Real-time adaptive radiotherapy management system

Abstract

This application discloses a real-time adaptive radiotherapy management system, relating to the field of medical technology. The system includes: a real-time data processing unit for acquiring real-time data of the target object during radiotherapy; an adaptive decision-making unit for determining a target radiotherapy plan matching the current state of the target object from multiple alternative radiotherapy plans based on the real-time data, wherein the multiple alternative radiotherapy plans are associated with different physiological motion phases or anatomical states of the target object; a plan execution sequence splitting unit for parsing and splitting each alternative radiotherapy plan into one or more hardware execution sequences; and an execution scheduling unit for sending at least one hardware execution sequence corresponding to the target radiotherapy plan to the control system of the radiotherapy equipment. This application solves the technical problem that existing adaptive radiotherapy systems, due to their rigid architecture and singular decision-making mode, cannot achieve highly real-time dynamic plan switching during treatment.

Description

Technical Field

[0001] This application relates to the field of medical technology, particularly to the field of radiotherapy technology, and specifically to a real-time adaptive radiotherapy management system. Background Technology

[0002] Existing technologies, such as adaptive methods based on treatment plans and adaptive modules in commercial treatment planning systems, mainly acquire images before or during treatment intervals, perform plan matching or re-optimization, and then execute the plan.

[0003] However, these solutions generally employ a tightly coupled monolithic design or simple integration of functional modules at the software architecture level, failing to effectively distinguish between real-time control services and non-real-time computing tasks. Furthermore, their decision-making models often rely on single, offline plan selection or optimization, lacking a real-time response mechanism for real-time data during treatment. More importantly, the systems typically treat the treatment plan as a whole, without designing it to be broken down into independently schedulable and rapidly switchable hardware control sequences.

[0004] The aforementioned rigid architecture and singular decision-making model prevent existing systems from achieving highly real-time dynamic plan switching during treatment beam irradiation, making it difficult to meet the urgent clinical demand for real-time adaptive radiotherapy. Summary of the Invention

[0005] This application provides a real-time adaptive radiotherapy management system to at least solve the technical problem that existing adaptive radiotherapy systems cannot achieve highly real-time dynamic plan switching during treatment due to rigid architecture and a single decision-making mode.

[0006] According to one aspect of this application, a real-time adaptive radiotherapy management system is provided, comprising: a real-time data processing unit for acquiring real-time data of a target object during radiotherapy, wherein the real-time data includes real-time image data and / or real-time motion data; an adaptive decision-making unit for determining a target radiotherapy plan matching the current state of the target object from multiple alternative radiotherapy plans based on the real-time data, wherein the multiple alternative radiotherapy plans are associated with different physiological motion phases or anatomical states of the target object; a plan execution sequence splitting unit for parsing and splitting each alternative radiotherapy plan into one or more hardware execution sequences; and an execution scheduling unit for sending at least one hardware execution sequence corresponding to the target radiotherapy plan to the control system of the radiotherapy equipment, wherein the hardware execution sequence includes control parameters for direct execution by the control system of the radiotherapy equipment.

[0007] Optionally, the real-time adaptive radiotherapy management system further includes: a multi-source data processing unit, used to acquire multi-dimensional data including static images, dynamic images, and physiological changes and physiological motion data of the target object, and combine them with clinical goals and constraints to form a comprehensive dataset; a plan generation unit, used to generate multiple alternative radiotherapy plans for the target object based on the comprehensive dataset, wherein the multiple alternative radiotherapy plans are associated with different physiological motion phases or anatomical states of the target object; a plan time alignment unit, used to map and calibrate the multiple alternative radiotherapy plans to a unified time coordinate system according to physiological motion data and the physical execution characteristics of the radiotherapy equipment, so that each alternative radiotherapy plan has relative time coding information; and a plan uploading unit, used to upload and store the multiple time-calibrated alternative treatment plans in the target storage area.

[0008] Optionally, the hardware execution sequence corresponding to the targeted radiotherapy plan includes at least one of the following sequences:

[0009] The multi-leaf collimator control point sequence is a set of instructions arranged sequentially on the time axis to control the position and state of each blade of the multi-leaf collimator. The multi-leaf collimator control point sequence is used to form a shooting field shape that changes over time by controlling the dynamic position of the multi-leaf collimator blades.

[0010] The gantry angle control sequence is a set of instructions arranged sequentially on the time axis for controlling the rotation angle of the radiotherapy equipment gantry. The gantry angle control sequence is used to control the gantry to rotate around the target object to the target angle according to a predetermined trajectory and time point.

[0011] The dose rate control sequence is a set of instructions arranged sequentially on the time axis for controlling the intensity or flux of radiation output from a radiotherapy device. The dose rate control sequence is used to dynamically control the output intensity of the radiation source based on a predetermined time function or event trigger.

[0012] Optionally, the adaptive decision unit includes: an image matching subunit, used to calculate the image similarity between the real-time image data and the image data used when formulating each alternative radiotherapy plan, when the real-time data includes real-time image data; and a first decision subunit, used to select the alternative radiotherapy plan with the highest image similarity as the target radiotherapy plan.

[0013] Optionally, the adaptive decision-making unit includes: a second decision subunit, configured to determine a target radiotherapy plan based on the following first or second operation when real-time motion data is included in the real-time data: the first operation includes: inputting real-time motion data into a neural network model, predicting organ state information corresponding to the real-time motion data based on prior knowledge learned by the neural network model during end-to-end training; obtaining the organ state information used to formulate each alternative radiotherapy plan, calculating the similarity between the similarity and the organ state information predicted by the neural network model, and selecting the alternative radiotherapy plan with the highest similarity as the target radiotherapy plan; the second operation includes: determining the in vivo deformation field of the target object based on the real-time motion data, determining the current organ state information of the target object based on the in vivo deformation field of the target object; and determining the target radiotherapy plan from multiple alternative radiotherapy plans based on the similarity between the organ state information used to formulate each alternative radiotherapy plan and the current organ state information of the target object.

[0014] Optionally, the adaptive decision-making unit includes: a third decision subunit, configured to: determine a first confidence level for each candidate radiotherapy plan based on real-time image data, given that the real-time data includes real-time image data and real-time motion data; determine a second confidence level for each candidate radiotherapy plan based on real-time motion data; perform a weighted calculation on the first and second confidence levels of each candidate radiotherapy plan to obtain a target confidence score for each candidate radiotherapy plan; and select the candidate radiotherapy plan with the highest target confidence score as the target radiotherapy plan.

[0015] Optionally, the execution scheduling unit includes: a phase matching subunit, used to perform phase matching between real-time data and a four-dimensional image sequence taken before radiotherapy of the target object to determine the target phase corresponding to the real-time data; a positioning subunit, used to locate the plan execution start point corresponding to the target phase on the time axis; a subsequence determination subunit, used to extract subsequent subsequences starting from the plan execution start point from the hardware control sequence associated with the target radiotherapy plan; and a sequence switching subunit, used to control the radiotherapy equipment to interrupt the currently executed subsequence and switch to the extracted subsequent subsequence.

[0016] Optionally, the phase matching subunit includes: a first matching module, used to perform phase matching between the real-time image data and a four-dimensional image sequence taken before radiotherapy of the target object when the real-time data includes real-time image data, to obtain a target phase corresponding to the real-time image data; and a second matching module, used to determine the internal deformation field of the target object based on the real-time motion data when the real-time data does not include real-time image data but includes real-time motion data, determine the current organ state information of the target object based on the internal deformation field, and perform phase matching between the current organ state information of the target object and the organ state information described by each image in the four-dimensional image sequence, to obtain a target phase corresponding to the real-time motion data.

[0017] Optionally, the real-time adaptive radiotherapy management system further includes: a data monitoring and feedback unit, used to monitor N monitoring indicators of the target object during the radiotherapy process, and when any monitoring indicator reaches the corresponding preset threshold, to notify the adaptive decision-making unit to perform a re-decision, wherein the re-decision is used to determine the target radiotherapy plan that matches the current state of the target object from multiple alternative radiotherapy plans based on the latest real-time data of the target object, where N is an integer greater than or equal to 1.

[0018] Optionally, the N monitoring indicators include at least one of the following: respiratory cycle, tumor offset, tumor shrinkage rate, and organ-at-risk damage rate.

[0019] Optionally, when determining the target radiotherapy plan, the adaptive decision-making unit is further configured to prioritize the alternative radiotherapy plans based on at least one of the following optimization objectives: a first optimization objective to constrain the shortening of radiotherapy duration; a second optimization objective to constrain the dose distribution of the tumor target area and organs at risk to meet preset dose requirements; and a third optimization objective to serve as a constraint to reduce the radiation dose to normal tissues.

[0020] According to another aspect of this application, a radiotherapy device is also provided, including the aforementioned real-time adaptive radiotherapy management system, and an accelerator treatment head, a multi-leaf collimator, a gantry, and an image acquisition device controlled by the real-time adaptive radiotherapy management system.

[0021] It should be noted that the technical solution described in this application systematically solves the technical problem that existing adaptive radiotherapy systems, due to their rigid architecture and singular decision-making mode, cannot achieve high real-time dynamic plan switching during treatment by constructing a modular dynamic decision-making and execution architecture. For example, by decomposing each alternative radiotherapy plan into a control sequence that can directly drive the hardware through a plan execution sequence splitting unit, it helps to achieve the "decomposability" and "switchability" of the radiotherapy plan, thereby overcoming the interruption delay caused by the overall plan loading. Secondly, by forming a closed-loop feedback system with the real-time data processing unit and the adaptive decision-making unit, the optimal plan can be dynamically selected based on continuously input real-time data, which helps to transform decision-making from a single offline judgment to a continuous online response. Finally, the execution scheduling unit can directly call the pre-generated hardware sequence for distribution, thereby bypassing the cumbersome re-parsing and verification process in the traditional architecture, which is conducive to improving the execution efficiency of the radiotherapy plan. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0023] Figure 1 This is a schematic diagram of an optional real-time adaptive radiotherapy management system according to an embodiment of this application;

[0024] Figure 2 This is a flowchart of an optional adaptive radiotherapy management method according to an embodiment of this application. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] It should also be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0028] According to an embodiment of this application, a real-time adaptive radiotherapy management system is provided, wherein... Figure 1 This is a schematic diagram of an optional real-time adaptive radiotherapy management system according to an embodiment of this application, such as... Figure 1 As shown, the system includes the following units:

[0029] The plan execution sequence splitting unit is used to parse and split each alternative radiotherapy plan for the target object into one or more hardware execution sequences, wherein the hardware execution sequence includes control parameters that are directly executed by the control system of the radiotherapy equipment.

[0030] In some embodiments, the target population includes, but is not limited to, individual patients, and can also extend to patient groups with similar anatomical features or pathological types. Multiple alternative radiotherapy plans can encompass multiple interchangeable radiotherapy regimens generated based on different imaging phases (e.g., different phases of the respiratory cycle), different organ motion states, different dosimetric goals (e.g., priority given to target coverage or priority given to organ-at-risk protection), and different treatment stages (e.g., after tumor regression during treatment). The planning management unit can store, version-manage, and index these multi-dimensional alternative radiotherapy plans through a structured database or distributed file system, providing a dynamically selectable plan library for real-time adaptive decision-making during subsequent treatment.

[0031] It should be noted that the multiple alternative radiotherapy plans for the target object are associated with different physiological motion phases or anatomical states of the target object. In other words, the alternative radiotherapy plans for the target object are not radiotherapy plans for objects similar to the target object (e.g., patients), but rather various radiotherapy plans generated based on the different physiological motion phases of the target object, or various radiotherapy plans corresponding to different anatomical states of the target object. For example, assuming there are N physiological motion phases for the target object, N alternative radiotherapy plans (N is an integer greater than or equal to 1) are generated, and there is a one-to-one correspondence between the N alternative radiotherapy plans and the N physiological motion phases. Assuming there are M anatomical states for the target object, M alternative radiotherapy plans (M is an integer greater than or equal to 1) are generated, and there is a one-to-one correspondence between the M alternative radiotherapy plans and the M anatomical states.

[0032] In some embodiments, the plan execution sequence splitting unit can parse the digital file of the candidate radiotherapy plan to extract control parameters, including the multi-leaf collimator blade position sequence, gantry rotation angle timing, dose rate change curve, and beam on / off time points. These control parameters are then decomposed along a time axis into a discrete, timestamped set of hardware control instructions (i.e., a hardware execution sequence). Each hardware execution sequence is encapsulated as a standardized data packet (e.g., a protocol frame containing instruction type, target value, and execution delay) that can be directly recognized and executed by the radiotherapy device's control system. Furthermore, the same candidate radiotherapy plan can be split into multiple sub-sequences that are independently scheduled during treatment, thereby facilitating real-time switchable mapping from macroscopic treatment plans to microscopic device actions.

[0033] The real-time data processing unit is used to acquire real-time data of the target object during radiotherapy, wherein the real-time data includes real-time image data and / or real-time motion data.

[0034] For example, the real-time data processing unit can continuously acquire structural image data of the treatment area through image acquisition components integrated into the radiotherapy equipment (such as cone-beam CT, megavolt-level electron beam imaging devices), and simultaneously acquire motion trajectory data of surface and internal feature points of the target object through optical surface monitoring systems, electromagnetic tracking beacons, or respiratory gating devices. The real-time data processing unit can also preprocess the acquired raw data (i.e., raw real-time image data and / or raw real-time motion data) (e.g., noise reduction, registration, reconstruction) to generate standardized real-time image sequences and motion parameter streams, and establish a time synchronization mechanism. This facilitates subsequent adaptive decision-making by the adaptive decision-making unit based on dynamic information strictly aligned with the treatment progress.

[0035] An adaptive decision unit is used to determine a target radiotherapy plan that matches the current state of the target object from multiple alternative radiotherapy plans based on real-time data.

[0036] For example, the adaptive decision-making unit can quickly register and calculate the similarity between the image data (such as 2D projection or 3D reconstructed images) provided by the real-time data processing unit at the current treatment moment and the reference images on which each alternative radiotherapy plan is based (e.g., using mutual information or deformation field norm as a metric). The decision logic built into the adaptive decision-making unit will select the alternative radiotherapy plan with the highest similarity score as the target radiotherapy plan. This approach is more suitable for scenarios where changes in anatomical structures are relatively slow or predictable, thus facilitating rapid "state-plan" mapping through a pre-set multi-state plan library.

[0037] For example, the adaptive decision unit can also analyze motion information (such as respiratory phase and tumor displacement) in real-time data, combine it with the current irradiated dose distribution (feedback from the dose monitoring unit), and use a pre-trained dose prediction model to extrapolate the possible final dose outcomes of continuing each alternative plan. The adaptive decision unit can evaluate the extrapolation results based on preset clinical goals (such as minimum target dose and maximum dose to organs at risk) and select the plan that best meets the clinical goals as the target radiotherapy plan. This approach enables dynamic prognostic assessment of the treatment process, helping to ensure that the final decision is based not only on the current state but also on the overall endpoint of the treatment.

[0038] The execution scheduling unit is used to send at least one hardware execution sequence corresponding to the target radiotherapy plan to the control system of the radiotherapy equipment.

[0039] For example, after receiving a switching instruction from the adaptive decision-making unit, the execution scheduling unit can retrieve the sequence segment corresponding to the target radiotherapy plan from the pre-stored hardware execution sequence library. For instance, the execution scheduling unit can directly send the sequence data packet (including at least one hardware execution sequence corresponding to the target radiotherapy plan) to the motion controller and servo system of the radiotherapy equipment through a low-latency communication interface (such as the real-time Ethernet protocol), and use a timestamp synchronization mechanism, which helps the new control instructions take effect at the specified treatment time point or the boundary of the equipment motion cycle, thereby facilitating seamless or micro-interruption switching between the current execution sequence and the target sequence.

[0040] For example, the execution scheduling unit can also maintain a priority queue containing sequences to be distributed. When the adaptive decision unit specifies multiple candidate sequences or multiple parallel treatment tasks exist, the execution scheduling unit can dynamically calculate priorities based on preset rules (such as treatment urgency, sequence execution duration, and equipment load status). For instance, the execution scheduling unit can distribute sequence data packets (including at least one hardware execution sequence corresponding to the target radiotherapy plan) to the corresponding equipment control subsystems in priority order, and continuously monitor the feedback information of the radiotherapy equipment during execution. If an execution deviation or emergency interruption signal is detected in the radiotherapy equipment, the execution scheduling unit will trigger queue rearrangement and sequence re-distribution, thereby improving the robustness and controllability of the treatment execution process.

[0041] In some embodiments, the real-time adaptive radiotherapy management system further includes: a multi-source data processing unit, used to acquire multi-dimensional data including static images, dynamic images, and physiological changes and physiological motion data of the target object, and combine them with clinical goals and constraints to form a comprehensive dataset; a plan generation unit, used to generate multiple alternative radiotherapy plans for the target object based on the comprehensive dataset, wherein the multiple alternative radiotherapy plans are associated with different physiological motion phases or anatomical states of the target object; a plan time alignment unit, used to map and calibrate the multiple alternative radiotherapy plans to a unified time coordinate system according to the physiological motion data and the physical execution characteristics of the radiotherapy equipment, so that each alternative radiotherapy plan has relative time coding information; and a plan uploading unit, used to upload and store the multiple time-calibrated alternative treatment plans in the target storage area.

[0042] For example, the multi-source data processing unit can receive static 3D images from a CT simulator, dynamic image sequences acquired by a 4D-CT scanner, and physiological motion data such as real-time respiratory waveforms obtained through a respiratory gating device. The multi-source data processing unit can also acquire target dose targets and dose limits for organs at risk such as the spinal cord and lungs set by clinicians. These multi-dimensional data are then spatially registered and fused to form a comprehensive dataset including organ anatomy, physiological motion information, and dose constraints. Subsequently, the planning unit generates multiple alternative radiotherapy plans in parallel based on the comprehensive dataset using an optimization algorithm: for example, Plan A corresponds to the end-expiratory phase and prioritizes left lung protection; Plan B corresponds to the mid-inspiratory phase and focuses on target homogeneity; and Plan C generates a backup plan for possible intestinal displacement, thereby achieving pre-correlation between the plan and the patient's potential physiological state.

[0043] For example, the planning time alignment unit can extract the periodic characteristics of physiological motion data (such as respiratory cycle duration) and combine them with the physical characteristics of the radiotherapy equipment (such as gantry rotation speed, MLC blade maximum movement speed, and dose rate rise time) to calculate the theoretical execution timeline for each alternative plan. For instance, the planning time alignment unit can map each control command (such as MLC position and gantry angle) in different alternative radiotherapy plans to a unified time coordinate system and insert timestamp tags to form a time-coded plan file. The plan upload unit then uploads these plan files to the core storage server of the radiotherapy network or the target storage area of ​​the cloud platform via a standardized protocol (such as DICOM RT), while simultaneously establishing an index relationship between alternative radiotherapy plans and patient IDs and timestamps for rapid retrieval and access by the subsequent real-time scheduling module.

[0044] In some embodiments, the hardware execution sequence corresponding to the targeted radiotherapy plan includes at least one of the following sequences:

[0045] The multi-leaf collimator control point sequence is a set of instructions arranged sequentially on the time axis to control the position and state of each blade of the multi-leaf collimator. The multi-leaf collimator control point sequence is used to form a shooting field shape that changes over time by controlling the dynamic position of the multi-leaf collimator blades.

[0046] The gantry angle control sequence is a set of instructions arranged sequentially on the time axis for controlling the rotation angle of the radiotherapy equipment gantry. The gantry angle control sequence is used to control the gantry to rotate around the target object to the target angle according to a predetermined trajectory and time point.

[0047] Dose rate control sequence, wherein the dose rate control sequence is a set of instructions arranged sequentially on the time axis for controlling the intensity or flux of radiation output by the radiotherapy equipment.

[0048] For example, the multi-leaf collimator control point sequence can be configured as a series of binary data streams with millisecond-level timestamps, each binary containing a leaf group number and target position coordinates. For instance, during volumetric rotational intensity-modulated therapy (IMRT), this sequence sequentially issues position commands for 160 pairs of leaves at 50-millisecond intervals, dynamically shaping a beam field that gradually transitions from a star shape to a rectangle during continuous gantry rotation. This sequence can be directly transmitted to the multi-leaf collimator servo controller via a high-speed real-time bus, driving the leaves to reach the target position at specified time points, achieving synchronization between beam modulation and three-dimensional tumor conformal therapy.

[0049] For example, the gantry angle control sequence can be encoded as an angle-time function table, including the starting angle, ending angle, angular velocity curve, and time stamps for key points. For instance, in segmented arc therapy, the sequence commands control the gantry to move at a non-uniform speed within the 0° to 180° range: maintaining a constant low speed in the critical irradiation segment from 45° to 90°, and accelerating through the non-irradiation segment, with the entire process achieving a control accuracy of 0.1 degrees per second. This sequence is parsed by a motion control card to generate pulse signals, driving servo motors to rotate the gantry precisely along a predetermined spatiotemporal trajectory.

[0050] For example, the dose rate control sequence can employ a hybrid mode of event triggering and timing control, including a base output rate, gradient change rate, and emergency cutoff command. For instance, during simultaneous boost irradiation, the sequence command maintains the dose rate at a high throughput of 600 MU / min during the irradiation phase in the tumor center, and then linearly reduces the dose rate to 200 MU / min within 10 milliseconds when the irradiation field sweeps across the projection area of ​​a sensitive organ. This sequence, through the analog-to-digital conversion interface of the dose servo system, adjusts the accelerator gun high voltage or target current in real time, thereby achieving sub-second intensity modulation response.

[0051] In some embodiments, the adaptive decision unit includes: an image matching subunit, configured to calculate the image similarity between the real-time image data and the image data used in formulating each alternative radiotherapy plan when the real-time data includes real-time image data; and a first decision subunit, configured to select the alternative radiotherapy plan with the highest image similarity as the target radiotherapy plan.

[0052] For example, the image matching subunit can first preprocess the real-time acquired cone-beam CT images (corresponding to the real-time image data mentioned above), such as extracting regions of interest containing the tumor target area and key organs at risk. Subsequently, the image matching subunit can use a deformation registration algorithm to elastically align the corresponding phases of the real-time images with the reference four-dimensional CT images (i.e., the image data used when formulating the alternative radiotherapy plans) corresponding to each alternative radiotherapy plan, and calculate the normalized cross-correlation coefficient between the two in the registered region as the image similarity. The first decision subunit receives the similarity scores of all alternative plans in real time, and can complete the comparison within one respiratory cycle (e.g., 3 seconds), and selects plan A (e.g., the plan corresponding to the end-expiratory phase) with the highest score and a similarity exceeding a preset threshold (e.g., 0.92) as the target radiotherapy plan, triggering the switching.

[0053] For example, to reduce computational latency, the image matching subunit can also employ a feature point clustering matching strategy. For instance, it can extract high-frequency feature point sets such as bone edges and surgical clips from real-time 2D perspective images, and perform rapid spatial pattern matching between these high-frequency feature point sets and pre-annotated similar feature point templates in the reference images of each candidate plan. The reciprocal of the average distance between matched point pairs is used as the similarity metric. After receiving the fusion similarity evaluation based on the latest L-frame (e.g., 3-frame) images (e.g., a time window of approximately 500 milliseconds), the first decision subunit selects Plan B (e.g., the plan corresponding to the deep inhalation phase) with consistently leading similarity and satisfactory stability as the target plan. The decision is then issued before the arrival of the next gantry angle control point, thus facilitating a closed-loop response from sub-second image feedback to plan switching.

[0054] In some embodiments, the adaptive decision-making unit includes: a second decision subunit, configured to determine a target radiotherapy plan based on the following first or second operation when real-time motion data is included in the real-time data: the first operation includes: inputting real-time motion data into a neural network model, predicting organ state information corresponding to the real-time motion data based on prior knowledge learned by the neural network model during end-to-end training; obtaining the organ state information on which each alternative radiotherapy plan is based, and calculating the similarity between the similarity and the organ state information predicted by the neural network model, and selecting the alternative radiotherapy plan with the highest similarity as the target radiotherapy plan; the second operation includes: determining the in vivo deformation field of the target object based on the real-time motion data, determining the current organ state information of the target object based on the in vivo deformation field of the target object; and determining the target radiotherapy plan from multiple alternative radiotherapy plans based on the similarity between the organ state information on which each alternative radiotherapy plan is based and the current organ state information of the target object.

[0055] For example, the first operation of the second decision subunit can be implemented by deploying a temporal convolutional neural network trained end-to-end. This network can take the motion trajectory of real-time surface optical markers or waveform data from respiratory sensors as input and directly output the predicted three-dimensional displacement vectors of internal tumors and key organs (such as prostate and liver tumors). For instance, during decision-making, the system can input motion data within the current time window (e.g., a 200-millisecond window) into the temporal convolutional neural network, then obtain the organ center coordinates and morphological parameters predicted by the neural network, and subsequently perform Euclidean distance and shape similarity calculations with the labeled information of organ states in each phase of 4D-CT, on which each alternative plan is based. For example, when the similarity between the predicted state and the "mid-expiration - tumor displacement 3mm" state in the plan library reaches 95%, the corresponding plan is selected as the target plan, thereby facilitating an indirect and rapid mapping from external motion signals to internal anatomical states.

[0056] For example, when the second decision subunit executes the second operation, it can utilize a deformation field reconstruction algorithm based on a biomechanical model to calculate the full-field tissue deformation field by solving the elasticity equations from real-time monitored surface point cloud motion data (e.g., acquired via a stereo camera) or partial in vivo reference point displacements (e.g., acquired via electromagnetic tracking). Then, based on this deformation field, the reference image is deformed to directly obtain the estimated 3D model and spatial coordinates of each organ at the current moment, serving as the estimated state of each organ. Subsequently, the estimated state is compared with the reference organ states associated with each alternative plan at the voxel level for overlap or surface distance calculation. For instance, when the residual deformation field between the estimated bladder fullness state and the "half-fullness state" reference model in the plan library is minimized, the optimized plan corresponding to that state is selected as the execution target, thereby helping to provide a more intuitive anatomical state matching path based on a physical model.

[0057] In some embodiments, the adaptive decision unit includes: a third decision subunit, configured to: determine a first confidence level for each candidate radiotherapy plan based on real-time image data, given that the real-time data includes real-time image data and real-time motion data; determine a second confidence level for each candidate radiotherapy plan based on real-time motion data; perform a weighted calculation on the first and second confidence levels for each candidate radiotherapy plan to obtain a target confidence score for each candidate radiotherapy plan; and select the candidate radiotherapy plan with the highest target confidence score as the target radiotherapy plan.

[0058] For example, the third decision subunit can improve decision robustness by fusing multi-source heterogeneous data: First, the third decision subunit can determine the grayscale mutual information value between real-time cone-beam CT images and reference images of each alternative plan, and then normalize the grayscale mutual information value to obtain the first confidence level (range 0-1); at the same time, the second confidence level is calculated by using the waveform correlation coefficient between the real-time respiratory signal and the phase template associated with each plan; subsequently, the system dynamically adjusts the weighting coefficients (e.g., image weight 0.7, motion weight 0.3) according to the current treatment stage (e.g., the initial stage trusts the image more, and the subsequent stage pays more attention to motion stability), and calculates a weighted comprehensive score for each plan; finally, the plan C with the highest comprehensive score of more than 0.85 is selected as the target radiotherapy plan. For example, when the image confidence level of plan C is 0.9 and the motion confidence level is 0.8, its target score reaches 0.87, so that it can still make a stable and reliable adaptive selection when the image quality fluctuates or the motion signal is transiently abnormal.

[0059] In some embodiments, the execution scheduling unit includes: a phase matching subunit, used to perform phase matching between real-time data and a four-dimensional image sequence taken before radiotherapy of the target object to determine the target phase corresponding to the real-time data; a positioning subunit, used to locate the plan execution start point corresponding to the target phase on the time axis; a subsequence determination subunit, used to extract subsequent subsequences starting from the plan execution start point from the hardware control sequence associated with the target radiotherapy plan; and a sequence switching subunit, used to control the radiotherapy equipment to interrupt the currently executed subsequence and switch to the extracted subsequent subsequence.

[0060] For example, the phase matching subunit can quickly project and register real-time images with a pre-radiotherapy four-dimensional image sequence. By comparing the projection similarity of each phase of the real-time image and the four-dimensional image sequence on key structures, the respiratory phase corresponding to the current moment (e.g., the 8th frame of the end-expiratory phase) is determined. The positioning subunit then maps the absolute position of this phase on the complete treatment timeline (e.g., 125 seconds after the start of treatment) in the four-dimensional image timestamp corresponding to the phase, and uses this point as the starting point for plan execution. The subsequence determination subunit then retrieves and extracts all control command packets from the 125-second timestamp until the end of the plan from the hardware control sequence associated with the pre-decomposed target radiotherapy plan. Finally, the sequence switching subunit sends a command to the accelerator control system to interrupt the current command stream and load and execute the new subsequence within the next beam pulse interval (typically less than 10 milliseconds) via the real-time control network, achieving seamless treatment path switching based on precise phase locking.

[0061] In another implementation scenario, when real-time data primarily consists of surface optical motion signals, the phase matching subunit can directly determine the target phase (e.g., the mid-inspiratory phase) by performing cross-correlation analysis between the real-time respiratory waveform and the respiratory signals synchronously recorded during four-dimensional image acquisition. The positioning subunit then calculates the corresponding logical starting point on the target plan timeline based on the relative time position of this phase within a standard respiratory cycle (e.g., 45% of the cycle) and the progress of the already executed treatment. The subsequence determination subunit then dynamically assembles a continuous instruction segment from the hardware control sequence associated with the target plan, starting from this logical point and adapting to possible subsequent phase changes. The sequence switching subunit then coordinates the gantry, multi-leaf collimator, and dose rate controller to trigger synchronous switching when the equipment moves to the next safely interruptible mechanical node (e.g., the gantry rotates through a specific angular tolerance zone), ensuring a smooth transition between the mechanical system and dose delivery and avoiding switching jitter or dose errors caused by motion prediction deviations.

[0062] In some embodiments, the phase matching subunit includes: a first matching module, configured to perform phase matching between the real-time image data and a four-dimensional image sequence taken before radiotherapy of the target object, when the real-time data includes real-time image data, to obtain a target phase corresponding to the real-time image data; and a second matching module, configured to determine the in vivo deformation field of the target object based on the real-time motion data, determine the current organ state information of the target object based on the in vivo deformation field, and perform phase matching between the current organ state information of the target object and the organ state information described by each image in the four-dimensional image sequence, to obtain a target phase corresponding to the real-time motion data, when the real-time data does not include real-time image data but includes real-time motion data.

[0063] For example, the first matching module can achieve phase locking through a real-time image-driven direct registration algorithm. For instance, when the system acquires cone-beam CT or two-dimensional orthogonal fluoroscopic images during treatment, the first matching module can use a fast registration algorithm based on grayscale or features to perform frame-by-frame similarity calculation between the real-time image and the 4D-CT sequence acquired before radiotherapy. For example, by minimizing the mutual information difference between the real-time image and the 4D-CT phase reference images projected in key organ regions, the respiratory phase corresponding to the current image (such as the 70th phase of the end-expiratory phase) can be identified, thereby achieving phase matching based on direct comparison of anatomical structures.

[0064] For example, the second matching module can perform phase inference through indirect reconstruction of motion data. For instance, when real-time images are unavailable and only motion data such as surface optical markers or respiratory waveforms are available, the second matching module can use a pre-trained deformation vector field model to calculate the three-dimensional displacement field of tumors and key organs in the body based on real-time motion data. Then, based on this displacement field, the equivalent spatial configuration of the organ is reconstructed, and the equivalent spatial configuration is morphologically compared with the organ models manually or automatically drawn in each phase of 4D-CT (such as calculating the centroid distance or surface distance). Finally, the 4D-CT phase with the highest similarity is selected as the target phase. For example, when the reconstructed liver displacement model matches the organ state of the mid-inspiratory phase of 4D-CT by 90%, it is determined that the current moment is the corresponding moment of the target phase, thus realizing phase tracking in the absence of direct image feedback information.

[0065] For example, the planning management unit can provide access to alternative radiotherapy plans through a RESTful API (Representational StateTransfer Application Programming Interface), the plan execution sequence splitting unit receives alternative radiotherapy plans through a message queue and returns a structured hardware control sequence, the real-time data processing unit subscribes to the imaging equipment data stream as a stream processing service, and the adaptive decision-making unit and the execution scheduling unit respectively implement low-latency decision communication and control command issuance through gRPC (gRPC Remote Procedure Calls) interfaces.

[0066] In some embodiments, the real-time adaptive radiotherapy management system further includes: a data monitoring and feedback unit, used to monitor N monitoring indicators of the target object during the radiotherapy process, and when any monitoring indicator reaches the corresponding preset threshold, to notify the adaptive decision-making unit to perform a re-decision, wherein the re-decision is used to determine the target radiotherapy plan that matches the current state of the target object from multiple alternative radiotherapy plans based on the latest real-time data of the target object, where N is an integer greater than or equal to 1.

[0067] For example, the data monitoring and feedback unit can continuously receive and analyze key indicators in the real-time data stream, including but not limited to respiratory cycle deviation (such as exceeding the preset baseline ±20%), tumor target center displacement (such as exceeding the planned boundary by 3mm), and target cumulative dose deviation based on real-time dose reconstruction of EPID (such as being less than the prescribed dose by 5%). When any indicator exceeds its clinical safety threshold, the unit immediately sends a decision reassessment request to the adaptive decision unit through event-driven messages, triggering the decision unit to recalculate and select the optimal alternative plan based on the latest imaging and motion data. For example, when a patient is detected to have a sudden cough that causes respiratory disturbance, the system can switch to a stable radiotherapy plan that is pre-set for cases of large respiratory fluctuations in a timely manner, thereby forming a closed-loop quality assurance and adaptive intervention.

[0068] In some embodiments, the N monitoring indicators include at least one of respiratory cycle, tumor offset, tumor shrinkage rate, and organ-at-risk damage rate.

[0069] For example, the N monitoring indicators tracked in real time by the data monitoring and feedback unit can be manifested as follows: quantifying the stability of the respiratory cycle through respiratory sensors (e.g., alarm is triggered if the cycle standard deviation is >0.5 seconds), calculating the centroid offset of the tumor target area based on real-time image registration (e.g., triggering the threshold if the displacement is >4mm for 3 consecutive frames), extrapolating the tumor shrinkage rate by comparing the cumulative dose with the planned dose (e.g., indicating that the plan needs to be adjusted if the tumor shrinkage is >15% in the middle of the treatment course), and using the dose-volume histogram to assess the risk of exceeding the radiation dose limit for organs at risk (e.g., the warning is triggered if the maximum radiation dose reaches 90% of the tolerable dose). If any indicator exceeds the corresponding threshold, the system's adaptive re-decision loop will be triggered, thereby realizing dynamic risk control of the treatment process and optimization of individualized radiotherapy plans.

[0070] In some embodiments, when determining a target radiotherapy plan, the adaptive decision unit is further configured to prioritize alternative radiotherapy plans based on at least one of the following optimization objectives:

[0071] The primary optimization objective is to constrain and shorten the duration of radiotherapy.

[0072] The second optimization objective is to constrain the dose distribution in the tumor target area and organs at risk to meet the preset dose requirements;

[0073] The third optimization objective serves as a constraint to reduce the radiation dose to normal tissues.

[0074] For example, the adaptive decision unit has a built-in multi-objective optimization engine. When multiple alternative plans meet the basic matching conditions, it can prioritize them according to a preset clinical strategy. For example, when the first optimization objective (shortest treatment duration) is the dominant condition, the system will select the plan with the fewest total jumps and the shortest gantry rotation path for priority execution. If the second optimization objective (dose compliance) is the priority condition, the system will score and rank the plans by calculating the degree of compliance of each plan with the target area D95 achievement rate and the Dmax limit of the organs at risk. In scenarios where patients are extremely sensitive to the protection of normal tissues, the third optimization objective (minimizing the dose to normal tissues) can be enabled as the dominant condition. The system will compare the relevant indicators of critical organs in each plan (such as lung V20 or rectal V50) and select the alternative plan with the lowest comprehensive irradiation volume as the final target radiotherapy plan, thereby realizing the flexible integration of clinical preferences in real-time decision-making.

[0075] For example, Figure 2 This is a flowchart of an optional real-time adaptive radiotherapy management method according to an embodiment of this application, such as... Figure 2 As shown, it includes the following steps:

[0076] Step S1: Generate and upload multiple alternative radiotherapy plans.

[0077] For example, before radiotherapy begins, the real-time adaptive radiotherapy management system can generate multiple alternative radiotherapy plans for the same patient based on different possible anatomical states, and upload the multiple alternative radiotherapy plans to the target storage area for centralized management.

[0078] For example, generating and uploading multiple alternative treatment plans may include the following steps:

[0079] Step S11: Collect multi-dimensional data and clinical input information.

[0080] First, the system can acquire simulated positioning CT images and 4D CT images of each phase of the patient, and simultaneously collect physiological data such as the patient's respiratory waveform and body surface movement trajectory. At the same time, clinicians input basic requirements and constraints such as the treatment site, prescription dose target, fractionation plan, and type of radiotherapy technique (e.g., IMRT, VMAT).

[0081] Step S12: Set treatment goals and plan parameters.

[0082] Based on the system input information from step S11 above, the dose coverage requirements for the tumor target area, the dose limits for organs at risk, and the tolerated dose for normal tissues are determined. On this basis, key physical parameters of the treatment plan are preliminarily determined, such as the direction of the radiation field and the range of gantry rotation, and time constraints for the treatment process are set.

[0083] Step S13: Generate static / dynamic alternative radiotherapy plans.

[0084] The system first generates an initial treatment plan based on reference CT images and the selected radiotherapy technique (such as IMRT or VMAT) using a dosimetric optimization algorithm. Subsequently, by combining 4D CT image sequences with synchronously recorded physiological motion data, a series of dynamic radiotherapy plans (i.e., 4D plans) corresponding to different respiratory phases or motion states are generated using a 4D optimization algorithm, with each plan associated with a specific anatomical state.

[0085] Step S14: Timeline alignment and system integration.

[0086] Furthermore, based on the collected motion pattern data and the accelerator's physical execution characteristics (such as dose rate ramp-up time and gantry movement speed), the system calibrates and aligns all control commands in the 4D plans on a unified time coordinate system, ensuring that each plan has accurate relative time encoding information. Finally, these time-tagged 4D plans are uploaded to the treatment management system's storage unit, completing the construction of a multi-alternative plan library and providing a foundation for subsequent real-time adaptive scheduling.

[0087] Step S2: Split the alternative radiotherapy plans into hardware execution sequences.

[0088] The real-time adaptive radiotherapy management system can perform parsing and transformation operations on each uploaded candidate radiotherapy plan, breaking it down into a set of discrete control sequences that can be directly executed by the linear accelerator hardware. Each sequence contains timing-based instruction parameters that can directly drive a specific control module. An optional implementation flow is as follows:

[0089] Step S21: The real-time adaptive radiotherapy management system can analyze alternative radiotherapy plans and extract core control parameters, including beam direction, multi-leaf collimator (MLC) leaf position sequence, gantry angle, dose rate, etc.

[0090] Step S22: The real-time adaptive radiotherapy management system can convert the control parameters extracted in step S21 into time-series control instructions that strictly correspond to hardware actions, based on the alternative radiotherapy plans that have been aligned with the time axis in step S14. Examples include: generating a sequence of control points showing the MLC blade position changing over time; generating trajectory instructions showing the gantry rotation angle changing over time; generating a modulation curve of the dose rate triggered by time or events; and marking each control action with a precise execution timestamp or trigger condition.

[0091] Step S23: The real-time adaptive radiotherapy management system can decompose the complete irradiation path in the alternative radiotherapy plan into a series of time-sequential "beam on-modulation-off" basic execution operations based on the combination of gantry angle, MLC shape and dose output changes, thereby obtaining time-sequential control commands.

[0092] Step S24: The real-time adaptive radiotherapy management system can encapsulate and distribute the generated time-sequential control commands according to the communication protocols of each control subsystem of the linear accelerator (such as the MLC controller, gantry servo system, and dose servo system), forming instruction data packets that each subsystem can independently receive and execute, thus preparing for subsequent real-time scheduling and rapid switching.

[0093] Step S3: Acquire real-time data during radiotherapy.

[0094] During radiotherapy irradiation using a linear accelerator, the real-time adaptive radiotherapy management system can simultaneously initiate real-time data acquisition and processing, continuously acquiring dynamic information reflecting the patient's current anatomical structure and physiological state, providing immediate input for adaptive decision-making. This includes, for example, the following steps:

[0095] Step S31: The real-time adaptive radiotherapy management system can periodically or continuously acquire two-dimensional / three-dimensional images of the treatment area through an image-guided system integrated into the radiotherapy equipment, such as cone-beam computed tomography (CBCT), electronic field imaging device (EPID), or magnetic resonance imaging (MR-linac). Simultaneously, it utilizes optical surface monitoring systems, respiratory gating devices, etc., to synchronously acquire physiological motion signals such as the patient's surface movement trajectory and respiratory waveforms.

[0096] Step S32: The real-time adaptive radiotherapy management system can preprocess the acquired raw data (such as noise reduction and correction). For example, it can quickly reconstruct or generate high-resolution "pseudo-CT" 3D images from limited 2D projection data or sparse 3D sampling data to more clearly reflect soft tissue contrast and organ deformation, thereby realizing the real-time conversion from limited data to synthetic images that can be used for dose calculation and evaluation.

[0097] Step S4: Adaptive decision-making and plan / sequence selection.

[0098] The adaptive decision-making unit can dynamically assess the patient's current status based on the real-time data obtained in step S3, and select the most suitable target radiotherapy plan from multiple pre-stored alternative radiotherapy plans or directly select the corresponding hardware execution subsequence for real-time adjustments to the radiotherapy. The logic of the adaptive decision-making is as follows:

[0099] Step S41: The real-time adaptive radiotherapy management system continuously monitors key physiological and anatomical indicators, such as respiratory cycle stability and the offset of the tumor target volume relative to the planned position. When any indicator exceeds a preset safety or effectiveness threshold, the plan switching decision process is automatically triggered.

[0100] Step S42: The core of the decision-making process involves rapidly registering and phase-matching the real-time acquired 3D images (or generated pseudo-3D images) with the 4D-CT image sequence acquired before treatment. This matching is used to determine which respiratory phase (e.g., the i-th phase) in the 4D-CT corresponds to the patient's current anatomical state, and thereby associate it with the logical position of that phase on the complete treatment timeline.

[0101] Step S43: Based on the matched phase, the adaptive decision unit extracts and schedules the execution of subsequent sub-sequences starting from the time point of that phase from the hardware execution sequence library that is already bound to each alternative plan, so as to achieve “seamless continuation” of treatment.

[0102] Optional execution examples are as follows:

[0103] At the first decision moment: Based on real-time images, the adaptive decision unit selects Plan A as the target radiotherapy plan. After matching with the 4D-CT sequence, it is determined that the current phase is i. The system then locates the instruction corresponding to the time start point of the i-th phase in the hardware control sequence of Plan A, and starts executing the subsequent sub-sequences from this point.

[0104] At the second decision point (after the change in status): Based on the new real-time images, the decision-making system reassesses and selects Plan B as the target radiotherapy plan. At this time, the phase matching results show that the patient's status has changed to phase i+2. The system then immediately locates and extracts the subsequent subsequence starting from the time start of phase i+2 from the hardware control sequence of Plan B, and controls the accelerator to switch to execute this new subsequence.

[0105] Through this mechanism, the real-time adaptive radiotherapy management system can achieve adaptive execution flow switching across plans and phases without interruption or with minimal interruption of treatment.

[0106] Step S5: Perform sequence scheduling and fast switching.

[0107] The execution scheduling unit receives sequence switching instructions from the adaptive decision-making unit and is responsible for rapidly deploying the selected target hardware execution sequence to the linear accelerator control system, achieving real-time updates of treatment parameters and seamless switching of the execution flow. Optional operation steps are as follows:

[0108] Step S51: The execution scheduling unit does not simply transmit instructions, but integrates a layer of security and logic verification. The execution scheduling unit can compare the real-time data stream (such as the current MLC position and rack angle) with the initial state of the sequence to be switched, ensuring that the switching action is within a safe window in terms of equipment mechanics and dosimetry (such as the beam intermittent period and the rack low-speed zone), and then generate an executable instruction package containing the switching time point, target sequence identifier, and necessary transition parameters.

[0109] Step S52: When multiple alternative radiotherapy plans meet the switching conditions, the scheduling service can select the target radiotherapy plan according to a preset priority strategy. This strategy may be based on single or combined optimization objectives such as: shortest remaining treatment time (to improve efficiency), optimal expected dose distribution (to improve efficacy), or minimizing the radiation dose to normal tissues (to reduce toxic side effects), ensuring that the switching decision meets real-time requirements while also conforming to the overall clinical treatment goals.

[0110] Step S6: Closed-loop monitoring and continuous adaptive decision-making.

[0111] Throughout the treatment process, the real-time adaptive radiotherapy management system operates continuously in a closed-loop manner, cyclically executing steps S3 to S5 to achieve dynamic and continuous adaptive management of the entire treatment process. For example, as a key component of the closed loop, the real-time adaptive radiotherapy management system can utilize radiation field images acquired during radiotherapy (such as through an electronic field imaging device, EPID) and, through a dose reconstruction algorithm, calculate the cumulative dose distribution already projected onto the patient's body in near real-time. By comparing this reconstructed dose with the planned expected dose, the accuracy of treatment execution can be monitored, and it can be determined whether a new round of adaptive adjustment (i.e., returning to step S4) is necessary due to dose deviation.

[0112] In some embodiments, this application also provides a radiotherapy device, which may include the above-described real-time adaptive radiotherapy management system, as well as an accelerator treatment head, a multi-leaf collimator, a gantry, and an image acquisition device controlled by the real-time adaptive radiotherapy management system.

[0113] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0114] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0115] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0116] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0117] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0118] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0119] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A real-time adaptive radiotherapy management system, characterized in that, include: A real-time data processing unit is used to acquire real-time data of the target object during radiotherapy, wherein the real-time data includes real-time image data and / or real-time motion data; An adaptive decision-making unit is configured to determine a target radiotherapy plan that matches the current state of the target object from a plurality of alternative radiotherapy plans based on the real-time data, wherein the plurality of alternative radiotherapy plans are associated with different physiological motion phases or anatomical states of the target object; The plan execution sequence splitting unit is used to parse and split each alternative radiotherapy plan into one or more hardware execution sequences, wherein the hardware execution sequence includes control parameters that are directly executed by the control system of the radiotherapy equipment; An execution scheduling unit is used to send at least one hardware execution sequence corresponding to the target radiotherapy plan to the control system of the radiotherapy equipment; The execution scheduling unit includes: a phase matching subunit, used to perform phase matching between the real-time data and the four-dimensional image sequence taken before radiotherapy of the target object to determine the target phase corresponding to the real-time data; a positioning subunit, used to locate the plan execution start point corresponding to the target phase on the time axis; a sub-sequence determination subunit, used to extract subsequent sub-sequences starting from the plan execution start point from the hardware control sequence associated with the target radiotherapy plan; and a sequence switching subunit, used to control the radiotherapy equipment to interrupt the currently executed sub-sequence and switch to execute the extracted subsequent sub-sequence.

2. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, The real-time adaptive radiotherapy management system also includes: The multi-source data processing unit is used to acquire multi-dimensional data, including static images, dynamic images, and physiological changes and physiological movements of the target object, and combine them with clinical goals and constraints to form a comprehensive dataset. The plan generation unit is used to generate multiple alternative radiotherapy plans for the target object based on the comprehensive dataset; The planning time alignment unit is used to map and calibrate the multiple alternative radiotherapy plans to a unified time coordinate system based on the physiological motion data and the physical execution characteristics of the radiotherapy equipment, so that each alternative radiotherapy plan has relative time coding information. The plan upload unit is used to upload and store multiple time-calibrated alternative treatment plans in the target storage area.

3. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, The hardware execution sequence corresponding to the target radiotherapy plan includes at least one of the following sequences: A multi-leaf collimator control point sequence, wherein the multi-leaf collimator control point sequence is a set of instructions arranged sequentially on the time axis for controlling the position and state of each blade of the multi-leaf collimator, and the multi-leaf collimator control point sequence is used to form a shooting field shape that changes over time by controlling the dynamic position of the multi-leaf collimator blades. A gantry angle control sequence, wherein the gantry angle control sequence is a set of instructions arranged sequentially on a time axis for controlling the rotation angle of the radiotherapy equipment gantry, and the gantry angle control sequence is used to control the gantry to rotate around the target object to a target angle according to a predetermined trajectory and time point; A dose rate control sequence, wherein the dose rate control sequence is a set of instructions arranged sequentially on a time axis for controlling the output radiation intensity or flux of a radiotherapy device, and the dose rate control sequence is used to dynamically control the output intensity of a radiation source based on a predetermined time function or event triggering.

4. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, The adaptive decision-making unit includes: An image matching subunit is used to calculate the image similarity between the real-time image data and the image data used when formulating each alternative radiotherapy plan, when the real-time data includes the real-time image data. The first decision subunit is used to select the candidate radiotherapy plan with the highest image similarity as the target radiotherapy plan.

5. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, The adaptive decision-making unit includes: The second decision subunit is configured to determine the target radiotherapy plan based on either the first or second operation, provided that the real-time data includes the real-time motion data: The first operation includes: inputting the real-time motion data into a neural network model, predicting organ state information corresponding to the real-time motion data based on the prior knowledge learned by the neural network model during end-to-end training; obtaining the organ state information on which each alternative radiotherapy plan is based, and calculating the similarity between the organ state information predicted by the neural network model and the alternative radiotherapy plan with the highest similarity as the target radiotherapy plan; The second operation includes: determining the in vivo deformation field of the target object based on the real-time motion data; determining the current organ status information of the target object based on the in vivo deformation field of the target object; and determining the target radiotherapy plan from the plurality of alternative radiotherapy plans based on the similarity between the organ status information used to formulate each alternative radiotherapy plan and the current organ status information of the target object.

6. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, The adaptive decision-making unit includes: The third decision subunit is configured to, when the real-time data includes the real-time image data and the real-time motion data, determine a first confidence level for each candidate radiotherapy plan based on the real-time image data; determine a second confidence level for each candidate radiotherapy plan based on the real-time motion data; perform a weighted calculation on the first and second confidence levels of each candidate radiotherapy plan to obtain a target confidence score for each candidate radiotherapy plan; and select the candidate radiotherapy plan with the highest target confidence score as the target radiotherapy plan.

7. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, The phase matching subunit includes: The first matching module is used to perform phase matching between the real-time image data and the four-dimensional image sequence taken before radiotherapy of the target object, when the real-time data includes the real-time image data, to obtain the target phase corresponding to the real-time image data. The second matching module is used to determine the internal deformation field of the target object based on the real-time motion data when the real-time data does not include the real-time image data but includes the real-time motion data, determine the current organ state information of the target object based on the internal deformation field, and perform phase matching between the current organ state information of the target object and the organ state information described by each image in the four-dimensional image sequence to obtain the target phase corresponding to the real-time motion data.

8. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, The real-time adaptive radiotherapy management system also includes: The data monitoring and feedback unit is used to monitor N monitoring indicators of the target object during the radiotherapy process, and when any monitoring indicator reaches the corresponding preset threshold, it notifies the adaptive decision-making unit to perform a re-decision. The re-decision is used to determine the target radiotherapy plan that matches the current state of the target object from the multiple alternative radiotherapy plans based on the latest real-time data of the target object. N is an integer greater than or equal to 1.

9. The real-time adaptive radiotherapy management system according to claim 8, characterized in that, The N monitoring indicators include at least one of the following: respiratory cycle, tumor offset, tumor shrinkage rate, and organ-at-risk damage rate.

10. The real-time adaptive radiotherapy management system according to claim 1, characterized in that, When determining the target radiotherapy plan, the adaptive decision-making unit is further configured to prioritize candidate radiotherapy plans based on at least one of the following optimization objectives: The primary optimization objective is to constrain and shorten the duration of radiotherapy. The second optimization objective is to constrain the dose distribution in the tumor target area and organs at risk to meet the preset dose requirements; The third optimization objective serves as a constraint to reduce the radiation dose to normal tissues.

11. A radiotherapy device, characterized in that, The system includes a real-time adaptive radiotherapy management system as described in any one of claims 1 to 10, and an accelerator treatment head, a multi-leaf collimator, a gantry, and an image acquisition device controlled by the real-time adaptive radiotherapy management system.

Images