidance system acquires structural data via a data acquisition device. The data format of the structural data is similar to that of the surface data.

[0039] In one possible implementation, the collision avoidance system generates an initial 3D model based on structural and surface data using any feasible modeling method and / or external modeling software. If the collision avoidance system generates the initial 3D model based on structural and surface data, the 3D information includes information from the 3D model that corresponds to the structural data. For example, the dimensions of the 3D model corresponding to the supporting structure.

[0040] In one possible implementation, the patient-related support structures are also considered when building the initial 3D model, thereby constructing a more accurate initial 3D model, which helps to improve the realism and reliability of subsequent simulations.

[0041] The planned path refers to the path that directs the movement of the treatment equipment. The first planned path is the planned path set before radiotherapy begins. The treatment equipment refers to the equipment that performs radiotherapy (such as a CT scanner). The collision avoidance system is communicatively connected to the treatment equipment.

[0042] In one possible implementation, the collision avoidance system generates motion control commands based on a first planned path and sends the motion control commands to the treatment device, controlling the treatment device to move along the first planned path after the start of radiotherapy to perform radiotherapy on the patient.

[0043] In one possible implementation, the collision avoidance system uses surface data and structural data as its basis. It queries a path preset table to find reference paths corresponding to the surface and structural data, and uses these reference paths as the first planned path. The path preset table is pre-set based on historical data and includes multiple sets of surface and structural data, as well as the reference path for each set.

[0044] The collision avoidance system incorporates historical surface and structural data from past radiotherapy procedures into a path preset table, and uses the first planned path from which no collisions actually occurred during past radiotherapy procedures as a reference path into the path planning table. The reference path includes the coordinates of multiple spatial points arranged in sequence.

[0045] In one possible implementation, the collision avoidance system performs a collision avoidance simulation during radiotherapy based on an initial 3D model to determine a first planned path.

[0046] Simulation refers to the simulation of the movement of therapeutic equipment.

[0047] In one possible implementation, to prevent collisions between the patient and / or supporting structure and the treatment equipment during radiotherapy, the collision avoidance system simulates multiple candidate first paths using simulation software, and selects any one of the candidate first paths from those that do not result in a collision as the first planned path. The simulation software can include any feasible simulation software, such as built-in system simulation software or external simulation software (e.g., ADAMS, SolidWorks, etc.).

[0048] A candidate first path refers to a planned path to be determined as the first planned path. In one possible implementation, the collision avoidance system obtains multiple candidate first paths through user input or other means. For example, a candidate first path includes paths connecting multiple spatial points arranged in sequence, where the spatial points in different candidate first paths are located differently or partially the same. Users include doctors, etc. Users can manually set multiple planned paths as candidate first paths based on the patient's treatment needs. Treatment needs include one or more of the following: the site requiring radiotherapy, the required radiation dose or duration for each site.

[0049] In one possible implementation, the process of the collision avoidance system simulating based on a single candidate first path includes: the collision avoidance system sets an initial 3D model in the simulation software, and sets the motion path of the virtual device based on the candidate first path, so that the virtual device performs virtual motion in the simulation software, and records whether the virtual device collides with the initial 3D model during the virtual motion. The virtual device refers to the model used in the simulation software to simulate the treatment equipment.

[0050] In one possible implementation, the collision avoidance system may also perform one or more simulations based on the initial 3D model and the target path library, and determine a first planned path based on the simulation results of one or more simulations. See Figure 5 and its related description for an explanation of this part. Figure 5 is an exemplary flowchart illustrating the determination of the first planned path according to an embodiment of this specification.

[0051] In one possible implementation, a 3D model accurately reflecting the patient's body surface data is constructed based on the patient's complete body surface data. This effectively eliminates the influence of errors in the body surface data, thereby significantly improving the accuracy and efficiency of path planning. Through simulation, potential collisions that may occur during the movement of the treatment device can be effectively predicted, allowing for adjustments to the device's movement path and reducing the likelihood of actual collisions with the patient.

[0052] In one possible implementation, the collision avoidance system can also acquire real-time scan data of the patient during treatment, update the initial 3D model based on the real-time scan data to obtain a real-time 3D model, and perform simulation based on the real-time 3D model.

[0053] Real-time scan data refers to the patient's body surface data acquired during radiotherapy. When the patient's posture or other factors change during radiotherapy, the real-time scan data will not match the body surface data acquired before the start of radiotherapy.

[0054] In one possible implementation, the real-time scan data may also include structural data acquired during radiotherapy.

[0055] In one possible implementation, the data acquisition device can continuously collect surface and structural data as real-time scan data during radiotherapy and send it to the anti-collision system.

[0056] A real-time 3D model refers to a 3D model obtained by updating an initial 3D model.

[0057] In one possible implementation, the collision avoidance system updates the initial 3D model based on real-time scan data using real-time 3D reconstruction technology (such as BundleFusion) to obtain a real-time 3D model.

[0058] For example, for real-time scanning data acquired by an optical camera, the collision avoidance system can preprocess the real-time scanning data and extract information related to the patient and / or supporting structure from the image using image recognition and processing technologies. This information is then converted into three-dimensional information, and the initial three-dimensional model is updated using real-time three-dimensional reconstruction technology (such as BundleFusion) to obtain a real-time three-dimensional model. Preprocessing includes one or more of the following: noise removal, resolution adjustment, and region of interest cropping. The region of interest includes areas of the patient requiring radiotherapy.

[0059] In one possible implementation, after obtaining the real-time 3D model, the collision avoidance system can update the real-time 3D model again based on subsequently acquired real-time scanning data, using the method described above for updating the initial 3D model.

[0060] In one possible implementation, after obtaining a real-time 3D model, the collision avoidance system can perform a simulation again based on the real-time 3D model to determine whether a collision may occur between the patient and / or the supporting structure and the treatment equipment.

[0061] In one possible implementation, by monitoring relevant information about the patient and / or supporting structure in real time and updating the initial 3D model based on the obtained information, the accuracy of the 3D model can be effectively improved, thereby increasing the precision of the simulation and facilitating the judgment of subsequent operations.

[0062] In one possible implementation, the collision avoidance system can also determine the first collision risk based on the real-time 3D model and the first planned path, and generate and issue a warning message in response to the first collision risk meeting the warning conditions.

[0063] The first collision risk refers to the collision risk corresponding to the first future time. Collision risk is used to characterize the probability of a collision between the treatment device and the patient and / or supporting structure when the device is in motion. In one possible implementation, collision risk is expressed numerically, with a higher numerical value indicating a higher collision risk.

[0064] The first future time refers to the future time after obtaining the real-time 3D model.

[0065] In one possible implementation, the first future time includes one or more future time points. The number of future time points is related to the motion characteristics of the treatment device.

[0066] Motion features are used to characterize the motion of the therapeutic device. In one possible implementation, motion features include one or more of motion velocity, motion acceleration, etc. The number of future time points can be positively correlated with motion velocity or motion acceleration; the greater the motion velocity or acceleration, the more future time points there are.

[0067] In one possible implementation, the number of future time points is related to the motion characteristics of the treatment device. This allows for the determination of more future time points when the treatment device is operating at high speeds, thereby predicting collision risks over a longer period and improving the safety of radiotherapy.

[0068] In one possible implementation, the collision avoidance system determines the first collision risk based on a real-time 3D model and a first planned path through various methods. For example, the collision avoidance system performs a simulation again based on the real-time 3D model and the first planned path, counting the number of collisions that occur during the simulation; the more collisions that occur, the greater the first collision risk.

[0069] For example, a collision avoidance system transforms and / or fuses the real-time 3D model and the first planned path into feature vectors to obtain path vectors. These path vectors are then used as cluster centers to cluster multiple historical path vectors. The first collision risk is determined based on the set of vectors corresponding to the cluster centers. Specifically, the collision avoidance system can extract a preset number of historical path vectors from the vector set, obtain the historical collision results corresponding to these historical path vectors from historical data, calculate the ratio of the number of collisions in the historical collision results to the preset number, and use this ratio as the first collision risk. The preset data is pre-set based on historical experience.

[0070] The historical path vector refers to the feature vector obtained by the collision avoidance system through feature vector transformation and / or fusion of the historical real-time 3D model and the historical first planned path. The historical collision result corresponding to the historical path vector refers to the actual collision result that occurred during the historical radiotherapy process. The collision result includes collisions that occurred and collisions that did not occur.

[0071] In one possible implementation, the collision avoidance system can also determine the first collision risk at each future time point in the first future time period based on a real-time 3D model and a first planned path, using a risk prediction model. See Figure 3 and its related description for details. Figure 3 is an exemplary schematic diagram of the risk prediction model shown in an embodiment of this specification.

[0072] Warning conditions refer to the criteria used to determine whether to generate and issue a warning. In one possible implementation, warning conditions are pre-set based on historical experience. For example, the risk of a first collision may exceed a risk threshold. The risk threshold is pre-set based on historical experience.

[0073] In one possible implementation, if the collision avoidance system determines the first collision risk at each future time point in the first future time through a risk prediction model, then the warning conditions include the average value of the first collision risks at multiple future time points being greater than a risk threshold, etc.

[0074] Warning information refers to information related to the content of the warning. For example, it might alert a user to a high risk of initial collision or remind a patient not to move unnecessarily. In one possible implementation, warnings can be issued through any feasible method, such as an audible alert.

[0075] In one possible implementation, collision risks in the future are dynamically calculated using a real-time 3D model and a first planned path. This improves the accuracy and timeliness of collision risk assessment, and issuing early warnings helps to take timely measures to avoid collisions, ensuring the safety of patients and treatment equipment.

[0076] In one possible implementation, the collision avoidance system can also modify the first planned path based on the real-time 3D model to obtain a second planned path, and determine the second collision risk based on the real-time 3D model and the second planned path.

[0077] In one possible implementation, in response to the second collision risk meeting the warning conditions, the collision avoidance system generates a warning message and issues a warning, and / or updates the second planned path.

[0078] The second planning path refers to the path that instructs the movement of the treatment equipment after obtaining the real-time 3D model.

[0079] In one possible implementation, the collision avoidance system generates motion control commands based on a second planned path and sends the motion control commands to the treatment device to control the treatment device to move along the second planned path to perform radiotherapy on the patient.

[0080] In one possible implementation, the collision avoidance system, based on a real-time 3D model, refines the first planned path to obtain multiple candidate second paths, and performs simulation on each candidate second path to obtain the second planned path. Here, a candidate second path refers to a planned path to be determined as the second planned path. The collision avoidance system can select any one of the candidate second paths that did not result in a collision in the simulation as the second planned path.

[0081] In one possible implementation, the collision avoidance system modifies the first planned path in several ways. For example, the collision avoidance system obtains the modified first planned path (i.e., candidate second paths) through user input. The user can manually modify the first planned path to generate multiple candidate second paths and input them into the collision avoidance system.

[0082] For example, a collision avoidance system can divide a first planned path into multiple spatial points based on a preset distance, and randomly adjust the coordinates of one or more of these spatial points to correct the first planned path and obtain multiple candidate second paths. The preset distance is pre-set based on historical experience.

[0083] The second collision risk refers to the collision risk corresponding to the second future time. The second future time refers to the future time after the second planned path is obtained. In one possible implementation, the second future time is later than or equal to the first future time. The second future time includes one or more future time points.

[0084] In one possible implementation, the method by which the collision avoidance system determines the second collision risk is similar to the method by which it determines the first collision risk, and its implementation method can be found in the method for determining the first collision risk.

[0085] In one possible implementation, the warning conditions also include a second collision risk exceeding a risk threshold. In response to the second collision risk meeting the warning conditions, the collision avoidance system generates a warning message and issues a warning and / or updates the second planned path. Furthermore, after obtaining the updated second planned path, the collision avoidance system can further determine the second collision risk corresponding to the updated second planned path, and in response to the new second collision risk meeting the warning conditions, update the second planned path again.

[0086] In one possible implementation, the collision avoidance system updates the second planned path in several ways. For example, the collision avoidance system updates the second planned path using the methods described above for determining the second planned path. Another example is that the collision avoidance system, based on the second planned path, directionally adjusts the second planned path and performs simulation on the adjusted second planned path. If no collision occurs in the simulation, the adjusted second planned path is used as the new second planned path; if a collision occurs in the simulation, the second planned path is directionally adjusted again.

[0087] In one possible implementation, the directional adjustment of the second planned path includes: the collision avoidance system dividing the second planned path into multiple spatial points based on a preset distance, and adjusting the coordinates of spatial points prone to collision towards the direction away from the patient, thereby obtaining the adjusted second planned path. The adjustment range is preset based on historical experience. The collision avoidance system can identify spatial points that are too close to or overlap with the real-time 3D model as potentially prone to collisions.

[0088] In one possible implementation, the first planned path is modified based on a real-time 3D model to obtain a second planned path, thereby enabling dynamic adjustment of the movement path of the treatment equipment during radiotherapy, which improves the accuracy, safety, and intelligence of the radiotherapy process.

[0089] Figure 3 is an exemplary schematic diagram of a risk prediction model according to some embodiments of this specification.

[0090] In one possible implementation, the collision avoidance system determines a first collision risk 350 based on a real-time 3D model 230 and a first planned path 310 using a risk prediction model 340. For an explanation of the real-time 3D model, the first planned path, and the first collision risk, please refer to Figure 2 and its related description.

[0091] A risk prediction model is a model used to determine the risk of a first collision. In one possible implementation, the risk prediction model can be a machine learning model. For example, the risk prediction model can include any one or a combination of Convolutional Neural Networks (CNN) models, Neural Networks (NN) models, or other custom model structures.

[0092] In one possible implementation, the risk prediction model takes into account a real-time 3D model and a first planned path as input, and outputs the first collision risk at each future time point in the first future time.

[0093] In one possible implementation, the collision avoidance system trains a risk prediction model based on a large number of first training samples with first labels, using methods such as gradient descent. The first training samples include real-time 3D models of the samples and first planned paths for the samples. The first label of the first training sample can be the actual collision risk of each sample at a future time point in the first future time of the sample corresponding to the first training sample.

[0094] In one possible implementation, the first training sample and the first label are determined based on historical data. For example, the collision avoidance system uses historical real-time 3D models and historical first planned paths from historical radiotherapy processes as the first training samples. For the first training samples where no collision actually occurred, the labels for all time points corresponding to the first training sample are set to 0. For the first training samples where a collision actually occurred, the label for the time point where the collision occurred is set to 1, and the labels for the time points where no collision occurred are set to values ​​between 0 and 1. The closer the time point where no collision occurred is to the time point where a collision occurred, the closer its label is to 1.

[0095] In one possible implementation, the risk prediction model can be trained as follows: Multiple first training samples with a first label are input into the initial risk prediction model; a loss function is constructed using the first label and the prediction results of the initial risk prediction model; the initial risk prediction model is iteratively updated based on the loss function; and the risk prediction model training is complete when the loss function of the initial risk prediction model satisfies a preset condition. The preset condition could be loss function convergence, the number of iterations reaching a set value, etc.

[0096] In one possible implementation, the inputs to the risk prediction model also include the number of data blind spots (320) and the activity level of the corresponding part (330) for each data blind spot.

[0097] A data blind spot refers to a patient's body part where the data acquisition device cannot collect surface data. For example, a patient's body part that is obscured by the treatment device during its movement.

[0098] In one possible implementation, the collision avoidance system determines data blind spots based on real-time scan data. For example, if the data acquisition device is an optical camera, the collision avoidance system uses image recognition and processing technologies to identify body parts of the patient that are obscured by the treatment equipment, and designates these body parts as data blind spots. Alternatively, if the data acquisition device is a 3D scanning device, since there are differences between the 3D point clouds obtained from scanning the human body and scanning the treatment equipment, the collision avoidance system can directly determine body parts of the patient that are obscured by the treatment equipment based on real-time scan data and designate these body parts as data blind spots.

[0099] Site activity refers to the activity level of the site corresponding to the patient's data blind spot. The site activity level for each data blind spot is determined separately. See Figure 4 and its related description for an explanation of how site activity is determined.

[0100] In one possible implementation, if the input to the risk prediction model also includes the number of data blind spots and the activity level of the corresponding part for each data blind spot, then the first training sample also includes the number of sample data blind spots and the activity level of the corresponding sample part for each sample data blind spot.

[0101] In one possible implementation, the activity level of the data blind zone helps to reflect the movement tendency of the corresponding part of the patient's body in the absence of data support. Incorporating the activity level of the data blind zone into the input of the risk prediction model helps to improve the accuracy of the output first collision risk.

[0102] In one possible implementation, by processing data such as the real-time 3D model and the first planned path through a risk prediction model, the self-learning capability of the machine learning model can be utilized to find patterns in a large amount of data, obtain the correlation between the real-time 3D model and the first planned path and the first collision risk, and improve the accuracy and efficiency of determining the user's first collision risk.

[0103] Figure 4 is an exemplary flowchart illustrating the process of obtaining a second planned path according to some embodiments of this specification. In one possible implementation, process 400 is executed by a radiotherapy anti-collision system based on 3D scanning and real-time simulation (hereinafter referred to as the anti-collision system).

[0104] Step 410: Determine one or more security buffers.

[0105] A safety buffer is a space reserved to prevent collisions between patients and treatment equipment. For example, a safety buffer is the space obtained by extending a certain distance outward from a local contour of a 3D model; therefore, the size of the safety buffer can be represented by the outward extension distance.

[0106] In one possible implementation, the collision avoidance system determines one or more safety buffers in various ways. For example, the collision avoidance system sets up safety buffers based on model boundaries and a preset space size. Here, the model boundaries refer to the boundary points of the initial 3D model or the real-time 3D model, and the edges formed by connecting these boundary points. The preset space size is pre-set based on historical experience, such as a space 5cm outside the model boundaries.

[0107] In one possible implementation, the collision avoidance system can also determine one or more safety buffer zones based on the regional characteristics of the data blind zone. See Figure 3 and its related description for an explanation of the data blind zone.

[0108] Region features refer to information related to data blind spots. In one possible implementation, region features include patient body parts corresponding to the data blind spot. Region features are determined simultaneously by the collision avoidance system during the process of identifying data blind spots.

[0109] For example, the collision avoidance system, based on the regional characteristics of the data blind zone, queries a preset feature table for reference regional features that are identical to the regional features, and determines the reference buffer corresponding to the reference regional features as the safety buffer corresponding to the data blind zone. Each data blind zone corresponds to one safety buffer.

[0110] The preset feature table is pre-set based on historical data and includes multiple reference area features and a reference buffer corresponding to each reference area feature. In one possible implementation, the collision avoidance system includes historical area features of historical data blind spots in historical radiotherapy processes as reference area features in the preset feature table, and also includes historical safety buffers (such as a 30cm space extending outward from the data blind spot) in the preset feature table.

[0111] In one possible implementation, the region feature also includes the area of ​​the data blind zone. The region area refers to the area occupied by the part corresponding to the data blind zone. The collision avoidance system can also determine one or more safety buffers based on the region area. For example, the larger the region area, the greater the outward extension distance of the safety buffer corresponding to the data blind zone.

[0112] In one possible implementation, the larger the data blind zone, the higher the potential collision risk. Increasing the size of the safety buffer corresponding to a larger data blind zone can help further reduce the potential collision risk.

[0113] In one possible implementation, different safety buffers are set according to different data blind spots, which improves the flexibility of ensuring the safety of radiotherapy.

[0114] In one possible implementation, the collision avoidance system can also determine the activity level of the corresponding parts in the data blind zone based on regional features, and determine one or more safety buffers based on the regional features and the activity level of the parts. The safety buffer corresponding to each data blind zone is determined based on the regional features and the activity level of the parts in that data blind zone.

[0115] In one possible implementation, the collision avoidance system determines the frequency and amplitude of activity of the patient's body parts before they appear in the data blind zone based on real-time scan data of the patient's body parts corresponding to the regional features. It then calculates a weighted sum of the activity frequency and amplitude, and uses this weighted sum as the activity level of the body part in the data blind zone. The activity frequency and amplitude can be determined using any feasible method, such as image comparison or changes in 3D point cloud data.

[0116] In one possible implementation, the collision avoidance system determines one or more safety buffers by querying a preset area table based on regional features and part activity levels. The preset area table is pre-set based on historical experience and includes multiple sets of regional features and part activity levels, as well as the outer extension distance of the safety buffer corresponding to each set of features and part activity levels. The outer extension distance of the safety buffer is positively correlated with the area of ​​the data blind zone and the part activity level corresponding to the data blind zone.

[0117] In one possible implementation, the collision avoidance system determines the outer extension distance of the safety buffer corresponding to the data blind zone based on regional characteristics and location activity through a buffer prediction model.

[0118] A buffer prediction model is a model used to determine the outer distance of a safety buffer. In one possible implementation, the buffer prediction model can be a machine learning model. For example, the buffer prediction model can include any one or a combination of Convolutional Neural Networks (CNN) models, Neural Networks (NN) models, or other custom model structures.

[0119] In one possible implementation, the collision avoidance system is based on a large number of second training samples with second labels, and a buffer prediction model is trained using methods such as gradient descent. The second training samples include sample region features and sample location activity, and the second label of the second training sample can be the actual outward expansion distance of the safety buffer corresponding to the second training sample.

[0120] In one possible implementation, the second training sample and the second label are determined based on historical data. For example, the collision avoidance system statistically analyzes historical radiotherapy processes that did not involve collisions and had high radiotherapy efficiency. The regional characteristics and site activity corresponding to the historical data blind spots in these historical radiotherapy processes are used as the second training sample, and the actual outward extension distance of the safety buffer zone used in these historical radiotherapy processes is used as the second label. Radiotherapy efficiency can be represented by radiotherapy time, etc.; the shorter the radiotherapy time, the higher the radiotherapy efficiency.

[0121] In one possible implementation, the training process of the buffer prediction model is similar to that of the risk prediction model, and its implementation method is described in the training process of the risk prediction model.

[0122] In one possible implementation, the higher the activity level of a part, the higher the potential collision risk. Based on the activity level of the part, the size of the safety buffer corresponding to the data blind zone can be determined in multiple ways, which helps to improve the efficiency of determining the safety buffer and further reduce the potential collision risk.

[0123] In one possible implementation, the collision avoidance system performs step 420 based on the real-time 3D model 310 and one or more safety buffers obtained in step 410 to obtain a second planned path 430.

[0124] Step 420: Modify the first planned path.

[0125] In one possible implementation, the collision avoidance system fuses the safety buffer with the real-time 3D model using any feasible method such as point cloud union to obtain a fused 3D model. Based on the fused 3D model, the first planned path is corrected to obtain a second planned path. The method for correcting the first planned path based on the fused 3D model is similar to the method for correcting the first planned path based on the real-time 3D model; its implementation can be found in Figure 2 and its related description.

[0126] In one possible implementation, by reserving a safety buffer zone, it is ensured that even if the patient makes a slight accidental movement during radiotherapy, a collision will not occur, thus improving the safety of the radiotherapy process.

[0127] Figure 5 is an exemplary flowchart illustrating the determination of a first planned path according to some embodiments of this specification. In one possible implementation, process 500 is executed by a radiotherapy anti-collision system based on 3D scanning and real-time simulation (hereinafter referred to as the anti-collision system).

[0128] Step 510: Determine the target path library.

[0129] A target landmark database refers to a database containing multiple planned paths. In one possible implementation, the collision avoidance system stores multiple target path databases, each corresponding to a type of treatment need. The collision avoidance system determines the corresponding type of target path database based on the current patient's treatment needs. See Figure 2 and its related description for an explanation of treatment needs.

[0130] In one possible implementation, the construction process of the target path library includes: for each treatment need, the collision avoidance system statistically analyzes a large amount of historical data on historical treatment needs that are the same as or similar to the treatment need; the historical radiotherapy processes corresponding to these historical treatment needs are identified as target historical data; the target historical data with better radiotherapy effects are identified as preferred historical data; and the planned paths used in the preferred historical data are added to the target path library as candidate planned paths. The radiotherapy effects are manually labeled by the user based on historical experience.

[0131] The collision avoidance system determines the similarity between treatment needs and historical treatment needs by converting them into feature vectors and calculating vector similarity. Vector similarity is negatively correlated with vector distance, which includes Euclidean distance, etc.

[0132] Step 520: Perform one or more simulations based on the initial 3D model and the target path library.

[0133] In one possible implementation, the process of a single simulation includes:

[0134] S1: The collision avoidance system selects a candidate path from the target path library based on a preset selection order, which is then used as the target path for this simulation. The preset selection order includes selecting the candidate path with the highest historical usage frequency among those that have not been selected before. Here, a candidate path that has not been selected before refers to a candidate path that has not been selected in the current simulation or multiple simulations. The historical usage frequency is determined by the collision avoidance system based on historical data.

[0135] For example, the current simulation has already been performed once, referred to as the first simulation. This simulation is the second simulation. If the target path library includes 5 paths (referred to as paths 1 to 5, and the historical usage counts of paths 1 to 5 are sorted from high to low), then the target planning path selected in the first simulation is path 1, the target planning path selected in the second simulation is path 2, and the unselected candidate planning paths include paths 3 to 5.

[0136] S2: Based on the target planning path and the initial 3D model, determine the test collision risk corresponding to the target planning path. The test collision risk is used to characterize the probability of the treatment device colliding with the patient and / or supporting structure when moving along the target planning path. The method for determining the test collision risk is similar to the method for determining the first collision risk, and its implementation method is shown in Figure 2 and its related description.

[0137] S3: In response to the test collision risk not meeting the termination condition, proceed to the next simulation. The termination condition includes the test collision risk being no less than the risk threshold. See Figure 2 and its related description for an explanation of the risk threshold. Figure 2 is an exemplary flowchart illustrating a radiotherapy collision avoidance method based on three-dimensional scanning and real-time simulation according to an embodiment of this specification.

[0138] S4: When the collision risk corresponding to the target planned path meets the termination condition, one or more simulations are terminated.

[0139] Step 530: Determine the first planned path based on the simulation results of one or more simulations.

[0140] For further explanation of the first planned path, please refer to Figure 2 and its related description.

[0141] In one possible implementation, the collision avoidance system determines the candidate planning path with the lowest test collision risk from one or more simulations as the first planning path. The simulation results include the test collision risk determined in each simulation.

[0142] In one possible implementation, a database of target pathways for the patient's treatment needs is combined to comprehensively evaluate and determine the optimal first planning pathway, thereby improving the efficiency of determining the first planning pathway.

[0143] It should be noted that the above descriptions of processes 400 and 500 are for illustrative purposes only and do not limit the scope of this specification. Those skilled in the art can make various modifications and changes to the processes under the guidance of this specification. However, these modifications and changes remain within the scope of this specification.

[0144] In one possible implementation, a computer-readable storage medium is also provided, which stores computer instructions. When a computer reads the computer instructions from the storage medium, the computer executes the method described in any of the above embodiments.

[0145] Furthermore, certain features, structures, or characteristics in one or more embodiments of this specification may be appropriately combined.

[0146] The embodiments use numbers describing the quantity of components and attributes. It should be understood that such numbers used in the embodiments are sometimes modified by the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in one possible implementation, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In one possible implementation, the numerical parameters should take into account the prescribed significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of range in some embodiments of this specification are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0147] If there is any inconsistency or conflict between the descriptions, definitions, and / or terms used in the materials referenced in this specification and the content described in this specification, the descriptions, definitions, and / or terms used in this specification shall prevail.

Claims

1. A method for radiotherapy collision avoidance based on three-dimensional scanning and real-time simulation, the method comprising: generating an initial three-dimensional model based on surface data of a patient, the initial three-dimensional model comprising at least three-dimensional information of the patient; performing a simulation of collision avoidance during a radiotherapy procedure based on the initial three-dimensional model to determine a first planned path during the radiotherapy procedure.

2. The method of claim 1, wherein, The method further comprises: obtaining real-time scanning data of the patient during the procedure; updating the initial three-dimensional model based on the real-time scanning data to obtain a real-time three-dimensional model; performing the simulation based on the real-time three-dimensional model.

3. The method of claim 2, wherein, The method further comprises: determining a first collision risk based on the real-time three-dimensional model and the first planned path, the first collision risk being a collision risk corresponding to a first future time; in response to the first collision risk satisfying an early warning condition, generating early warning information and issuing an early warning.

4. The method of claim 3, wherein, The determining a first collision risk based on the real-time three-dimensional model and the first planned path comprises: determining the first collision risk based on the real-time three-dimensional model and the first planned path by a risk prediction model, the risk prediction model being a machine learning model.

5. The method of claim 4, wherein, The input of the risk prediction model further comprises a number of data blind areas and a part activity corresponding to each data blind area, the part activity being an activity level of a part of the patient in the data blind area.

6. The method of claim 3, wherein, The first future time comprises one or more future time points, the number of the one or more future time points being related to a motion feature of a treatment device.

7. The method of claim 2, wherein, The method further comprises: correcting the first planned path based on the real-time three-dimensional model to obtain a second planned path; determining a second collision risk based on the real-time three-dimensional model and the second planned path, the second collision risk being a collision risk corresponding to a second future time, the second future time being later than or equal to the first future time; in response to the second collision risk satisfying the early warning condition, generating early warning information and issuing an early warning, and / or updating the second planned path.

8. The method of claim 1, wherein, The method further comprises: determining one or more safety buffers; correcting the first planned path based on the real-time three-dimensional model and the one or more safety buffers to obtain a second planned path.

9. The method of claim 8, wherein, The determining one or more safety buffers comprises: determining a data blind area; determining the one or more safety buffers based on a region feature of the data blind area.

10. The method of claim 9, wherein, The region feature comprises a region area of the data blind area, and the determining the one or more safety buffers based on the region feature of the data blind area comprises: determining the one or more safety buffers based on the region area.

11. The method of claim 10, wherein, The determining the one or more safety buffers based on the region feature of the data blind area further comprises: determining a part activity corresponding to the data blind area based on the region feature; determining the one or more safety buffers based on the region feature and the part activity.

12. The method of claim 1, wherein, The method further comprises: obtaining structure data of a support structure supporting the patient; generating the initial three-dimensional model based on the surface data and the structure data, the initial three-dimensional model further comprising the structure data.

13. The method of claim 1, wherein, The simulation of anti-collision in the radiotherapy process based on the initial three-dimensional model comprises: determining a target path library; performing one or more simulation simulations based on the initial three-dimensional model and the target path library, each simulation simulation process comprising: selecting a candidate planning path from the target path library as the target planning path of this simulation simulation based on a preset selection order; determining the test collision risk corresponding to the target planning path based on the target planning path and the initial three-dimensional model; in response to the test collision risk not meeting the end condition, performing the next simulation simulation; in response to the collision risk corresponding to the target planning path meeting the end condition, ending one or more simulation simulations; determining the first planning path based on the simulation results of one or more simulation simulations.

14. A radiotherapy anti-collision system based on three-dimensional scanning and real-time simulation, the system comprising a model generation module and a simulation simulation module; The model generation module is configured to generate an initial three-dimensional model based on the surface data of a patient, the initial three-dimensional model comprising at least three-dimensional information of the patient; The simulation simulation module is configured to perform simulation of anti-collision in the radiotherapy process based on the initial three-dimensional model, and determine the first planning path in the radiotherapy process.

15. A computer readable storage medium, the storage medium stores computer instructions, when the computer reads the computer instructions in the storage medium, the computer executes the method of any one of claims 1 to 13.

Images