Artificial intelligence-based intelligent decision-making method and electronic device

Abstract

Provided in the present disclosure are an artificial intelligence-based intelligent decision-making method and an electronic device, relating to the technical field of medical treatment. An artificial intelligence-based intelligent decision-making method, comprising: acquiring reference information of a target object, wherein the reference information comprises a pre-plan and a pre-plan image, the pre-plan comprises at least one fraction plan, each fraction plan comprises a plurality of sub-fields, and the pre-plan comprises a dose for each fraction and a dose for each sub-field; acquiring current information of the target object, wherein the current information comprises at least one of a current image and a current dose; and an intelligent decision-making model generating a decision instruction according to the reference information and the current information of the target object, wherein the decision instruction comprises executing the pre-plan, executing an adjusted plan, or issuing an alarm. When target region deformation, positional deviation, or dose deviation occurs during radiotherapy, the intelligent decision-making model is utilized to achieve efficient and accurate real-time decision-making, thereby reducing or even eliminating treatment deviation problems caused by tumor motion, deformation, etc.

Description

An AI-based intelligent decision-making method and electronic device Technical Field

[0001] This disclosure relates to the field of medical technology, and in particular to an intelligent decision-making method and electronic device based on artificial intelligence. Background Technology

[0002] Radiation therapy is widely used in cancer treatment. Typically, a radiation therapy plan (also known as a treatment plan) is developed for a patient before treatment begins. For example, if multiple low-dose irradiations are used, the treatment plan includes several fractions, meaning that during a multi-day treatment period, a portion of the dose can be delivered to the patient each time according to the treatment plan. However, the anatomical structure of the tumor and other tissues (such as surrounding tissues) may change due to growth, deformation, contraction, patient respiration, patient movement, etc., therefore the treatment plan may need to be updated.

[0003] Adaptive radiotherapy (ART) is a technique that dynamically adjusts treatment plans by flexibly adjusting the distribution of radiation doses through real-time monitoring of the patient's tumor morphology and location. This technique aims to improve treatment precision while minimizing damage to surrounding healthy tissues.

[0004] However, existing adaptive radiotherapy primarily relies on feedback to clinicians when a target displacement or dose deviation is detected, allowing clinicians to make decisions. This approach depends heavily on human decision-making, is limited by the clinician's experience, and often involves a high degree of subjectivity, especially when dealing with individualized patient characteristics. Summary of the Invention

[0005] This disclosure provides an intelligent decision-making method and electronic device based on artificial intelligence. In radiotherapy, when target area deformation, positional shift, or dose deviation occurs, the intelligent decision-making model is used to achieve efficient and accurate real-time decision-making, thereby reducing or even eliminating treatment deviation problems caused by tumor movement, deformation, etc. during radiotherapy.

[0006] On one hand, this application discloses an intelligent decision-making method based on artificial intelligence, including: acquiring reference information of a target object, the reference information including a pre-plan and a pre-plan image, wherein the pre-plan includes at least one fractional plan, each fractional plan includes multiple subfields, and the pre-plan includes the dose for each fraction and the dose for each subfield; the pre-plan image is related to the pre-plan; acquiring current information of the target object, the current information including at least one of current image and current dose; the intelligent decision-making model generates a decision instruction based on the reference information and current information of the target object; wherein the decision instruction includes executing the pre-plan, executing the adjusted plan, or issuing an alarm.

[0007] In some embodiments, the intelligent decision model generates decision instructions based on reference information and current information of the target object, including: obtaining volume deviation based on a pre-planned image in the reference information of the target object and a current image in the current information; and / or obtaining dose deviation based on a dose in the reference information of the target object and a current dose; the intelligent decision model generates decision instructions based on volume deviation and / or dose deviation.

[0008] In some embodiments, the intelligent decision-making model generates decision instructions based on volume deviation and / or dose deviation, including: executing a pre-plan when the volume deviation or dose deviation is less than or equal to a first threshold; executing an adjusted plan when the volume deviation or dose deviation is greater than the first threshold but less than or equal to a second threshold; issuing an alarm when the volume deviation or dose deviation is greater than the second threshold; or, performing a weighted calculation on the volume deviation and dose deviation to obtain a calculation result; executing a pre-plan when the calculation result is less than or equal to the first threshold; executing an adjusted plan when the calculation result is greater than the first threshold but less than or equal to the second threshold; and issuing an alarm when the calculation result is greater than the second threshold.

[0009] In some embodiments, the current information also includes current clinical information. The intelligent decision model generates decision instructions based on the reference information and current information of the target object. Specifically, the intelligent decision model generates decision instructions based on the volume deviation, the dose deviation and the current clinical information of the target object.

[0010] In some embodiments, the pre-planning includes at least two sub-planning sessions.

[0011] In some embodiments, obtaining the pre-plan of the target object includes: obtaining the fractionation plan of the target object for the current fraction; obtaining the current information of the target object includes: obtaining at least one of the real-time treatment image of the target object for the current fraction and obtaining the real-time cumulative dose of the target object for the current fraction.

[0012] In some embodiments, the pre-planned image includes the three-dimensional pre-planned volume of the target object; obtaining the real-time treatment image of the target object in the current session includes: obtaining the real-time treatment image of the target object in the current session, and generating the current three-dimensional volume based on the real-time treatment image of the current session; obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: obtaining the volume deviation based on the three-dimensional pre-planned volume and the current three-dimensional volume of the target object.

[0013] In some embodiments, obtaining the real-time cumulative dose of the target object in the current fraction includes: obtaining the actual dose of the target object to each subfield in the current fraction; obtaining the cumulative dose of the current fraction based on the actual dose of each subfield of the target object; obtaining the dose deviation based on the dose in the target object reference information and the current dose includes: obtaining the dose deviation based on the total dose of the target object to the current subfield in the pre-planned dose and the cumulative dose of the current fraction.

[0014] In some embodiments, obtaining a pre-plan for the target object includes: obtaining a fractionation plan for the current fraction of the target object; obtaining current information about the target object includes: obtaining an image guidance image of the target object before the current fractionation treatment; obtaining volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: obtaining volume deviation based on the pre-planned image in the reference information of the target object and the image guidance image of the target object; executing the pre-plan includes executing the fractionation plan for the current fractionation.

[0015] In some embodiments, the pre-planned image includes the three-dimensional pre-planned volume of the target object; acquiring an image guide image of the target object before the current treatment includes: acquiring the image guide image of the target object before the current treatment; generating the current three-dimensional volume based on the image guide image of the target object; obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: obtaining the volume deviation based on the three-dimensional pre-planned volume and the current three-dimensional volume of the target object.

[0016] In some embodiments, if volume deviation is acceptable, the method further includes: generating a current fractionation plan by dose prediction or dose recalculation; and the intelligent decision model executing the current fractionation plan based on whether the fractionation plan dose meets the prescription requirements.

[0017] In some embodiments, the intelligent decision-making model further includes, before executing the adjusted plan, generating the adjusted plan; the intelligent decision-making model acquiring the adjusted plan; wherein at least a portion of generating the adjusted plan is performed in parallel with generating decision instructions.

[0018] In some embodiments, the adjusted plan can be generated during the stop-beam period or during the exit-beam period.

[0019] In some embodiments, the intelligent decision model generates a termination instruction based on dose deviation and / or volume deviation.

[0020] In some embodiments, generating an adjusted plan includes generating at least two adjusted plans; the method further includes: evaluating and determining an adjusted plan from the at least two adjusted plans.

[0021] In some embodiments, the method further includes displaying the decision-making process or the decision result.

[0022] In some embodiments, the intelligent decision-making model is obtained based on the following steps: acquiring multiple sample data to establish a sample database, the sample data including multimodal information of sample objects, wherein the multimodal information includes images, treatment plans, prescriptions, and decision information of clinicians, and the decision information of clinicians includes at least one set of decision information; the intelligent decision-making model learns from the sample data.

[0023] In some embodiments, the multimodal information also includes clinical information.

[0024] In some embodiments, it may be an image format, a text format, or a video format.

[0025] In some embodiments, the method also includes storing process information and outcome information of the decision-making process.

[0026] In some embodiments, the intelligent decision-making model includes multiple intelligent decision-making models for different diseases; the intelligent decision-making model generates decision instructions based on the reference information and current information of the target object, specifically including: the intelligent decision-making model corresponding to the current disease of the target object generates decision instructions based on the reference information and current information of the target object.

[0027] In some embodiments, the intelligent decision-making model includes multiple intelligent decision-making models for different diseases; before the intelligent decision-making model learns from the sample data, it also includes classifying the sample data according to the disease.

[0028] In some embodiments, when there are new diseases different from those in the sample library, the method further includes: obtaining a disease intelligent decision-making model similar to the new disease; migrating and updating its parameters to generate an intelligent decision-making model for the new disease.

[0029] In some embodiments, the method further includes: periodically or periodically acquiring sample data; and incrementally training the intelligent decision model on the updated sample data.

[0030] In some embodiments, the method further includes: preprocessing multiple sample data to achieve consistency and comparability of different sample data.

[0031] In some embodiments, the multimodal information of the same sample data is encoded separately using a neural network to realize feature interaction between information of different modalities; and the relationship between feature interaction is captured and decision is made.

[0032] In some embodiments, the intelligent decision-making model is obtained through the following steps: training a first pre-trained model based on multiple sample data; performing refined training on the first pre-trained model, the refined training including at least one of model compression and hardware acceleration processing on the pre-trained model; optimizing the first pre-trained model using a reinforcement learning algorithm to obtain a second pre-trained model; the intelligent decision-making model is the second pre-trained model.

[0033] In some embodiments, the method further includes: performing online reinforcement learning on the second pre-trained model to generate a third pre-trained model; the intelligent decision-making model is the third pre-trained model.

[0034] On the other hand, embodiments of the present invention provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the intelligent decision-making method provided in this disclosure.

[0035] This invention provides a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the electronic device to execute the intelligent decision-making method provided in this disclosure.

[0036] This invention provides a computer program product, including a computer program that, when executed by a processor, implements the intelligent decision-making method provided in this disclosure.

[0037] The technical solution provided in this disclosure enables the intelligent decision-making model to learn from doctors' decision-making experience through massive sample training and to autonomously generate decisions to execute pre-plans, execute adjusted plans, or issue alarms. In the treatment of the target object, the model makes decisions based on the current information of the target object and the image deviation and / or volume deviation of the reference information, thereby reducing or even eliminating treatment deviation problems caused by tumor movement, deformation, etc. during radiotherapy.

[0038] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0039] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0040] Figure 1 is a schematic diagram of the implementation environment of radiotherapy according to an embodiment of this disclosure;

[0041] Figure 2 is a flowchart illustrating an intelligent decision-making method according to an embodiment of this disclosure;

[0042] Figure 3 is a flowchart illustrating an intelligent decision-making method according to an embodiment of this disclosure;

[0043] Figure 4 is a flowchart illustrating an intelligent decision-making method according to an embodiment of this disclosure;

[0044] Figure 5 is a flowchart illustrating an intelligent decision-making method according to an embodiment of this disclosure;

[0045] Figure 6 is a flowchart illustrating an intelligent decision-making method according to an embodiment of this disclosure;

[0046] Figure 7 is a flowchart illustrating an intelligent decision-making method according to an embodiment of this disclosure;

[0047] Figure 8 is a flowchart illustrating an intelligent decision-making method according to an embodiment of this disclosure;

[0048] Figure 9 is a schematic diagram illustrating the generation of an intelligent decision-making model according to an embodiment of this disclosure;

[0049] Figure 10 is a schematic diagram illustrating the generation of an intelligent decision-making model according to an embodiment of this disclosure;

[0050] Figure 11 is a schematic diagram illustrating the generation of an intelligent decision-making model according to an embodiment of this disclosure;

[0051] Figure 12 is a block diagram of an electronic device used to implement the intelligent decision generation method of the embodiments of the present disclosure. Detailed Implementation

[0052] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0053] The terminology used in this specification is for the purpose of describing particular exemplary embodiments only and is not restrictive. The singular forms “a,” “an,” and “the” used herein may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms “comprising” and “including” as used herein indicate only the presence of the stated features, integers, steps, operations, components, and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, components, parts, and / or combinations thereof.

[0054] It should be understood that the terms “system,” “module,” and / or “block” used herein are methods for distinguishing different components, elements, parts, sections, or assemblies at different levels in ascending order. However, these terms may be replaced with other expressions if the same purpose can be achieved.

[0055] Generally, as used herein, the terms "module" or "block" refer to logic, or a collection of software instructions, now in hardware or firmware. The modules or blocks described herein can be implemented as software and / or hardware and can be stored in any type of non-transitory computer-readable medium or other storage device. In some embodiments, software modules / units / blocks can be compiled and linked into an executable program. It should be understood that software modules can be invoked from other modules / units / blocks or from themselves, and / or can be invoked in response to detected events or interrupts. Software modules / units / blocks configured to execute on a computing device (e.g., processor 210 as shown in Figure 2) can be provided on computer-readable media, such as optical discs, digital video discs, flash drives, magnetic disks, or any other tangible media, or as digital downloads (and may be initially stored in a compressed or installable format, requiring installation, decompression, or decryption before execution). The software code herein can be stored, in part or in whole, in the storage device of the computing device performing the operation and applied in the operation of the computing device. Software instructions can be embedded in firmware, such as electronically programmable read-only memory (EPROM). It should also be understood that hardware modules / units / blocks may be included in connected logical components, such as gates and flip-flops, and / or may include programmable units, such as programmable gate arrays or processors. The modules / units / blocks or computing device functions described herein may be implemented as software modules / units / blocks, but can be represented in hardware or firmware. Typically, the modules / units / blocks described herein refer to logical modules / units / blocks that can be combined with other modules / units / blocks or divided into submodules / subunits / subblocks, although they are physical organization or storage devices. This description may apply to a system, an engine, or a part thereof.

[0056] It is understood that, unless the context explicitly states otherwise, when a module or block is referred to as being "connected to" or "coupled to" another module or block, it may be directly connected to or coupled to another module or block, or communicate with another module or block, or there may be intermediate units, engines, modules, or blocks present. In this specification, the term "and / or" may include any one or more of the relevant listed items or a combination thereof.

[0057] These and other features, characteristics, functions and operating methods of related structural elements, as well as the assembly and manufacturing economy of components, will become more apparent from the following description of the accompanying drawings, which form part of this specification. However, it should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of this specification. It should also be understood that the drawings are not drawn to scale.

[0058] First, the application scenarios involved in the embodiments of this disclosure are described. The intelligent decision-making method based on artificial intelligence provided in the embodiments of this disclosure can be applied to the field of medical technology, specifically to the scenario of adaptive radiotherapy.

[0059] Figure 1 is a schematic diagram of an implementation environment shown in an embodiment of the present disclosure. The implementation environment includes an image acquisition device 101, a treatment planning device 102, a control computer device 103, and a radiotherapy device 104.

[0060] The image acquisition device 101 is used to acquire images of the tumor site (i.e., the target area) and surrounding normal tissue of the subject to radiotherapy (such as a patient). In some embodiments, the image acquisition device 101 can be at least one of the following: a computed tomography (CT) device, an emission computed tomography (ECT) device, a magnetic resonance imaging (MRI) device, a positron emission tomography (PET) device, and an ultrasound examination device.

[0061] The radiotherapy device 104 is a device for performing radiotherapy on a subject (such as a patient). In some embodiments, the radiotherapy device 104 may include a gantry 1041, a treatment head 1042, an image guide device 1043, and a support device 1044.

[0062] The gantry 1041 can be a rotatable gantry, a C-arm gantry, a drum-shaped gantry, or a robotic arm, etc., used to drive the rotation of the treatment head. This invention does not limit the specific structure of the gantry. The treatment head 1042 can be mounted on the gantry and is used to emit a radiation beam to irradiate the target, such as gamma rays, MV-level X-rays, proton rays, etc. For example, the treatment head 1042 can be a gamma knife treatment head for rotationally focused radiotherapy, an accelerator treatment head for intensity-modulated radiotherapy, or other radiotherapy heads. In this embodiment, the treatment head 1042 can correspond to different radiotherapy modes, such as rotationally focused radiotherapy and intensity-modulated radiotherapy. Image guidance device 1043 is used for positioning the subject to radiotherapy and for real-time image guidance during radiotherapy. The image guidance device can also be a computed tomography (CT) device, an emission computed tomography (ECT) device, a magnetic resonance imaging (MRI) device, a positron emission tomography (PET) device, etc. This embodiment of the invention does not specifically limit the image guidance device; only the X-ray tube and detector shown in the accompanying drawings are used as examples. For example, as shown in FIG1, image guidance device 1043 includes an X-ray tube 1043A and a detector 1043B. Detector 1043B can receive the imaging beam emitted by X-ray tube 1043A passing through the subject to radiotherapy and generate a geopotential image. This geopotential image is used to indicate the morphology and / or location of the target area of ​​the subject to radiotherapy. The image guidance device shown in FIG1 includes an X-ray tube and a detector arranged opposite each other. In some embodiments, the image guidance device may also include two or more X-ray tubes and detectors. The support device 1044 is used to support and move the patient to be radiotreated, and may be a treatment bed.

[0063] In some embodiments, the image acquisition device can be integrated into the radiotherapy equipment, and the image guidance device and the image acquisition device can be the same. The image used to generate the treatment plan can be an image acquired by the image acquisition device or an image acquired by the image guidance device. This application embodiment uses the example shown in Figure 1 for illustration only.

[0064] In some embodiments, when the subject to be radiotherapy is on the support device 1044, the rotation of the gantry 1041 can drive the treatment head 1042 to irradiate the subject in 360 degrees, thereby completing the radiotherapy.

[0065] Treatment planning equipment 102 is used to acquire planned images of the patient to be radiotreated from imaging equipment 101 and positional images of the patient to be radiotreated from image guidance device 1043 in radiotherapy equipment 104, in order to formulate, optimize, and evaluate treatment plans. Treatment planning equipment 102 may operate a radiotherapy planning system (TPS), which provides functions for formulating, optimizing, and evaluating treatment plans. For example, an RT pro TPS system.

[0066] In some embodiments, the treatment planning device 102 may include a TPS client 1021 and a TPS server 1022.

[0067] The TPS client 1021 can be at least one of the following devices: smartphone, smartwatch, desktop computer, laptop, virtual reality terminal, augmented reality terminal, wireless terminal, and laptop computer. For example, in some embodiments, a user can trigger the TPS server 1022 to execute an adaptive treatment plan optimization process by running a radiotherapy planning system on the TPS server 1022 through the TPS client 1021, and then display the optimized treatment plan. This effectively saves user time and provides a more intuitive presentation of the optimized treatment plan, allowing the user to evaluate it.

[0068] The TPS server 1022 can be a standalone physical server, a server cluster consisting of multiple physical servers, a distributed file system, or at least one of the following cloud servers providing basic cloud computing services: cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and big data or artificial intelligence platforms. This disclosure does not limit the specific implementation of these services. In some embodiments, the number of TPS servers 1022 can be greater or less, and this disclosure does not limit the implementation of these services. Of course, the TPS server 1022 can also include other functions to provide more comprehensive and diverse services. In some embodiments, the TPS server 1022 is used to provide background services for the TPS client 1021, such as performing an adaptive treatment plan optimization process.

[0069] Commonly used radiotherapy techniques include stereotactic radiotherapy (SRT), intensity-modulated radiotherapy (IMRT), and helical tomotherapy (HMT). Stereotactic radiotherapy delivers a single, high-dose irradiation with higher precision and is widely used to treat benign and malignant tumors, such as brain tumors, intracranial and extracranial vascular malformations. IRT uses multiple, low-dose irradiations and is widely used to treat tumors in various systems of the body, such as lung cancer, esophageal cancer, cervical cancer, and nasopharyngeal carcinoma of the head and neck. Helical tomotherapy utilizes a linear accelerator that rotates 360 degrees while the treatment bed moves unidirectionally to achieve rotational tomographic scanning. This invention provides an artificial intelligence-based intelligent decision-making method applicable to any type of radiotherapy.

[0070] This invention provides an intelligent decision-making method based on artificial intelligence. In radiotherapy, when target area deformation, positional shift, or dose deviation occurs, the intelligent decision-making model can quickly and accurately make a decision, thereby reducing or even eliminating treatment deviation problems caused by tumor movement, deformation, etc. during radiotherapy.

[0071] An intelligent decision-making method based on artificial intelligence, as shown in Figure 2, is provided in this embodiment of the invention, including:

[0072] S201. Obtain reference information for the target object. The reference information includes a pre-plan and a pre-plan image, wherein the pre-plan includes at least one fractionation plan, each fractionation plan includes multiple subfields, and the pre-plan includes the dose for each fraction and the dose for each subfield; the pre-plan image is associated with the pre-plan.

[0073] A radiotherapy plan (i.e., a treatment plan) includes information about the radiation delivery from the radiotherapy equipment to ensure the accuracy and effectiveness of the treatment. Each patient requiring radiotherapy has their own personalized radiotherapy plan, tailored to individual differences, tumor size, shape, location, and the specific conditions of surrounding tissues. The development of a treatment plan begins with determining the radiation site. Doctors determine the tumor's location and the target area to be irradiated based on the patient's tumor type and images acquired by the imaging equipment. Next, the target area is delineated, precisely outlining the area to be irradiated based on the images. After determining the target area, doctors design appropriate irradiation fields and allocate appropriate radiation doses based on the tumor's size, shape, and location. This aims to ensure that radiation accurately covers the tumor area while minimizing damage to surrounding healthy tissues. Finally, dose distribution verification is performed. Before actual radiotherapy, physical models or computer simulations are used to confirm that the predetermined radiation dose distribution meets expectations, ultimately forming the treatment plan. The treatment plan includes a prescription, namely the total dose to the target area and organs at risk, as well as the number of fractions and the execution plan for each fraction. The treatment plan is also sent to the radiotherapy equipment, which then performs the irradiation according to the treatment plan to complete the treatment.

[0074] Generally, a radiotherapy plan (also known as a treatment plan) is generated for cancer patients before treatment begins. According to the radiotherapy plan, radiation can be delivered to the patient in multiple fractions, i.e., multi-fractionation radiotherapy. Each fraction delivers a specific radiation dose to the patient. Multi-fractionation radiotherapy can also be distributed over multiple days, meaning the pre-plan includes multiple fractionation plans. For example, each fractionation plan includes multiple subfields. The intelligent decision-making method based on artificial intelligence provided by this invention can be used before the first fraction, between fractions of multiple treatments, or within a single fraction (for example, between multiple subfields or within a subfield).

[0075] In this embodiment of the invention, the pre-plan in the reference information is the treatment plan. In this embodiment, the pre-plan includes at least one fractionation plan. For example, a stereotactic radiotherapy treatment plan may have only one fractionation. The pre-plan may also include at least two fractionation plans, i.e., a multi-fractionation treatment plan. This embodiment of the invention uses a pre-plan including multiple fractionation treatment plans as an example for illustration; a single-fractionation treatment plan can refer to a multi-fractionation treatment plan.

[0076] In intensity-modulated radiotherapy (IMRT), a subfield refers to the division of the total irradiation area into multiple smaller irradiation zones. The dose distribution and shape of these subfields can be precisely adjusted according to the shape, size, and location of the tumor to achieve more accurate irradiation. IMRT can be VMAT, IMRT, etc., and the embodiments of this invention can be used for any type of IMRT. The pre-planning includes the dose for each fraction and the dose for each subfield.

[0077] In this embodiment of the invention, the pre-planned image is related to the pre-planning process. For example, it may be some or all images of the patient's treatment process; another example is an image used to delineate the target area. The pre-planned image may be acquired by different image acquisition devices, or it may be a fusion of images acquired by different image acquisition devices. Unless otherwise specified, the target object in this application may be the patient.

[0078] S202. Obtain current information about the target object. Current information includes at least one of: current image and current dose. For example, current information may only include the current image. Current information may also only include the current dose. Current information may also include both the current image and the current dose. For example, the current image may be an image acquired by an image acquisition device or an image guidance device. For example, the radiotherapy equipment also includes a detector plate disposed opposite the treatment head, which can be used to receive the radiation emitted from the treatment head that passes through the patient, thereby obtaining the treatment dose information. In this embodiment of the invention, the current dose can be obtained through the detector plate.

[0079] It should be noted that the order of steps 201 and 203 is not limited in the embodiments of the present invention. The specification and drawings are illustrated with step 201 first as an example.

[0080] S203. The intelligent decision-making model generates decision instructions based on the reference information and current information of the target object. These decision instructions include executing a pre-planned action, executing an adjusted plan, or issuing an alarm. In this embodiment of the invention, the intelligent decision-making model can be a trained artificial intelligence model. After learning from massive amounts of sample data, the intelligent decision-making model possesses the ability to issue decision instructions, thus enabling it to make such instructions. For example, the adjusted plan could be adjusting the shape and angle of the firing field, or optimizing the number of subfields, etc. For example, the intelligent decision-making model could generate different instructions based on different thresholds, or it could generate different instructions through comprehensive reasoning. For example, generating different instructions based on different thresholds could be based on a single threshold or multiple thresholds. That is, there could be multiple thresholds, and different instructions could be triggered based on multiple different thresholds.

[0081] The adjusted plan is generated based on the specific circumstances of dose deviation and / or volume deviation. This application does not impose any restrictions on the optimization elements of the adjusted plan; the above is merely an example.

[0082] For example, the adjusted plan can be generated by fine-tuning a pre-plan, or by deep optimization of a pre-plan, or by re-optimization. As another example, the adjusted plan can be generated by fine-tuning within a subfield, or by deep adjustment before, between, or before a subfield. This application does not restrict the timing of generating the adjusted plan; the above examples are merely illustrations. For instance, fine-tuning can also be performed between subfields.

[0083] This invention provides an intelligent decision-making method based on artificial intelligence. The intelligent decision-making model learns from clinical decision-making experience through training on a large number of samples. With rich clinical decision-making experience, it can quickly and accurately make decisions when the target area deforms, shifts in position, or deviates in the radiotherapy process, thereby reducing or even eliminating treatment deviations caused by tumor movement, deformation, etc. during radiotherapy.

[0084] An intelligent decision-making method provided in this embodiment of the invention includes an intelligent decision-making model that generates decision instructions based on reference information and current information of a target object. This includes: obtaining a volume deviation based on a pre-planned image in the reference information of the target object and a current image in the current information; and / or obtaining a dose deviation based on a dose in the reference information of the target object and a current dose; and the intelligent decision-making model generates decision instructions based on the volume deviation and / or dose deviation.

[0085] For example, the intelligent decision-making model can generate decision instructions based on reference information and current information of the target object, or it can generate volume deviation based solely on the pre-planned image in the reference information and the current image in the current information. In this case, the intelligent decision-making model generates decision instructions based on the volume deviation. As shown in Figure 3, the intelligent decision-making method includes:

[0086] S301. Obtain reference information for the target object. Refer to step 201 in Figure 2 for details.

[0087] S302. Obtain the current information of the target object. See step 202 in Figure 2 for details.

[0088] S303. Generate a volume deviation based on the pre-planned image in the reference information and the current image in the current information. For example, the volume deviation can be obtained by registering the pre-planned image and the current image. For example, the volume deviation can be obtained by registering the contours of the pre-planned image and the contours of the current image. For example, the volume deviation can also be obtained by registering features of the pre-planned image and features of the current image. This invention does not specifically limit the method of obtaining the volume deviation; the above examples are merely illustrative.

[0089] S304. Generate decision instructions based on volume deviation. The intelligent decision-making model generates decision instructions based on volume deviation. For example, when the volume deviation is less than or equal to a first threshold, execute the pre-planned treatment; when the volume deviation is greater than the first threshold but less than or equal to a second threshold, execute the adjusted plan; when the volume deviation is greater than the second threshold, trigger an alarm. Alternatively, it could be that when the volume deviation is less than the first threshold, execute the pre-planned treatment; when the volume deviation is greater than or equal to the first threshold but less than or equal to the second threshold, adjust the treatment plan; when the volume deviation is greater than the second threshold, trigger an alarm.

[0090] For example, volume deviation can be the deviation distance of the projected profile of a three-dimensional region of interest (ROI), the difference in the coverage area of ​​the ROI projection area, or the deviation between the three-dimensional region surrounded by a specified percentage dose and the three-dimensional region of the target area.

[0091] The three-dimensional region of interest (ROI) includes the target area and the organ at risk. For example, the deviation distance of the ROI projection profile can be the deviation distance of the target area projection profile, the deviation distance of the organ at risk profile, or the deviation distance of the projection profile including both the target area and the organ at risk. The difference in coverage area of ​​the ROI projection region can also be the difference in coverage area between the target area and / or the organ at risk projection regions.

[0092] For example, the threshold can be a value exceeding the target area. For example, an alarm is triggered if the value exceeds 2mm beyond the target area boundary in any direction; if the value exceeds 2mm beyond the target area, the treatment plan is adjusted; and the pre-planned treatment is executed within the target area. For example, the target area can be GTV (Gross Tumor Volume), CTV (Clinical Tumor Volume), ITV (Internal Target Volume), or PTV (Planning Target Volume).

[0093] In this application, the specific values ​​of the first threshold and the second threshold are not limited, and different values ​​can be set according to clinical needs.

[0094] For example, the adjusted plan may involve adjusting the shape and angle of the radiation field, and may also optimize the number of subfields. The adjusted plan is generated based on the specific circumstances of dose deviation and / or volume deviation. This application does not limit the optimization elements of the adjusted plan; the above is merely an example.

[0095] Alternatively, the intelligent decision-making model can generate decision instructions based on the reference information and current information of the target object, or it can generate dose deviations based solely on the dose and current dose in the target object's reference information. The intelligent decision-making model generates decision instructions based on the dose deviations. As shown in Figure 4, the intelligent decision-making method includes:

[0096] S401. Obtain reference information for the target object. See step 201 in Figure 2 for details.

[0097] S402. Obtain the current information of the target object. See step 202 in Figure 2 for details.

[0098] S403. Generate a dose deviation based on the dose in the target object reference information and the current dose. For example, the dose deviation can be the difference between the dose in the target object reference information and the current dose.

[0099] S404. Generate decision instructions based on dose deviation. The intelligent decision model generates decision instructions based on dose deviation. For example, when the dose deviation is less than or equal to a first threshold, execute the pre-planned action; when the dose deviation is greater than the first threshold but less than or equal to a second threshold, execute the adjusted plan; when the dose deviation is greater than the second threshold, trigger an alarm.

[0100] For example, dose deviation can be the deviation between the planned irradiation dose and the actual irradiation dose. Dose deviation can also be the deviation between the patient's actual absorption and the planned dose. The planned irradiation dose and the planned dose can be the dose pre-planned in the reference information.

[0101] For example, the first threshold is 1 MU and the second threshold is 5 MU. MU is a dose control unit used by radiotherapy equipment to achieve a specific dose distribution. In this embodiment of the invention, the first and second thresholds can also be other values ​​set as needed. This embodiment of the invention does not impose specific limitations on this; for example, the first threshold can be set to 2 MU and the second threshold to 7 MU, or the first threshold can be set to 1 MU and the second threshold to 7 MU. This embodiment of the invention uses a first threshold of 1 MU and a second threshold of 5 MU as an example for illustration.

[0102] Alternatively, the intelligent decision-making model can generate decision instructions based on reference information and current information of the target object. This may include generating volume deviation based on the pre-planned image in the reference information and the current image in the current information, and generating dose deviation based on the dose in the target object's reference information and the current dose. The intelligent decision-making model then generates decision instructions based on the volume deviation and dose deviation. As shown in Figure 5, the intelligent decision-making method includes:

[0103] S501. Obtain reference information for the target object. Refer to step 201 in Figure 2 for details.

[0104] S502. Obtain the current information of the target object. See step 202 in Figure 2 for details.

[0105] S503. Generate volume deviation based on the planned image in the reference information and the current image in the current information. See step 303 in Figure 3 for details.

[0106] S504. Generate a dose deviation based on the dose in the target object reference information and the current dose. See step 403 in Figure 4 for details.

[0107] S505. The volume deviation and dose deviation are weighted to obtain the calculation result. "Weighted" means "multiplied by a weight". For example, different weights are set for volume deviation and dose deviation. The volume deviation and dose deviation are weighted to obtain the calculation result, that is, the volume deviation and dose deviation are multiplied by a coefficient respectively.

[0108] S506. Generate decision instructions based on the calculation results. The intelligent decision model generates decision instructions based on the calculation results. For example, when the calculation result is less than or equal to the first threshold, execute the pre-planned action; when the calculation result is greater than the first threshold but less than or equal to the second threshold, execute the adjusted plan; when the calculation result is greater than the second threshold, issue an alarm.

[0109] It should be noted that the embodiments of the present invention do not limit the order between steps 503 and 504 in the above method. The specification only uses the above as an example for illustration.

[0110] In this embodiment of the invention, there is no limitation on whether the volume deviation and / or dose deviation must equal a threshold to trigger intelligent decision-making. Taking the intelligent decision-making model generating decision instructions based on volume deviation and dose deviation as an example, it could be that if the calculated result is less than a first threshold, the pre-plan is executed; if the calculated result is equal to the first threshold but less than a second threshold, the adjusted plan is executed; and if the calculated result is equal to the second threshold, an alarm is triggered. Alternatively, if the calculated result is less than the first threshold, the pre-plan is executed; if the calculated result is equal to the first threshold but less than or equal to the second threshold, the adjusted plan is executed; and if the calculated result is greater than the second threshold, an alarm is triggered.

[0111] In this embodiment, the current information also includes current clinical information. The intelligent decision-making model generates decision instructions based on the reference information and current information of the target object. Specifically, the intelligent decision-making model generates decision instructions based on the volume deviation, dose deviation, and current clinical information of the target object. For example, clinical information includes the patient's medication, surgery, other treatments, physical condition, complications, etc. In this embodiment, the input to the intelligent decision-making model includes dosimetric information, imaging information, and clinical information. Because personalized clinical information for the patient is added, personalized treatment decisions can be achieved.

[0112] In this embodiment of the invention, the pre-planning includes at least two sub-plannings. The intelligent decision-making method provided in this embodiment of the invention can perform intelligent decision-making between sub-plannings or within sub-plannings. The following describes the intelligent decision-making model performing intelligent decision-making within and between sub-plannings respectively.

[0113] This invention provides an intelligent decision-making method. Obtaining a pre-plan for a target object includes: obtaining the current treatment plan for the target object. Obtaining current information about the target object includes: obtaining at least one of: obtaining real-time treatment images of the target object for the current treatment session, and obtaining the real-time cumulative dose of the target object for the current treatment session. For example, the pre-plan includes at least two treatment sessions. For instance, the pre-plan includes ten treatment sessions. If the patient is currently undergoing the second treatment session, then obtaining the current treatment plan for the target object is equivalent to obtaining the second treatment session plan. For example, obtaining real-time treatment images of the target object for the current treatment session allows for further determination of volume deviation based on the real-time treatment images and the pre-planned images. For example, obtaining the real-time cumulative dose of the target object for the current treatment session allows for further determination of dose deviation based on the real-time cumulative dose and the pre-planned dose. The intelligent decision-making model executes the current treatment plan, executes an adjusted plan, or issues an alarm based on the volume deviation and / or dose deviation.

[0114] This invention provides an intelligent decision-making method that acquires current information about a target object, including acquiring real-time treatment images of the target object in the current treatment session. An intelligent decision-making model generates a volume deviation based on a pre-planned image in reference information and the real-time treatment images of the target object in the current treatment session. The intelligent decision-making model then generates a decision instruction based on the volume deviation. In this decision instruction, executing the pre-planned treatment means executing the treatment plan for the current treatment session. It should be noted that the real-time treatment images can be generated by an image-guided device during treatment. When the image-guided device includes a set of X-ray tubes and detectors that can rotate with the gantry, the real-time treatment image can be a tomographic CBCT image. When the image-guided device is a CT scanner, the real-time treatment image can be a CT image. When the image-guided device is an MRI scanner, the real-time treatment image can be an MRI image. Or, when the image-guided device includes two sets of X-ray tubes and detectors, the real-time treatment image is a dual-KV image. This application does not limit the type of real-time treatment image; the above examples are used for illustrative purposes only.

[0115] This invention provides an intelligent decision-making method. During the current treatment phase, an image guidance device can acquire real-time treatment images of the target object for the current phase. A volume deviation is generated based on the pre-planned image in reference information and the real-time treatment images of the target object for the current phase. The intelligent decision-making model then generates instructions to execute the phase plan, adjust the phase plan, or issue an alarm based on this volume deviation. This allows for deviation adjustment during the real-time treatment phase, improving treatment accuracy.

[0116] This invention provides an intelligent decision-making method. The pre-planned image includes a three-dimensional pre-planned volume of a target object. Obtaining the real-time treatment image of the target object in the current segment includes: obtaining the real-time treatment image of the target object in the current segment, and generating the current three-dimensional volume based on the real-time treatment image of the current segment. Obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: obtaining the volume deviation based on the three-dimensional pre-planned volume of the target object and the current three-dimensional volume. It should be noted that the pre-planned image in this embodiment includes a pre-generated three-dimensional pre-planned volume, and the pre-planned image can also include different types of three-dimensional volumes such as CT, MRI, and CBCT. For example, if the real-time treatment image can be a tomographic CBCT image, then generating the current three-dimensional volume based on the real-time treatment image in the current segment includes obtaining multiple consecutive tomographic CBCT images and generating the current three-dimensional volume based on the multiple consecutive tomographic CBCT images. This application does not limit the treatment image used in the embodiments; only tomographic CBCT images are used as an example for illustration.

[0117] For example, as shown in Figure 6, this is a schematic diagram of an intelligent decision-making model that makes decisions based on volume deviation within multiple iterations, according to an embodiment of the present invention, including:

[0118] S601. Obtain the current treatment plan for the target patient. For example, the pre-plan includes at least two treatment sessions, and these at least two sessions are identical. For instance, the pre-plan includes 10 treatment sessions. If the patient is currently receiving their second treatment, obtaining the current treatment plan for the target patient is equivalent to obtaining the treatment plan for the second session.

[0119] S602. Obtain a pre-planned image of the target object, wherein the pre-planned image includes a three-dimensional pre-planned volume of the target object. For example, the three-dimensional pre-planned volume is a three-dimensional pre-planned volume generated from a simulated localization CT image.

[0120] S603. Obtain the real-time treatment image of the target object for the current treatment session, and generate the current three-dimensional volume based on the real-time treatment image of the current treatment session. For example, obtain the real-time treatment image of the target object for the second treatment session. This real-time treatment image can, for example, be a three-dimensional volume image (i.e., the current three-dimensional volume) generated by multiple two-dimensional projections generated by the image guidance device.

[0121] S604. Obtain the volume deviation based on the three-dimensional pre-planned volume and the current three-dimensional volume of the target object.

[0122] S605. Execute the current batch plan, the adjusted plan, or an alarm based on the volume deviation.

[0123] This application provides an intelligent decision-making method for obtaining the real-time cumulative dose of a target object in the current session, comprising: obtaining the actual dose of the target object to each subfield in the current session; obtaining the cumulative dose of the current session based on the actual dose of each subfield of the target object; and obtaining the dose deviation based on the total dose of the target object to the current subfield in the pre-planned dose and the cumulative dose of the current session. For example, a radiotherapy device includes a detector plate located opposite the treatment head. Obtaining the actual dose of the target object to each subfield in the current session can be achieved by the detector plate receiving radiation from the treatment head and calculating the actual dose based on back projection. Alternatively, the cumulative dose of the current session can be obtained based on the actual dose of each subfield of the target object. The dose deviation can be the difference between the total dose of the target object to the current subfield in the pre-planned dose and the cumulative dose of the current session.

[0124] For example, as shown in Figure 7, this is a schematic diagram illustrating an intelligent decision-making model that makes decisions based on dose deviation within fractions, according to an embodiment of the present invention, including:

[0125] S701. Obtain the current treatment plan for the target patient. For example, the pre-plan includes a treatment plan for at least two treatments, such as a pre-plan including a treatment plan for 10 treatments. If the patient is currently undergoing the second treatment, then obtaining the current treatment plan for the target patient is equivalent to obtaining the treatment plan for the second treatment.

[0126] S702. Obtain the total dose from the target object's pre-planned dose to the current subfield. For example, if the total prescribed dose is 30 Gy, comprising 10 fractions, then each fraction is 3 Gy. For example, each fraction comprises 180 subfields. If the current fraction is the 10th subfield of the second fraction being executed, obtaining the total dose from the target object's pre-planned dose to the current subfield is equivalent to obtaining the planned dose from the second fraction to the 10th subfield.

[0127] S703. Obtain the actual dose of the target object to each subfield in the current fraction. For example, obtaining the actual dose of the target object to each subfield in the current fraction is the actual dose of each subfield from the 0th subfield of the second fraction to the 10th subfield of the current fraction.

[0128] S704. Obtain the current fractional cumulative dose based on the actual dose of each subfield of the target object. For example, the cumulative dose is the sum of the doses of each subfield from the first subfield to the current 40th subfield. For example, the sum of doses can be obtained through the relationship in the dose calibration process.

[0129] S705. The dose deviation is obtained based on the total dose to the current subfield in the pre-planned dose for the target and the current cumulative dose in fractions.

[0130] S706. Execute the current fractional plan, the adjusted plan, or issue an alarm based on the dose deviation. For example, the first threshold is 1 MU and the second threshold is 5 MU. The intelligent decision model generates decision instructions based on the dose deviation: when the dose deviation is less than or equal to 1 MU, execute the pre-plan; when the dose deviation is greater than 1 MU but less than or equal to 5 MU, execute the adjusted plan; when the dose deviation is greater than 5 MU, issue an alarm.

[0131] In this embodiment of the invention, the order of the above steps is not limited; the above is only used as an example for illustration.

[0132] The following is an embodiment of the present invention providing an intelligent decision-making method for multiple stages.

[0133] The embodiments provided in this application offer an intelligent decision-making method. Obtaining a pre-plan for a target object includes: obtaining the current fractional plan for the target object. Obtaining current information about the target object includes: obtaining an image guidance image of the target object before the current fractional treatment; obtaining volume deviation based on the pre-planned image in the target object's reference information and the current image in the current information includes: obtaining the volume deviation based on the pre-planned image in the target object's reference information and the image guidance image of the target object; executing the pre-plan includes executing the fractional plan for the current fraction. For example, the image guidance image may be generated by an image guidance device during the previous fractional treatment. When the image guidance device is a CT scanner, the image guidance image may be a CT image. When the image guidance device is an MRI scanner, the image guidance image may be an MRI image. Or, when the image guidance device includes two sets of X-ray tubes and detectors, the real-time treatment image is a dual-KV projection image. This application does not limit the type of image guidance image; the above examples are used for illustrative purposes only.

[0134] This invention provides an intelligent decision-making method that, before each fractionated treatment, acquires an image guide image of the target object from the previous fraction. Based on the pre-planned image and the target object's image guide image in the reference information, a volume deviation is generated. The intelligent decision-making model then generates instructions to execute the fractionated treatment plan, adjust the fractionated treatment plan, or issue an alarm based on this volume deviation. If deviations occur during fractionated treatment due to target area deformation, contraction, or other reasons, the intelligent decision-making model can adjust for these deviations across multiple fractionated treatment sessions, thereby improving the accuracy of the fractionated treatment.

[0135] The embodiments provided in this application offer an intelligent decision-making method. The pre-planned image includes a three-dimensional pre-planned volume of the target object. Obtaining an image guidance image of the target object before the current treatment session includes: acquiring the image guidance image of the target object before the current treatment session, and generating the current three-dimensional volume based on the image guidance image of the target object. Obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: obtaining the volume deviation based on the three-dimensional pre-planned volume of the target object and the current three-dimensional volume. For example, the pre-planned image may also include different types of images, such as CT images, MRI images, CBCT images, etc. For example, a three-dimensional volume is generally generated after a CT image scan. The pre-planned image in this embodiment includes the already generated three-dimensional pre-planned volume. For example, the real-time treatment image is a tomographic CBCT image. Generating the current three-dimensional volume based on the real-time treatment image of the current session includes acquiring multiple two-dimensional projections from different angles and generating the current three-dimensional volume based on the multiple two-dimensional projections from different angles. This application does not limit the implementation quality image; only a tomographic CBCT image as the real-time treatment image is used as an example for illustration.

[0136] For example, the volume deviation can be obtained by comparing the planned 3D volume of the target object with the current 3D volume. This can be achieved by registering the planned 3D volume with the current 3D volume. This invention will not elaborate on the method of obtaining the volume deviation, but will only use the above example for illustration.

[0137] For example, as shown in Figure 8, this is a schematic diagram of an intelligent decision-making model that makes decisions based on volume deviation between stages according to an embodiment of the present invention. The intelligent decision-making method includes:

[0138] S801. Obtain the current treatment plan for the target patient. For example, the pre-plan includes at least two treatment sessions, and these at least two sessions can be identical. For instance, the pre-plan includes 10 treatment sessions. If the patient is currently undergoing their second treatment, obtaining the current treatment plan for the target patient is equivalent to obtaining the treatment plan for the second session.

[0139] S802. Obtain a pre-planned image of the target object, wherein the pre-planned image includes the three-dimensional pre-planned volume of the target object. See step 602 for details.

[0140] S803. Before this fractionated treatment, an image-guided image of the target object is acquired, and the current three-dimensional volume is generated based on the image-guided image of the target object. For example, the image-guided device is an MRI scanner, and the image-guided image can be an MRI image. For example, the patient is currently undergoing the second treatment, and the image-guided image can be generated by the image-guided device during the first fractionated treatment.

[0141] S804. Calculate the volume deviation based on the pre-planned 3D volume and the current 3D volume of the target object. For example, refer to step 604.

[0142] S805. Execute the current segment's plan, the adjusted plan, or issue an alarm based on the volume deviation. That is, the intelligent decision-making model generates decision instructions based on the volume deviation. For example, when the volume deviation is less than or equal to a first threshold, execute the pre-planned action; when the volume deviation is greater than the first threshold but less than or equal to a second threshold, execute the adjusted plan; when the volume deviation is greater than the second threshold, issue an alarm.

[0143] This application provides an intelligent decision-making method that, if the volume deviation is acceptable, further includes: obtaining the current fractionated planned dose through dose prediction or dose recalculation; and the intelligent decision-making model generating a decision instruction based on whether the fractionated planned dose meets the prescription requirements. For example, in step 804 above, the volume deviation is obtained based on the three-dimensional pre-planned volume and the current three-dimensional volume of the target object. If the volume deviation is clinically acceptable, the current fractionated planned dose is obtained through dose prediction or dose recalculation before step 805. If the fractionated planned dose meets the dose requirements, the current fractionated plan is executed.

[0144] The intelligent decision-making method provided in this application further includes, before executing the adjusted plan, the intelligent decision-making model: generating the adjusted plan; the intelligent decision-making model acquiring the adjusted plan, wherein at least a portion of generating the adjusted plan is parallel to generating the decision instruction. For example, generating the adjusted plan can be parallel to steps 201 and 202 shown in Figure 2. Alternatively, generating the adjusted plan can be parallel between steps 202 and 203 shown in Figure 2. This application does not impose specific limitations on the parallel generation of the adjusted plan; the above examples are merely illustrative.

[0145] For example, the adjusted plan can be generated based on reference information and current information of the target object. Alternatively, the adjusted plan can be generated based on image deviation and / or dose deviation. The adjusted plan can also be generated in other ways, which are not limited in this application. Alternatively, the intelligent decision-making model of this application can directly obtain the adjusted plan. For example, the adjusted plan can be generated by other software or a treatment plan generation intelligent decision-making model, and after generation, it is sent to the intelligent decision-making model of this application embodiment, so that the intelligent decision-making model of this application can directly obtain the adjusted plan. During the process, the generation of the adjusted plan and the generation of decision instructions can be independent and parallel.

[0146] For example, the adjusted plan can be generated during the treatment pause or during the treatment exit. For example, the treatment pause can be included during the execution of the treatment plan, or it can be a treatment pause command triggered by dose deviation and / or volume deviation. For example, the intelligent decision-making method provided in this application embodiment further includes generating a treatment pause command based on dose deviation and / or volume deviation. Thus, an adjusted plan is generated during the treatment pause, and the intelligent decision-making model obtains and executes the generated adjusted plan.

[0147] In this embodiment of the invention, at least a portion of the generation of the adjusted plan is parallel to the generation of the decision instruction. The intelligent decision model generates the adjusted plan during the decision-making process and can directly obtain the adjusted plan to execute it, thereby shortening the decision-making time and enabling rapid decision-making.

[0148] The intelligent decision-making method provided in this application allows for the generation of an adjusted plan based on a pre-plan with fine-tuning, deep optimization based on a pre-plan, or re-optimization. For example, the adjusted plan can be generated with fine-tuning within a subfield, or with deep adjustment before, between, or before subfield iterations. This application does not limit the timing of generating the adjusted plan; the above examples are merely illustrations. For instance, fine-tuning can also be performed between iterations.

[0149] For example, the adjusted treatment plan can be generated by optimizing a pre-plan. For example, a 10-fraction treatment plan is obtained for the target area of ​​the patient to be radiotreated; before the start of the second fraction, a current image of the patient is acquired, and the treatment plan is optimized based on the current image; wherein the current image is used to indicate the morphology and / or location of the target area of ​​the patient. Alternatively, during the second fraction, within the continuous beam exit segment, a current image of the patient is acquired, and the treatment plan is optimized based on the current image. Alternatively, within the non-exit segment before the continuous beam exit segment, a current image of the patient is acquired, and the treatment plan is optimized based on the current image. This application does not impose specific limitations on the generation of the adjusted plan; the above examples are merely illustrative.

[0150] The intelligent decision-making method provided in this application includes generating at least two adjusted plans and then selecting and determining the final adjusted plan from among the at least two adjusted plans. For example, the intelligent decision-making model obtains multiple (e.g., four) adjusted plans, evaluates these plans, determines the final (one) adjusted plan, and executes it. Alternatively, generating the adjusted plan may involve generating multiple (e.g., four) adjusted plans, evaluating them, determining the final (e.g., one) adjusted plan, and then sending it to the intelligent decision-making model for execution.

[0151] The intelligent decision-making method provided in this application further includes: displaying the decision-making process or decision results. For example, the decision-making process or decision results can be displayed on a clinician's electronic device, such as a computer or mobile phone. For example, displaying the decision-making process may include displaying dose deviation and / or volume deviation, or displaying the progress of generating an adjusted plan. Decisions can also be displayed, such as displaying the execution of a pre-plan, displaying the execution of an adjusted plan, or displaying an alarm. This application does not limit the display method; for example, it may use different colors to display different decisions, text displays, or progress displays, or it may be a combination of different display methods.

[0152] In this embodiment of the invention, as shown in Figure 9, the intelligent decision-making model is obtained based on the following steps:

[0153] Step S901: Acquire multiple sample data and establish a sample database. For example, this could involve building a comprehensive medical big data sample platform, i.e., a sample database. This platform, through cooperation with multiple medical institutions, collects a large amount of clinical data on radiotherapy, i.e., sample data, which can cover various types of diseases.

[0154] Sample data includes multimodal information about the sample objects. For example, multimodal information includes, but is not limited to, images, treatment plans (images generating the plan, segmentation maps, dose field maps, and plan documents), treatment records, projected images of EPID, prescriptions, clinician decision-making information, and feature information. For example, multimodal information may also include motion change information of the sample objects, including but not limited to patient positioning and anatomical changes between sessions. For example, multimodal information also includes clinical information, such as patient medications, surgeries, other treatments, and physical condition. The inclusion of clinical information in multimodal information allows intelligent decision-making models to learn from richer or complementary data, thereby enhancing decision support.

[0155] Therefore, the inputs to the intelligent decision-making model include dosimetric information, imaging information, and clinical information. By adding personalized clinical information for patients, personalized treatment decisions can be achieved.

[0156] The sample images include, but are not limited to, one or more of CT, CBCT, MRI, and PET. The treatment plan includes at least two fractionated treatment plans. The clinician's decision information includes at least one fractionated decision made by the clinician based on the images of the sample subject. That is, the sample plan is a multi-fragmented treatment plan, and during this multi-fragmented treatment process, the clinician makes at least one fractionated decision. This decision may be to execute the plan, adjust the plan, or stop treatment, etc.

[0157] For example, inclusion criteria can be established before building a sample database. Inclusion criteria establish a strict set of standards for data inclusion. For example, the data for inclusion must have undergone at least two fractionated radiotherapy sessions and include the pre-planning of the previous fraction, multimodal information for subsequent fractions, and the physician's decisions regarding plan adjustments and their rationale. Further criteria can include good image quality, clear and accurate delineation of the target area and organs at risk, and reliable dose calculation results.

[0158] In this embodiment of the invention, the multimodal information can be in image format, text format, or video format. For example, an image can be in image format, clinical information, prescriptions, etc., can be in text format, and the clinical decision-making process can be in video format. This application does not limit the format of the multimodal information; the above examples are merely used for illustrative purposes.

[0159] Step S902: Multimodal information from multiple sample data points in the sample database is encoded separately using a neural network to achieve feature interactions between different modalities. This allows the intelligent decision-making model to capture the complex relationship between features and decisions. For example, multimodal information utilizes hierarchical feature fusion and feature interaction techniques of neural networks to integrate information from different modalities into a more comprehensive and abstract feature representation.

[0160] Step S903: The intelligent decision-making model performs deep learning on the sample data. For example, the input layer of the intelligent decision-making model includes the sample plan and multimodal information. The goal of this intelligent decision-making model is to achieve accurate decision output by learning the complex correlation between the sample data and the decision-making experience of clinicians. For example, the intelligent decision-making model may be for executing the plan, adjusting the plan, or stopping treatment.

[0161] The intelligent decision-making method provided in this application includes multiple intelligent decision-making models for different diseases. The intelligent decision-making models generate decision instructions based on reference information and current information of the target object. Specifically, the intelligent decision-making model corresponding to the current disease type generates decision instructions based on the reference information and current information of the target object. For example, the intelligent decision-making models include multiple intelligent decision-making models for different cancers such as lung cancer, gastric cancer, and cervical cancer. The intelligent decision-making models for different cancers learn from sample data corresponding to their respective diseases, thereby improving the accuracy of intelligent decisions for specific diseases.

[0162] The intelligent decision-making method provided in this application, before the intelligent decision-making model learns from the sample data, further includes: classifying the sample data according to disease type, so as to train corresponding intelligent decision-making models according to different diseases. For example, in order to facilitate subsequent model training and evaluation, the sample data is divided according to tumor type to construct a disease-specific database, and a single-disease intelligent decision-making model is trained based on the single-disease database.

[0163] The intelligent decision-making method provided in this application includes preprocessing multiple sample data during the construction of the intelligent decision-making model to achieve consistency and comparability of different sample data. For example, preprocessing includes, but is not limited to: accurate registration of image information from different modalities, performing delineation consistency checks, standardizing spatial information, intensity, and dose units of the data.

[0164] The intelligent decision-making method provided in this application, in the construction of the intelligent decision-making model, further includes encoding multimodal information of multiple sample data in a sample database using neural networks to achieve feature interaction between information of different modalities; based on the relationship between feature interaction capture and decision-making experience, it provides a wider range of features for the intelligent decision-making model. For example, multimodal information of sample data is extracted and encoded separately using neural networks, and multimodal feature fusion technology is used to fuse the information of each modality, further integrating the feature representations of different modalities through modal interaction technology. For example, multimodal feature fusion strategies include, but are not limited to, concatenation, element-level cascading, attention mechanisms, etc. Modal interaction technologies include, but are not limited to, multimodal fusion layers, graph convolutional networks, etc. Fusion strategies include, but are not limited to, early fusion, intermediate layer fusion, and late fusion.

[0165] For example, as shown in Figure 10, this is a schematic diagram of establishing a sample database according to an embodiment of this application, including:

[0166] S1001. Obtain multiple sample data sets. For example, this could be sample data for different diseases obtained from different hospitals. For example, the sample data could include lung cancer, stomach cancer, cervical cancer, etc.

[0167] S1002. Classify the sample data according to the disease. For example, classify the sample data according to the disease to form sub-sample databases corresponding to different diseases, such as lung cancer sample database, gastric cancer sample database, cervical cancer sample database, etc., so that the intelligent decision-making models for different diseases can learn from the sample data corresponding to the disease.

[0168] S1003. Preprocessing of multiple sample data. For example, image information from different modalities is precisely aligned using a registration algorithm. Consistency is assessed using quantitative indicators by comparing the target area or organ-at-risk delineation results from different doctors or at different time points. All images are then standardized to a unified coordinate system and resampled to a uniform spatial resolution. The intensity distribution and dose units of the images are adjusted and standardized. For example, registration methods include, but are not limited to, rigid, affine, and deformation registration. Quantitative indicators include, but are not limited to, the Dice coefficient and Hausdorff distance. This application does not impose a fixed order on the preprocessing methods; the order can be flexibly adjusted.

[0169] It should be noted that the embodiments of this application do not impose a specific order on the above method steps, and the order of multiple steps can be interchanged. For example, the order of steps 1002 and 1003 can be adjusted.

[0170] The intelligent decision-making method provided in this application allows for the periodic or regular acquisition of sample data to update the sample database required for intelligent model construction. The intelligent decision-making model then performs incremental training on the updated sample data. To continuously improve the performance of the decision-making model, an iterative optimization strategy is implemented, including periodically or regularly acquiring new sample data to update the sample database and performing incremental training based on the newly acquired data, thereby iteratively optimizing the model. For example, the decisions of clinicians or the intelligent decision-making model can be logged, and multimodal information also includes the logs, thereby increasing the sample diversity for the intelligent decision-making model to learn and enhancing decision support.

[0171] The intelligent decision-making method provided in this application also includes saving the process information and result information of the decision. Therefore, the information of the decision process and the result information can be used as new sample data for the intelligent decision-making model to relearn.

[0172] The intelligent decision-making method provided in this application further includes, in the case of a new disease different from those in the sample database: selecting an intelligent decision-making model for a disease similar to the new disease, and migrating and updating its parameters to generate a decision-making model for the new disease. In the intelligent decision-making method provided in this invention, the intelligent decision-making model is a specific intelligent decision-making model trained for different diseases. If a new disease appears, the similarity between the new disease and existing diseases is analyzed. If similar, the parameters of the selected existing disease intelligent decision-making model are migrated and updated to the new disease model to generate a decision-making model for the new disease.

[0173] The intelligent decision-making method provided in this application embodiment further includes the construction of the intelligent model as follows:

[0174] The first pre-trained model is then trained. Specifically, a lightweight first pre-trained model can be trained based on a portion of the sample data in the sample database.

[0175] The first pre-trained model is subjected to refined training, which includes at least one of model compression and hardware acceleration processing of the pre-trained model.

[0176] The first pre-trained model is optimized using a reinforcement learning algorithm to obtain the second pre-trained model; the intelligent decision-making model is the second pre-trained model. Specifically, the second pre-trained model can be trained using reinforcement learning based on partial sample data from the database or newly added sample data.

[0177] For example, based on multiple sample data initially entered into the database, they are divided into training, validation, and test sets. A lightweight first pre-trained intelligent decision-making model is constructed using lightweight convolutional neural networks and multimodal information fusion technology, based on the pre-planned information from the previous iteration and the multimodal information from the current iteration. Model compression and hardware acceleration techniques are employed to reduce computational resource consumption and improve the model's inference speed, thereby achieving rapid decision-making. Reinforcement learning algorithms such as deep Q-networks or proximal policy optimization are selected to train the intelligent decision-making model. The refined model serves as the initial policy for reinforcement learning, and the action space is defined as "execution plan" and "adjustment plan." During training, the intelligent decision-making model continuously learns and optimizes its decision-making strategy through interaction with the environment to maximize cumulative rewards and obtain better decision-making strategies. The obtained reinforcement learning model is established as the second pre-trained model, which is also an intelligent decision-making model for iterations.

[0178] The intelligent decision-making method provided in this application further includes the following steps in constructing the intelligent model: generating a third pre-trained model through online reinforcement learning on a second pre-trained model. The intelligent decision-making model is the third pre-trained model. Specifically, the third pre-trained model utilizes a highly realistic simulation environment and animal experimental models to accurately reproduce the real-time treatment scenario during radiotherapy, serving as the basis for training the third pre-trained model. After the model performance reaches clinical application standards, real clinical data is further used as input for online reinforcement learning to continuously optimize the model's decision-making ability. The third pre-trained model is also an intra-fractional intelligent decision-making model. For example, the training of the third pre-trained model aims to achieve real-time decision-making during intra-fractional radiotherapy. Through real-time feedback obtained from a highly realistic radiotherapy simulation environment and animal experimental models, the model can dynamically make optimal treatment decisions. First, a highly realistic radiotherapy simulation environment is constructed and an animal experimental model is incorporated to reproduce the real-time scenario of radiotherapy treatment. Then, the second pre-trained model is used as the initial strategy for online reinforcement learning. In this process, the action space is defined as including decision options such as "execute plan," "adjust plan," and "stop treatment." Online reinforcement learning algorithms, such as Actor-Critic, are used to train a third pre-trained model in simulated environments and animal experimental models. This model can learn and adapt to individual differences and real-time changes in patients during radiotherapy, thereby achieving optimal treatment decisions in intra-fractionated radiotherapy. After training, the resulting third pre-trained model serves as an intelligent prediction model for intra-fractionated adaptive radiotherapy and undergoes clinical validation. Finally, once the model meets the high standards required for real-time clinical application, actual clinical data will be used as input for online reinforcement learning to further improve the model's decision-making performance and adaptability.

[0179] For example, as shown in Figure 11, this is a schematic diagram of constructing an intelligent decision-making model according to an embodiment of the present invention, including:

[0180] S1101. Training to form the first pre-trained model. For example, the lung cancer sample data in the sample database includes 1000 samples. The data is divided into training set, validation set and test set in a ratio of 8:1:1. A lightweight first pre-trained model is formed by using a lightweight convolutional neural network (CNN) architecture and combining it with a cross-validation strategy.

[0181] S1102. Refine the training of the first pre-trained model. For example, the model undergoes compression, hardware acceleration, and other techniques to reduce the number of model parameters and optimize computational complexity, thereby further improving the decision response speed of the intelligent decision-making model.

[0182] S1103. The first pre-trained model is optimized using a reinforcement learning algorithm to obtain a second pre-trained model. For example, the first pre-trained model is used as the initial policy for reinforcement learning, and the policy is continuously iterated and optimized based on the rich dataset in the database and the interactive feedback with the environment. This second prediction model can be the intelligent decision-making model in the embodiments of this application.

[0183] S1104. The second pre-trained model is subjected to online reinforcement learning to generate a third pre-trained model. For example, in this process, the complex dynamic characteristics of a real radiotherapy environment are simulated, and real-time data streams from animal experimental models are used as data input and support for the online reinforcement learning algorithm to construct the third pre-trained model. After clinical validation, this model can be continuously updated and iterated using real clinical data through online reinforcement learning. This third prediction model can be the intelligent decision-making model in the embodiments of this application.

[0184] This application provides an electronic device, including: at least one storage device containing at least one set of instructions for intelligent decision-making; and at least one processor for communicating with the at least one storage device, wherein, when executing the at least one set of instructions, the at least one processor is configured to control the system to perform a method of any one of the embodiments of the present invention.

[0185] This application provides a non-transitory computer-readable medium including at least one set of instructions for generating intelligent decisions, wherein when executed by one or more processors of a computer device, the at least one set of instructions causes the computing device to perform the method of any one of the embodiments of the present invention.

[0186] According to embodiments of this disclosure, this disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the intelligent decision-making method provided by this disclosure.

[0187] In some embodiments, the electronic device may be the radiotherapy planning device shown in FIG1 above. FIG12 shows a schematic block diagram of an example electronic device 1200 that can be used to implement embodiments of the present disclosure. The electronic device 1200 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 1200 may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0188] As shown in Figure 12, the electronic device 1200 includes a computing unit 1201, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 1202 or a computer program loaded from a storage unit 1208 into a random access memory (RAM) 1203. The RAM 1203 can also store various programs and data required for the operation of the electronic device 1200. The computing unit 1201, ROM 1202, and RAM 1203 are interconnected via a bus 1204. An input / output (I / O) interface 1205 is also connected to the bus 1204.

[0189] Multiple components in electronic device 1200 are connected to I / O interface 1205, including: input unit 1206, such as keyboard, mouse, etc.; output unit 1207, such as various types of displays, speakers, etc.; storage unit 1208, such as disk, optical disk, etc.; and communication unit 1209, such as network card, modem, wireless transceiver, etc. Communication unit 1209 allows electronic device 1200 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0190] The computing unit 1201 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1201 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. The computing unit 1201 performs the various methods and processes described above, such as intelligent decision-making methods. For example, in some embodiments, the intelligent decision-making method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1208. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 1200 via ROM 1202 and / or communication unit 1209. When the computer program is loaded into RAM 203 and executed by the computing unit 1201, one or more steps of the intelligent decision-making method described above may be performed. Alternatively, in other embodiments, the computing unit 1201 may be configured to perform intelligent decision-making methods by any other suitable means (e.g., by means of firmware).

[0191] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard parts (ASSPs), systems-on-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0192] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0193] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory (EPROM or flash memory), optical fibers, compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0194] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user, such as a cathode ray tube (CRT) or liquid crystal display (LCD) monitor; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0195] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0196] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0197] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this disclosure can be achieved, and this is not limited herein.

[0198] Furthermore, this specification uses specific terms to describe embodiments thereof. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that "an embodiment," "one embodiment," or "an alternative embodiment" mentioned twice or more in different places in this specification do not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this specification can be appropriately combined.

[0199] Furthermore, those skilled in the art will understand that aspects of this specification can be described and illustrated in several patentable ways, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Therefore, aspects of this specification can be implemented entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of software and hardware implementations, all of which are generally referred to herein as “units,” “modules,” or “systems.” Furthermore, aspects of this specification can take the form of computer program products embodied in one or more computer-readable media, containing computer-readable program code embodied thereon.

[0200] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this specification are not intended to limit the order of the processes and methods described herein. Although various examples have been discussed in the foregoing disclosure of some embodiments of the invention that are currently considered useful, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the spirit and scope of the embodiments described herein. For example, while the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a purely software solution, such as an installation on an existing server or mobile device.

[0201] Similarly, it should be noted that, in order to simplify the description disclosed in this specification and thus aid in the understanding of one or more embodiments of the invention, multiple features may sometimes be grouped into a single embodiment, drawing, or description thereof in the foregoing description of the embodiments of this specification. However, the method described in this specification should not be construed as reflecting an intention that the claimed object to be scanned requires more features than expressly recited in each claim. In fact, the embodiments contain fewer features than all the features of the individual embodiments disclosed above.

[0202] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An intelligent decision-making method based on artificial intelligence, characterized in that, include: Acquire reference information for the target object, the reference information including a pre-plan and a pre-plan image, wherein the pre-plan includes at least one fractionation plan, each fractionation plan includes multiple subfields, and the pre-plan includes the dose for each fractionation and the dose for each subfield; the pre-plan image is associated with the pre-plan. Obtain current information about the target object, the current information including at least one of: current image and current dose; The intelligent decision-making model generates decision instructions based on the reference information and current information of the target object; The decision instructions include executing a pre-planned action, executing an adjusted action, or issuing an alarm.

2. The intelligent decision-making method according to claim 1, characterized in that, The intelligent decision-making model generates decision instructions based on reference information and current information of the target object, including: The volume deviation is obtained based on the pre-planned image in the reference information of the target object and the current image in the current information; and / or, The dose deviation is obtained based on the dose in the target object reference information and the current dose; The intelligent decision-making model generates decision instructions based on the volume deviation and / or dose deviation.

3. The intelligent decision-making method according to claim 2, characterized in that, The intelligent decision-making model generates decision instructions based on the volume deviation and / or dose deviation, including: When the volume deviation or dose deviation is less than or equal to a first threshold, the pre-planned procedure is executed; when the volume deviation or dose deviation is greater than the first threshold but less than or equal to a second threshold, the adjusted procedure is executed; when the volume deviation or dose deviation is greater than the second threshold, an alarm is triggered; or... The volume deviation and the dose deviation are weighted and calculated to obtain the calculation result; If the calculation result is less than or equal to the first threshold, execute the pre-planned action; if the calculation result is greater than the first threshold but less than or equal to the second threshold, execute the adjusted plan; if the calculation result is greater than the second threshold, trigger an alarm.

4. The intelligent decision-making model according to claim 2, characterized in that, Current information also includes current clinical information; The intelligent decision-making model generates decision instructions based on the reference information and current information of the target object. Specifically, the intelligent decision-making model generates decision instructions based on the volume deviation, the dose deviation, and the current clinical information of the target object.

5. The intelligent decision-making method according to claim 2, characterized in that, The preliminary plan includes at least two phased plans.

6. The intelligent decision-making method according to claim 5, characterized in that, Obtaining the pre-plan for the target object includes: obtaining the current session plan for the target object; The acquisition of current information of the target object includes at least one of: acquiring real-time treatment images of the target object in the current session and acquiring the real-time cumulative dose of the target object in the current session.

7. The intelligent decision-making method according to claim 6, characterized in that, The pre-planned image includes the three-dimensional pre-planned volume of the target object; The step of obtaining the real-time treatment image of the target object in the current segment includes: obtaining the real-time treatment image of the target object in the current segment, and generating the current three-dimensional volume based on the real-time treatment image of the current segment; The process of obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: The volume deviation is obtained based on the planned three-dimensional volume and the current three-dimensional volume of the target object.

8. The intelligent decision-making method according to claim 6, characterized in that, The process of obtaining the real-time cumulative dose of the target object in the current fraction includes: Obtain the actual dose of the target object to each subfield in the current fraction; The current cumulative dose is obtained based on the actual dose of each subfield of the target object; The method of obtaining the dose deviation based on the dose in the target object reference information and the current dose includes: obtaining the dose deviation based on the total dose to the current subfield in the target object pre-plan and the current fractional cumulative dose.

9. The intelligent decision-making method according to claim 2, characterized in that, The pre-plan for acquiring the target object includes: Obtain the current allocation plan for the target object. The acquisition of current information of the target object includes: acquiring an image guide image of the target object before this fractional treatment; The step of obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the image guidance image of the target object; The execution plan includes the execution of the current sub-session plan.

10. The intelligent decision-making method according to claim 9, characterized in that, The pre-planned image includes the three-dimensional pre-planned volume of the target object; The step of obtaining the image guide image of the target object before the treatment session includes: obtaining the image guide image of the target object before the treatment session; The current three-dimensional volume is generated based on the image of the target object; The step of obtaining the volume deviation based on the pre-planned image in the reference information of the target object and the current image in the current information includes: obtaining the volume deviation based on the three-dimensional pre-planned volume of the target object and the current three-dimensional volume.

11. The intelligent decision-making method according to claim 10, characterized in that, If the volume deviation is acceptable, it also includes: The current fractionation plan is generated by dose prediction or dose recalculation; The intelligent decision-making model executes the current dose plan based on whether the dose plan of the current dose meets the prescription requirements.

12. The intelligent decision-making method according to claim 1, characterized in that, The intelligent decision-making model also includes the following before executing the adjusted plan: Generate the revised plan; The intelligent decision-making model obtains the adjusted plan; In this process, at least a portion of the generated adjusted plan is generated in parallel with the generated decision instructions.

13. The intelligent decision-making method according to claim 12, characterized in that, The adjusted plan can be generated during the stop-beam period or the exit-beam period.

14. The intelligent decision-making method according to claim 12, characterized in that, The method of generating the adjusted plan includes generating at least two adjusted plans; the method further includes: The revised plan is selected from at least two revised plans.

15. The intelligent decision-making method according to claim 13, characterized in that, The intelligent decision-making model generates a termination instruction based on dose deviation and / or volume deviation.

16. The intelligent decision-making method according to claim 1, characterized in that, Also includes: Display the decision-making process or decision results.

17. The intelligent decision-making method according to claim 1, characterized in that, It also includes saving information about the decision-making process and the results.

18. The intelligent decision-making method according to claim 1, characterized in that, The intelligent decision-making model is obtained based on the following steps: Multiple sample data are acquired to establish a sample database. The sample data includes multimodal information of sample objects, wherein the multimodal information includes images, treatment plans, prescriptions, and clinician decision information, and the clinician decision information includes at least one set of decision information. Intelligent decision-making models learn from sample data.

19. The intelligent decision-making method according to claim 18, characterized in that, The multimodal information also includes clinical information.

20. The intelligent decision-making method according to claim 18, characterized in that, The multimodal information can be in image format, text format, or video format.

21. The intelligent decision-making method according to claim 18, characterized in that, The intelligent decision-making model includes multiple intelligent decision-making models for different diseases; The intelligent decision-making model generates decision instructions based on reference information and current information about the target object, specifically including: The intelligent decision-making model corresponding to the current disease type generates decision instructions based on the reference information and current information of the target object.

22. The intelligent decision-making method according to claim 18, characterized in that, The intelligent decision-making model includes multiple intelligent decision-making models for different diseases; before the intelligent decision-making model learns from the sample data, it also includes: The sample data were categorized according to the type of disease.

23. The intelligent decision-making method according to claim 22, characterized in that, In cases where new diseases differ from those in the sample database, this also includes: Acquire intelligent decision-making models for diseases similar to the new diseases; The parameters are migrated and updated to generate intelligent decision-making models for new diseases.

24. The intelligent decision-making method according to claim 18, characterized in that, Also includes: Acquire sample data periodically or at regular intervals; The intelligent decision-making model is incrementally trained on the updated sample data.

25. The intelligent decision-making method according to claim 16, characterized in that, Also includes: The multiple sample data are preprocessed to achieve consistency and comparability among different sample data.

26. The intelligent decision-making method according to claim 18, characterized in that, Also includes For the multimodal information of the same sample data, neural networks are used to encode it separately, realizing feature interaction between information of different modalities; Based on the relationship between feature interaction capture and decision-making.

27. The intelligent decision-making method according to claim 18, characterized in that, The intelligent decision-making model is obtained through the following steps: Based on the multiple sample data, a first pre-trained model is trained and formed; The first pre-trained model is subjected to refined training, wherein the refined training includes at least one of model compression and hardware acceleration processing of the pre-trained model; The first pre-trained model is optimized using a reinforcement learning algorithm to obtain the second pre-trained model; The intelligent decision-making model is the second pre-trained model.

28. The intelligent decision-making method according to claim 27, characterized in that, Also includes: The second pre-trained model is subjected to online reinforcement learning to generate a third pre-trained model; The intelligent decision-making model is the third pre-trained model.

29. An electronic device comprising: At least one storage device containing at least one set of instructions for intelligent decision-making; as well as, At least one processor is configured to communicate with the at least one storage device, wherein, when executing the at least one set of instructions, the at least one processor is configured to control the system to perform the method of any one of claims 1-28.

Images