Processing device, processing program, processing method, and processing system

Abstract

The present invention more efficiently controls a medical treatment beam.　A first trained determination model, which was generated by performing training on the basis of a plurality of first training medical images in which at least a portion of a person is included as a photographic subject and photographic subject label information indicating the location of a disease in each of the first training medical images, is subjected to additional training on the basis of one or a plurality of second training medical images acquired by imaging a subject of a medical treatment and subject label information indicating the location of the disease in the second training medical image or images, thereby acquiring a second trained determination model. Control information for a medical treatment beam with which the subject is irradiated during the medical treatment is generated on the basis of the location of the disease as determined by inputting a medical image for determination, which was acquired by imaging the subject, into the second trained determination model.

Description

Processing device, processing program, processing method, and processing system

[0001] The present disclosure relates to a processing device, a processing program, a processing method, and a processing system configured to perform processing related to controlling a treatment beam irradiated to a subject of treatment.

[0002] 2. Description of the Related Art Conventionally, there have been known processing devices for detecting the location of a disease such as a tumor and irradiating the disease location with a treatment beam to treat the disease such as a tumor. For example, Patent Document 1 describes a "motion-tracking irradiation device comprising: a linac that irradiates a tumor with a treatment beam; a tumor marker embedded near the tumor; a first X-ray fluoroscopy device that images the tumor marker from a first direction; a second X-ray fluoroscopy device that images the tumor marker from a second direction simultaneously with the first X-ray fluoroscopy device; first and second image input units that digitize the first and second fluoroscopic images output from the first and second X-ray fluoroscopy devices; first and second recognition processing units that perform template matching using a gray-normalized cross-correlation method in which a pre-registered template image of the tumor marker is applied to the image information digitized by the first and second image input units in real time at a predetermined frame rate, thereby determining first and second two-dimensional coordinates of the tumor marker; a central processing unit that calculates the three-dimensional coordinates of the tumor marker from the first and second two-dimensional coordinates calculated by the first and second recognition processing units; and an irradiation control unit that controls the irradiation of the treatment beam of the linac using the determined three-dimensional coordinates of the tumor marker."

[0003] Japanese Patent Application Laid-Open No. 2000-167072

[0004] Therefore, in light of the above-described technology, the present disclosure aims to provide a processing device, a processing program, a processing method, and a processing system that can control treatment beams more efficiently through various embodiments.

[0005] According to one aspect of the present disclosure, there is provided a processing device including at least one processor, wherein the at least one processor is configured to: acquire a second trained judgment model by additionally learning a first trained judgment model generated by learning based on a plurality of first training medical images that include at least a portion of a human body part as a subject and subject label information indicating the location of a disease in each of the first training medical images, based on one or more second training medical images obtained by photographing a subject of treatment and subject label information indicating the location of the disease in the second training medical images; and execute processing to generate control information for a treatment beam to be irradiated to the subject during the treatment based on the location of the disease determined by inputting judgment medical images obtained by photographing the subject into the second trained judgment model.

[0006] According to one aspect of the present disclosure, there is provided a processing program that, when executed by at least one processor, causes the at least one processor to function to generate control information for a treatment beam to be irradiated to the subject during the treatment based on the location of the disease determined by inputting judgment medical images obtained by photographing the subject into the second trained judgment model, using a first trained judgment model generated by learning based on a plurality of first training medical images that include at least some parts of a human as subjects and subject label information indicating the location of the disease in each of the first training medical images.

[0007] According to one aspect of the present disclosure, there is provided a processing method executed by at least one processor, the processing method including: a step of acquiring a second trained judgment model by additionally learning a first trained judgment model generated by learning based on a plurality of first training medical images including at least a portion of a human body part as a subject and subject label information indicating the location of a disease in each of the first training medical images, based on one or more second training medical images obtained by photographing a subject of treatment and subject label information indicating the location of the disease in the second training medical images; and a step of generating control information for a treatment beam to be irradiated to the subject during the treatment based on the location of the disease determined by inputting the judgment medical images obtained by photographing the subject into the second trained judgment model.

[0008] According to one aspect of the present disclosure, there is provided a processing system including at least one processor, wherein the at least one processor is configured to: acquire a first trained judgment model by additionally learning the first trained judgment model generated by learning based on a plurality of first training medical images that include at least a portion of a human body part as a subject and subject label information indicating the location of a disease in each of the first training medical images, based on one or more second training medical images obtained by photographing a subject of treatment and subject label information indicating the location of the disease in the second training medical images; and generate control information for a treatment beam to be irradiated to the subject during the treatment based on the location of the disease determined by inputting the judgment medical images obtained by photographing the subject into the second trained judgment model.

[0009] According to one aspect of the present disclosure, there is provided a processing method, executed by at least one processor, for generating a second trained judgment model used to generate control information for a treatment beam to be irradiated to a subject in treating the subject from a judgment image obtained by photographing the subject, the processing method including: a step of inputting one or more second training medical images obtained by photographing the subject of the treatment and subject label information indicating the location of the disease in the second training medical images to the first trained judgment model generated by the at least one processor learning based on a plurality of first training medical images including at least some parts of a human as subjects and subject label information indicating the location of the disease in each of the first training medical images; and a step of generating a second trained judgment model by additionally training the first trained judgment model.

[0010] According to the present disclosure, it is possible to provide a processing device, a processing program, a processing method, and a processing system that are capable of controlling a treatment beam more efficiently.

[0011] It should be noted that the above effects are merely illustrative for the sake of convenience and are not limiting. In addition to or instead of the above effects, any effect described in this disclosure or an effect obvious to a person skilled in the art may be achieved.

[0012] FIG. 1A is a block diagram showing a configuration of a processing system 1 according to an embodiment of the present disclosure. FIG. 1B is a diagram conceptually showing a configuration of a medical device 200 according to an embodiment of the present disclosure. FIG. 2 is a block diagram showing a configuration of a processing device 100 according to an embodiment of the present disclosure. FIG. 3A is a diagram conceptually showing a learning object management table stored in the processing device 100 according to an embodiment of the present disclosure. FIG. 3B is a diagram conceptually showing an example of a CT image and a simulation image according to an embodiment of the present disclosure. FIG. 3C is a diagram conceptually showing an example of a subject management table stored in the processing device 100 according to an embodiment of the present disclosure. FIG. 4A is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. FIG. 4B is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. FIG. 5 is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. FIG. 6 is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. FIG. 7 is a diagram conceptually showing the timing of generation of treatment beam control information according to an embodiment of the present disclosure.

[0013] 1. Configuration of Processing System 1 The processing system 1 according to the present disclosure is a system used to generate control information for a treatment beam irradiated to a treatment subject. In particular, the processing system 1 acquires a second trained judgment model by additionally training a previously generated first trained judgment model based on second trained medical images acquired by photographing the treatment subject and subject label information indicating the location of a disease, and controls the treatment beam based on the location of a disease determined by inputting the judgment-use medical images acquired by photographing the treatment subject into the second trained model.

[0014] Such treatment beams are primarily radiation beams such as X-rays, proton beams, and heavy particle beams. The irradiation position is preset so that the beam is appropriately irradiated to the disease location identified through CT images or the like during the treatment planning stage. However, during the actual treatment stage, the pre-identified disease location may move due to movements such as breathing of the treatment subject. Therefore, the treatment beam must be irradiated when the disease location overlaps with the pre-set irradiation position by imaging the subject at any interval during the treatment stage. The processing system 1 can more efficiently control the treatment beam.

[0015] 1A is a block diagram showing the configuration of a processing system 1 according to an embodiment of the present disclosure. According to FIG. 1A, the processing system 1 includes at least a medical device 200 including an imaging device 200-1 for imaging a subject of treatment and an irradiation device 200-2 for irradiating the subject with a treatment beam, and a processing device 100 for executing a process for generating control information for the treatment beam. The processing device 100 and the medical device 200 are communicatively connected to each other via at least one of a wireless communication network and a wired communication network.

[0016] Here, FIG. 1B is a diagram conceptually illustrating the configuration of a medical device 200 according to an embodiment of the present disclosure. Specifically, FIG. 1B is a diagram conceptually illustrating a radiation irradiation device, which is an example of the medical device 200 described in FIG. 1A. According to FIG. 1B, the medical device 200 is used to irradiate a treatment beam to a subject lying on a treatment table 210, also known as an examination table, couch, operating table, etc., installed relative to the medical device 200. Therefore, the medical device 200 includes a head 200-2a of an irradiation device 200-2 that emits the treatment beam. The medical device 200 is capable of moving a gantry 200-2b, on which the head 200-2a is installed, two-dimensionally or three-dimensionally relative to a base 200-2c so as to align the irradiation direction of the treatment beam with a preset direction.

[0017] Furthermore, as described above, since the location of the disease constantly moves due to the subject's movements such as breathing, it is necessary to image the subject to determine whether the location of the disease overlaps with the irradiation position of the predetermined treatment beam. Therefore, the medical device 200 includes, as the imaging device 200-1, an X-ray tube 200-1b that irradiates X-rays to image the subject, and a flat panel detector (FPD) 200-1a that images the X-rays irradiated from the X-ray tube 200-1b and transmitted through the subject. This allows the capture of a medical image for determination used to determine the location of the disease in the subject.

[0018] 1B shows an example of an X-ray image using the X-ray tube 200-1b and its flat panel detector 200-1a, but the medical image for determination is not limited to an X-ray image. The medical image for determination may be any image that can be used to determine the location of a disease, and in addition to an X-ray image, for example, an MR image, a CT image, a PET image, or a combination of these can also be suitably used.

[0019] 1B illustrates an example in which the imaging device 200-1 includes a pair of X-ray tubes 200-1b and flat panel detectors 200-1a, but two pairs of X-ray tubes and flat panel detectors with different X-ray irradiation directions may be used to acquire medical images for assessment. Alternatively, three or more pairs of X-ray tubes and flat panel detectors with different X-ray irradiation directions may be installed, and the optimal pair or two pairs of X-ray tubes and flat panel detectors may be selected depending on the orientation of the subject, the location of the disease, the orientation of the gantry 200-2b, etc. to acquire medical images for assessment.

[0020] Also, although FIG. 1B shows the use of a flat panel detector as the detector, other detectors such as a scintillator can also be used.

[0021] 1A , in the processing system 1, typically, when a medical image for determination captured by the imaging device 200-1 is transmitted to the processing device 100 via a communication network, the processing device 100 determines the disease location by inputting the received medical image for determination into a second trained determination model. The processing device 100 then generates control information based on the determined location and transmits the control information to the medical device 200. The medical device 200 controls the treatment beam irradiated from the irradiation device 200-2 based on the received control information. As a result, the treatment beam is irradiated onto the subject placed on the treatment couch 210, and the disease is treated.

[0022] In the present disclosure, the processing system 1 is exemplified as being configured only with the processing device 100 and the medical device 200. However, in addition to these devices, the processing system 1 can also be configured by combining various devices, such as an electronic medical record device, a database device, an interview device, various diagnostic devices, a model generation device, and a treatment planning device, depending on the application and purpose. Specifically, as described in detail below, the processing system 1 performs each process in multiple stages, such as a generation stage of the first and second trained determination models, a treatment planning stage, and a treatment stage. For example, the generation process of simulation images (first training medical images) used to generate the first trained determination model and the learning process using these images are executed by a server device or a model generation device. Furthermore, the process related to the generation of a treatment plan in the treatment planning stage is executed by a treatment planning device. Furthermore, the process related to positioning in the treatment stage and the process related to the generation of control information for controlling the treatment beam are executed by the processing device 100. By distributing the processes among multiple devices in this way, it becomes possible to perform treatment planning and learning for another subject in parallel, for example, while a treatment is actually being performed by irradiating a treatment beam.

[0023] Furthermore, in the present disclosure, a "subject" may be any person who can be the target of treatment, and is not limited to only those with specific attributes. Therefore, such a subject may include any person, such as a patient, a test subject, a diagnosed person, or a healthy person. Note that, although the term "person" is used simply in the present disclosure, this may refer to the subject, or may also refer to other people in addition to the subject.

[0024] In the present disclosure, the term "disease" refers to any disease that can be treated using a treatment beam, and is not limited to a specific disease. Examples of such diseases include brain tumors, head and neck cancer, esophageal cancer, lung cancer, breast cancer, hepato-biliary-pancreatic cancer, rectal cancer, cervical cancer, prostate cancer, skin cancer, malignant lymphoma, or a combination thereof.

[0025] Furthermore, although the present disclosure includes descriptions such as "first training medical image," "second training medical image," "first trained judgment model," and "second trained judgment model," these are merely names given to distinguish training medical images and trained judgment models from one another and do not imply a limitation to a specific number or order. Furthermore, unless otherwise specifically mentioned, the names with "first" or "second" attached may have different meanings or may have the same meaning.

[0026] Furthermore, in the present disclosure, "control information" may refer to any information used to control a treatment beam. Examples of such control information include, but are not limited to, the following: Information for instructing at least one of turning on and off irradiation of a treatment beam; Information for instructing at least one of the irradiation direction and intensity of a treatment beam; Information indicating the location of a disease determined based on a medical image for assessment; and Combinations of the above information. In the following, unless otherwise specified, a case will be described in which information for instructing at least one of turning on and off irradiation of a treatment beam is used as control information.

[0027] 2. Configuration of the Processing Device 100 FIG. 2 is a block diagram showing the configuration of the processing device 100 according to an embodiment of the present disclosure. According to FIG. 2, the processing device 100 includes a processor 111, a memory 112, and a communication interface 113. These components are electrically connected to one another via control lines and data lines. The processing device 100 does not need to include all of the components shown in FIG. 2; some components may be omitted, or other components may be added. For example, an external memory, a database device, a server device, a terminal device, or the like connected to the processing device 100 in a communicable manner may be used together with the processing device 100. Furthermore, some processing may be distributed and executed among other devices. In other words, the processing device 100 is not limited to a single device, but may be distributed across multiple devices depending on the information handling and processing load.

[0028] Such a processing device 100 can be used in various forms, such as a server device installed on the cloud, an on-premise server device installed in a specific facility or equipment, a terminal device used by medical personnel at a medical institution that operates the medical device 200, or a control device connected to the medical device 200 via a communication network.

[0029] The processor 111 functions as a control unit that controls other components of the processing system 1 based on a processing program stored in the memory 112. Specifically, the processor 111 executes processing related to generating control information for a treatment beam based on the processing program. Specifically, the processor 111 executes, based on the processing program stored in the memory 112, the following processes, among others: "a process of generating a first trained judgment model by learning based on a plurality of first training medical images including at least a portion of a human body as a subject and subject label information indicating the location of a disease in each of the first training medical images," "a process of acquiring a second trained judgment model by additionally training the first trained judgment model based on one or more second training medical images acquired by photographing a subject of treatment and subject label information indicating the location of a disease in the second training medical images," and "a process of generating control information for a treatment beam to be irradiated to a subject during treatment based on the location of a disease determined by inputting a judgment medical image acquired by photographing the subject into the second trained judgment model." The processor 111 is mainly composed of one or more CPUs, but may also be appropriately combined with a GPU, FPGA, or the like.

[0030] Note that the processing related to the generation of the first trained decision model does not necessarily have to be performed by the processing device 100, and may be performed by a device other than the processing device 100.

[0031] The memory 112 is composed of RAM, ROM, non-volatile memory, HDD, etc., and functions as a storage unit. The memory 112 stores instructions and commands for various control operations of the processing system 1 according to this embodiment as programs. Specifically, the memory 112 stores programs executed by the processor 111, such as "a process for generating a first trained judgment model by learning based on a plurality of first training medical images including at least a portion of a human body part as a subject and subject label information indicating the location of a disease in each of the first training medical images," "a process for acquiring a second trained judgment model by additionally training the first trained judgment model based on one or more second training medical images acquired by photographing a subject of treatment and subject label information indicating the location of a disease in the second training medical images," and "a process for generating control information for a treatment beam to be irradiated to a subject during treatment based on the location of a disease determined by inputting a judgment medical image acquired by photographing the subject into the second trained judgment model." In addition to the programs, the memory 112 also stores information from a subject management table and a subject management table. Furthermore, the memory 112 stores the first trained judgment model and the second trained judgment model as programs.

[0032] The memory 112 may include a database device for storing information, particularly information on the subject management table and the subject management table, via a communication network. The program for the first trained determination model does not necessarily have to be stored in the memory 112, but may be stored in a memory of a device other than the processing device 100.

[0033] The communication interface 113 functions as a communication unit for transmitting and receiving determination-use medical images and control information between the medical device 200 and other devices via a wired or wireless network. Examples of the communication interface 113 include a wired communication connector such as a USB, SCSI, or wired LAN, a wireless communication transmitting / receiving device for broadband wireless communication such as wireless LAN, Bluetooth (registered trademark), or LTE, or an infrared wireless communication, and various connection terminals for printed circuit boards or flexible circuit boards.

[0034] 3. Other Devices Although the specific configuration of the medical device 200 is not specifically illustrated, it includes a processor that functions as a control unit, a memory that functions as a storage unit, a communication interface that functions as a communication unit, etc. These elements are electrically connected to each other via control lines and data lines. Note that the medical device 200 does not need to include all of these components; it may be configured with some components omitted, or other components may be added.

[0035] 4. Various Information Used in Processing of the Processing System 1 Fig. 3A is a diagram conceptually illustrating a learning object management table stored in the processing device 100 according to an embodiment of the present disclosure. According to Fig. 3A, the object management table stores various information used to generate the first trained determination model. Specifically, the object management table stores CT images, simulation images, and object label information in association with object ID information.

[0036] The "subject ID information" is information for identifying a simulation image, which is a first training medical image used for training, and is information unique to each first training medical image. The subject ID information is newly generated each time a CT image used to generate the first training medical image is received.

[0037] A "CT image" refers to each of a plurality of CT images, or the entire set of a plurality of CT images, in which cross sections of a human subject are taken along a predetermined range (e.g., one or more body parts or the entire body) along the body axis. That is, the CT image includes at least a portion of the human body as the subject. Such a CT image may be image data, which is the image itself taken by a CT device, or processed image data obtained by subjecting the image data to image processing. Furthermore, a CT image may be either a still image or a video image, and may be in any image format, such as a black-and-white image or a color image. It may be a two-dimensional image, or may be a three-dimensional or four-dimensional image.

[0038] Here, the first trained judgment model is typically not a personalized trained model trained only on simulation images generated from CT images of a specific treatment subject, but a general-purpose trained model trained on multiple simulation images each containing multiple different subjects. Therefore, it is preferable that CT images are also taken of multiple people, rather than storing images of only a specific subject.

[0039] A "simulation image" is an image that simulates a two-dimensional X-ray image by integrating the CT values ​​of multiple CT images taken consecutively along the body axis direction along a projection line connecting the X-ray tube 200-1b and the flat panel detector 200-1a. Typically, a DRR (Digital Reconstructed Radiograph) image is used as such a simulation image. The simulation image is used as a first training medical image for training the first trained determination model.

[0040] As described above, the simulation image is obtained by integrating the CT values ​​of multiple CT images along the projection line connecting the CT tube 200-1b and the flat panel detector 200-1a. At this time, it is preferable to randomly shift the projection direction and position within a predetermined range. This makes it possible to obtain multiple simulation images with slightly different imaging directions and / or positions, thereby enabling more efficient collection of first training medical images. Furthermore, by learning from such multiple first training medical images, it is possible to compensate for deviations in imaging direction and position, as well as deviations in tumor position, between the simulation image and the actual medical image for evaluation (e.g., X-ray image) captured for treatment beam control information.

[0041] Here, the first trained judgment model is a general-purpose trained model as described above, and therefore, it is preferable that the generated simulation image is generated based on multiple CT images acquired with multiple people as subjects, rather than based on CT images acquired only from a specific subject.

[0042] Here, the simulation images are generated from multiple CT images of multiple cross sections. However, other images may also be used, such as images generated using a generative trained model, images drawn by input from an operator, simulated two-dimensional X-ray images, and pseudo X-ray images generated from CT images using Monte Carlo simulation. Image processing such as contrast adjustment, noise addition, and edge enhancement may be randomly performed within a predetermined range on the generated simulation images to generate multiple simulation images with different contrasts, noise, and edges. This allows multiple simulation images to be obtained from CT images taken of a single subject, thereby enabling more efficient collection of first training medical images. Furthermore, by using the multiple simulation images obtained as described above as the first training medical images, it is possible to compensate for differences in image quality between the simulation images and the actual medical images (e.g., X-ray images) used for evaluation and control of the treatment beam.

[0043] "Subject label information" is information indicating the location of a specific disease identified by a medical professional such as a doctor. The subject label information is typically generated by using information indicating the location of a disease for each CT image acquired as described above (e.g., contour information such as a clinical target volume (CTV), a planned specimen volume (PTV), a volume of interest (VOI), or a region of interest (ROI)). Such contour information can be generated by projecting the contour information added to the CT image in the same direction as the projection line used when generating the simulation image. Then, a body model including information indicating the disease location is reconstructed from each CT image into which information indicating the disease location has been input, and a simulation image is generated from the body model. This eliminates the need to add label information to each simulation image, enabling efficient generation of subject label information.

[0044] Such object label information can be various information such as a specific location where a disease is observed, an area including the location, a region where a disease is observed, or a combination of these. Note that the object label information is not limited to information obtained by a medical professional. It can also be obtained by preparing an image in which the location of a disease is specified in advance, calculating the degree of match between the image and a simulation image through image analysis processing, and then based on this degree of match. The object label information is used to generate the first trained determination model.

[0045] 3B is a diagram conceptually illustrating an example of a CT image and a simulation image according to an embodiment of the present disclosure, specifically, a diagram illustrating the relationship between the CT image stored in the object management table of FIG. 3A and a simulation image generated from the CT image.

[0046] CT images are typically acquired by capturing cross sections in the thickness direction of a human body over a predetermined range (one or more regions or the entire body) at spatially regular intervals (e.g., slice interval = 5 mm) in a direction from the head to the legs. In Fig. 3B, a plurality of CT images including at least CT image Bm-7 to CT image Bm and CT image Bn-7 to CT image Bn are captured over one or more regions at spatially regular intervals (e.g., slice interval = 5 mm).

[0047] As described above, the multiple CT images acquired in this manner were taken at regular intervals in the direction from the head to the legs. Therefore, by sequentially overlapping the CT images in this direction and integrating the CT images at the same coordinate positions, it is possible to virtually form a three-dimensional body model. The simulation images are pseudo X-ray images obtained by virtually transmitting radiographs of the pseudo body model from a predetermined direction. That is, according to FIG. 3B , simulation image Bm is generated as a pseudo X-ray radiograph of the body model reconstructed from CT images Bm-7 to Bm, and simulation image Bn is generated as a pseudo X-ray image of the body model reconstructed from CT images Bn-7 to Bn.

[0048] The CT image and the simulation image thus obtained are typically used as the CT image and the simulation image shown in FIG. 3A.

[0049] 3C is a diagram conceptually illustrating a subject management table stored in the processing device 100 according to an embodiment of the present disclosure. According to FIG. 3C, the subject management table stores treatment plan information, training medical images, subject label information, assessment medical images, and disease locations in association with subject ID information.

[0050] The "subject ID information" is information for identifying each subject who is the target of treatment and is information unique to each subject. The subject ID information is newly generated, for example, each time a subject registers for treatment at a medical institution or each time a subject is registered as a subject for radiation therapy.

[0051] "Treatment plan information" refers to information indicating a radiation therapy plan generated prior to treatment, for example, based on CT images acquired by imaging a subject using a CT scanner. Such treatment plan information includes various information, such as the intensity (e.g., energy dose) of the treatment beam, the irradiation direction, the irradiation field, the dose fractionation, and the irradiation dose. Among these pieces of information, the irradiation field indicates the irradiation position of the treatment beam and specifies a position (including both a specific point and a range including that point) in three-dimensional coordinate space in a pseudo-body model reproduced from the subject's CT images. The irradiation position is identified by a medical professional, such as a physician, identifying the location of the disease in each CT image.

[0052] The "training medical image" is a medical image acquired by photographing a subject using the imaging device 200-1 of the medical device 200 immediately before treatment or on the day of treatment, and is used as the second training medical image. In other words, the training medical image is an image in which the subject of treatment is the subject, and differs in this respect from the simulation image, which is the first training medical image that also includes subjects other than the subject of treatment. A typical training medical image is an X-ray image captured by the imaging device 200-1, but other images such as MR images, CT images, PET images, or any combination thereof can also be suitably used. The training medical image is used to generate a second trained judgment model, which is a personalized trained judgment model, by additionally training the first trained judgment model generated as a general-purpose trained judgment model.

[0053] The "subject label information" is information indicating the location of a specific disease in the training medical image of Figure 3C, which is the second training medical image. Such subject label information can be a variety of information, such as a specific point where the disease is present, an area including that point, a region where the disease is present, or a combination of these. Here, the disease assigned as label information is typically the same disease as the disease assigned to the first training medical image. Examples of such diseases include the tumors exemplified above. That is, if "tumor" is used as the target of the subject label information for the first training medical image, "tumor" will also be used as the subject label information for the second training medical image. Note that "same" here can include cases where the tumors are the same type, such as pharyngeal cancer, as well as cases where the tumors are different types.

[0054] The subject label information is typically obtained by inputting the second training medical image into a first trained judgment model that has been generated in advance. However, this method is not limited to this. The subject label information may be obtained by a medical professional such as a doctor modifying the information obtained from the first trained judgment model, by the medical professional identifying the location of the disease, or based on the degree of match with an image in which the location of the disease has been identified in advance. The subject label information is used to generate the second trained judgment model.

[0055] The "medical image for determination" is a medical image acquired by imaging the subject using the imaging device 200-1 of the medical device 200 after treatment for the subject has begun, and is information used to generate control information for irradiating the treatment beam. The medical image for determination is input into the second trained determination model and used to acquire information on the location of the disease. A typical example of the medical image for determination is an X-ray image captured by the imaging device 200-1, but in addition to these images, any of MR images, CT images, PET images, or combinations thereof can also be suitably used.

[0056] Here, the irradiation position of the treatment beam set as the treatment plan information is set in a three-dimensional coordinate space in which a pseudo-body model reproduced from CT images is placed. Therefore, if there is a misalignment between the coordinate space in which the medical image for assessment is formed and the coordinate space in which the irradiation position of the treatment beam is set, the treatment beam cannot be irradiated at the irradiation position set in the treatment plan in the first place. Therefore, the coordinate space in which the medical image for assessment is formed and the coordinate space in which the irradiation position of the treatment beam is set are formed by positioning the treatment couch 210 with the subject on it in advance so that they correspond to each other.

[0057] "Disease location" is information indicating the location of a disease contained in a medical image for assessment. Such a disease location is obtained by inputting the medical image for assessment into the second trained assessment model. The disease location can be indicated as various information, such as a specific point where the disease is observed, an area including that point, a region where the disease is observed, or a combination of these. The disease location is used to generate treatment beam control information based on a predetermined treatment beam irradiation position.

[0058] Although not specifically shown in Figure 3C, it is also possible to add various information such as medical interview information, electronic medical record information, and subject attribute information to the subject management table in association with the subject ID information.

[0059] 5. Processing Flow Executed by the Processing Device 100 (1) Processing Flow from the Treatment Planning Stage to the Treatment Stage Therapies using treatment beams (e.g., radiation therapy) are an important treatment method, alongside surgical therapy and chemotherapy, for treating tumors such as brain tumors, head and neck cancer, esophageal cancer, lung cancer, breast cancer, hepato-biliary-pancreatic cancer, rectal cancer, cervical cancer, prostate cancer, skin cancer, malignant lymphoma, or a combination thereof. However, these therapies are required to completely cure tumors or alleviate symptoms by minimizing the dose irradiated to normal cells surrounding tumor cells. Therefore, appropriate and individualized treatment plans must be created for each case, including the intensity, irradiation direction, irradiation field, dose fractionation, irradiation dose, and concomitant chemotherapy. Therefore, these treatments are generally divided into a treatment planning stage, which is performed in advance of the day of treatment; a positioning stage, which corrects for differences between the subject's position and posture during CT imaging in the treatment planning stage and the subject's position and posture during X-ray imaging in the treatment stage; and a treatment stage, in which the treatment is actually performed.

[0060] Fig. 4A is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. Specifically, Fig. 4A is a diagram showing a processing flow executed by the processor 111 of the processing device 100 from the treatment planning stage to the treatment stage. The processing flow is mainly executed by the processor 111 of the processing device 100 reading and executing a program stored in the memory 112.

[0061] 4A shows the steps of the process in chronological order, but the steps do not have to be performed consecutively. For example, after the steps of the treatment planning stage are completed, the steps of the positioning stage and treatment stage may be performed after an arbitrary period of time.

[0062] As shown in FIG. 4A , in the treatment planning stage, the processor 111 acquires CT images of a subject by receiving them from a CT scanner or a PACS (Personal Image Processing System) via the communication interface 113 (S101). Then, as shown in FIG. 3B , the processor 111 generates simulation images from a body model reconstructed based on the received CT images. The processor 111 then stores treatment plan information set by a medical professional, such as a doctor, based on the generated simulation images in the memory 112 (S102). As described in FIG. 3C , the treatment plan information includes various information such as the intensity (e.g., energy amount) of the treatment beam, the irradiation direction, the irradiation field indicating the irradiation position of the treatment beam, the dose fractionation, and the irradiation dose. Although not specifically shown, the processor 111 may store the acquired CT images and the body model (image data) in a three-dimensional coordinate space reconstructed based on the CT images in association with subject ID information.

[0063] Each process in this treatment planning stage is carried out in advance rather than on the day of treatment, since various considerations must be made, such as creating fixtures to fix the subject to the treatment table 210 and planning the treatment.

[0064] Systems using markers (e.g., tumor markers), such as the technology described in Japanese Patent Application Laid-Open No. 2000-167072 (Patent Document 1), have been known for some time. Therefore, the processing system 1 can also use such marker-based technology. This allows the processing system 1, which does not use markers, to switch to marker-based processing when detecting the location of a disease is difficult due to some inconvenience. However, in the case of marker-based technology, marker images are also included in the CT images taken in advance to place markers inside the subject's body and in the simulation images reproduced from the CT images. If such images are used as first training medical images, the presence or absence of a disease and its location would be determined based on the presence or absence of a marker, which is not desirable as training images. Therefore, in such cases, image processing may be performed to remove the marker images from the CT images or simulation images.

[0065] Next, a positioning step is performed immediately before treatment or on the day of treatment. This positioning step begins with placing the subject on the treatment couch 210 installed in the medical device 200, with the subject's movement restricted using a fixture or the like. After that, when an X-ray image is captured by the imaging device 200-1 of the medical device 200, the processor 111 of the processing device 100 receives the X-ray image from the medical device 200 via the communication interface 113 (S103). The processor 111 compares the received X-ray image with a body model reconstructed from the CT images captured during the treatment planning phase and calculates the positional deviation of the subject (S104). The processor 111 then generates control information for moving the position of the treatment couch 210 based on the calculated positional deviation (S105) and transmits the control information to the medical device 200 via the communication interface 113. The medical device 200 moves the position of the treatment table 210 based on the received control information to precisely set the positional relationship between the subject and the irradiation device 200-2, thereby ensuring that the coordinate space in which the X-ray image is formed and the coordinate space in which the irradiation position of the treatment beam is set correspond accurately to each other.

[0066] Such a positioning step is performed to adjust according to the subject's condition when the treatment is actually performed, and is therefore preferably performed immediately before the treatment or on the day the treatment is performed.

[0067] Next, in the treatment stage, the processor 111 executes a control information generation process for generating control information for the treatment beam (S106), and transmits the generated control information to the medical device 200 via the communication interface 113 (S107). Note that the processes of S106 and S107 are repeated every time a medical image for assessment is received from the medical device 200.

[0068] Each time the medical device 200 receives the control information, it operates the irradiation device 200-2 based on the control information and executes the process of irradiating the subject with a treatment beam. In this way, during the treatment stage, steps S106 to S107 in the processing device 100 and the irradiation process in the medical device 200 are repeatedly executed in accordance with the treatment plan information.

[0069] (B) Processing flow of control information generation processing Here, Fig. 4B is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. Specifically, Fig. 4B is a diagram showing the processing flow of the control information generation processing of S116 in Fig. 4A. This processing flow is mainly performed by the processor 111 of the processing device 100 reading and executing a program stored in the memory 112.

[0070] 4B, the processor 111 reads from the memory 112 a first trained determination model that has been generated in advance based on a simulation image, which is a first training medical image, and object label information (S111). Note that the process for generating the first training determination model will be described later in FIG. 5.

[0071] Next, the processor 111 acquires, via the communication interface 113, training medical images (second training medical images), which are X-ray images of the subject captured by the imaging device 200-1 of the medical device 200 (S112), and acquires subject label information assigned to the acquired training medical images (S113). As described above, irradiation of the treatment beam is affected by movements such as breathing of the subject. Therefore, the training medical images are acquired over a period of at least one respiratory cycle, i.e., at least one respiratory phase, by having the subject take at least one deep breath. That is, the processor 111 acquires multiple training medical images over at least one respiratory cycle and acquires subject label information for each of them. At this time, the processor 111 inputs each of the multiple training medical images into a first trained determination model generated in advance, thereby acquiring subject label information for each training medical image as output from the first trained determination model. The processor 111 stores the acquired learning medical image (second learning medical image) and subject label information in the subject management table in association with the subject ID information of the photographed subject.

[0072] The processor 111 reads the training medical image (second training medical image) and the subject label information associated with the subject's ID information from the subject management table, and performs additional training on the first trained judgment model based on this information. The processor 111 then acquires a second trained judgment model personalized for the subject through this additional training and stores the second trained judgment model in the memory 112 (S114). As described above, the first trained judgment model is a general-purpose trained model trained based on training medical images (first training medical images) including subjects other than the subject of treatment. In FIG. 4B , a personalized second trained judgment model is acquired by additionally training the general-purpose first trained judgment model using the subject's training medical images, etc., in S114. Therefore, it is possible to quickly generate a second trained judgment model that enables more accurate judgment for each subject with a short training time. The process for generating the second trained judgment model will be described later in FIG. 6.

[0073] Next, the processor 111 acquires a medical image for determination, which is an X-ray image of the subject captured by the imaging device 200-1 of the medical device 200, via the communication interface 113 (S115), and stores the acquired medical image for determination in the subject management table in association with the subject's subject ID information. The processor 111 reads the medical image for determination of the subject from the subject management table and inputs the image for determination into the generated second trained determination model (S116). The processor 111 then acquires information indicating the disease location of the disease contained in the medical image for determination as output from the second trained determination model (S117), and stores the acquired disease location in the subject management table.

[0074] After reading the disease location from the subject management table, the processor 111 executes a process of generating control information (S118). Examples of such control information include, as described above, information for instructing at least one of turning on and off irradiation of the treatment beam, information for instructing at least one of the irradiation direction and intensity of the treatment beam, information indicating the disease location determined based on the medical image for determination, and a combination of the above information.

[0075] FIG. 7 is a diagram conceptually illustrating the timing of generating treatment beam control information according to an embodiment of the present disclosure. Specifically, FIG. 7 is a diagram for explaining the generation of control information in relation to the irradiation position of the treatment beam and the disease position acquired in S117. According to FIG. 7, the irradiation position of the treatment beam, which was previously set as the irradiation field in the treatment planning stage, is shown as irradiation position 31 in a two-dimensional coordinate space. In contrast, FIG. 7 shows the disease positions acquired by performing the processes of S115 to S117 in FIG. 4B each time a medical image for assessment is acquired as disease positions 32a to 32e. In other words, it is shown that disease positions 32a to 32e move each time a medical image for assessment is acquired due to movement such as breathing of the subject.

[0076] When a disease location 32a that does not overlap the preset irradiation location 31 is acquired, the processor 111 skips the process of S118 in FIG. 4B without generating control information. Next, when a disease location 32b that partially overlaps the preset irradiation location is acquired, the processor 111 skips the process of S118 in FIG. 4B without generating control information. Next, when a disease location 32c that completely overlaps the preset irradiation location 31 is acquired, the processor 111 generates control information for instructing to turn on irradiation of the treatment beam. Next, when a disease location 32d that partially overlaps the preset irradiation location 31 is acquired, the processor 111 generates control information for instructing to turn off irradiation of the treatment beam. Finally, when a disease location 32e that does not overlap the preset irradiation location 31 is acquired, the processor 111 skips the process of S118 in FIG. 4B without generating control information.

[0077] That is, the processor 111 generates control information for instructing to turn on irradiation of the treatment beam while all of the acquired disease locations overlap with the preset irradiation location 31, and generates control information for instructing to turn off irradiation of the treatment beam when all of the locations no longer overlap with the preset irradiation location 31. In this way, by generating control information based on the positional relationship between the irradiation location 31 and the disease location, it is possible to reduce irradiation of the treatment beam to normal cells surrounding the tumor cells.

[0078] 7 illustrates a case in which processor 111 generates control information based on whether the entire acquired disease location overlaps with irradiation location 31. However, instead of this, processor 111 may generate control information to instruct turning on irradiation of the treatment beam while at least a portion of the disease location overlaps with irradiation location 31 (disease locations 32b to 32d in FIG. 7 ), and to instruct turning off irradiation of the treatment beam when at least a portion of the disease location no longer overlaps with irradiation location 31 (disease locations 32a and 32e in FIG. 7 ). Furthermore, processor 111 may generate control information to turn on or off the treatment beam based on whether the overlapping area between irradiation location 31 and the disease location exceeds a predetermined threshold.

[0079] 7, the case where the irradiation position 31 and each disease position have ranges has been described. However, instead, each disease position may indicate a specific point, and the processor 111 may generate control information for turning the treatment beam on or off based on whether the point is included in the irradiation position 31.

[0080] For convenience of explanation, FIG. 7 illustrates the irradiation position 31 and each disease location in a two-dimensional coordinate space defined by the x-axis and y-axis. However, instead, the irradiation position 31 and each disease location may be identified in a three-dimensional coordinate space including the z-axis, and the processor 111 may generate control information for turning the treatment beam on or off based on the degree of overlap between the two locations in the three-dimensional coordinate space. In such a case, although one pair of X-ray tube and flat panel detector is used in FIG. 1B , it is possible to obtain the disease location in three-dimensional coordinate space by using at least two pairs of X-ray tube and flat panel detector to identify the disease location from different directions. Note that, in this case, the disease locations identified in each X-ray image obtained using the two pairs of X-ray tube and flat panel detector should typically be identified at positions that coincide with each other in three-dimensional coordinate space. Therefore, the processor 111 may generate control information for turning off the treatment beam if the disease locations identified in each X-ray image differ by more than a predetermined threshold.

[0081] 7 illustrates a case in which the processor 111 of the processing device 100 determines whether to turn the treatment beam irradiation on or off and generates control information for instructing at least one of on and off. However, the processor 111 may generate information about each disease location as control information, and the processor of the medical device 200 may generate information for instructing at least one of on and off based on the control information using the method shown in FIG. 7. Furthermore, the processor 111 may generate information specifying the positional relationship between the irradiation location 31 and each disease location (e.g., information indicating whether the irradiation location 31 overlaps with the disease location) as control information, and the processor of the medical device 200 may generate information for instructing at least one of on and off based on the control information.

[0082] FIG. 7 illustrates a case in which control information is generated to instruct at least one of turning the treatment beam on and off. However, instead of or in addition to this, the control information may include information instructing the intensity and method of the treatment beam. For example, the processor 111 generates control information to increase the intensity as the overlap area between the irradiation position 31 and the disease position increases and decrease the intensity as the overlap area decreases. This enables a treatment beam with a higher intensity to be irradiated closer to the center of a disease such as a tumor. The processor 111 also generates control information to change the irradiation direction to track the disease position. This means that if the irradiation direction is fixed, the treatment beam needs to be turned on and off depending on the overlap with the decoy disease position, which requires additional treatment time. However, by tracking the irradiation direction to the treatment position, the off time can be minimized, enabling more efficient treatment beam irradiation.

[0083] (C) Processing flow for generating the first trained judgment model Figure 5 is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. Specifically, Figure 5 is a diagram showing a processing flow for generating the first trained judgment model read out in S111 of Figure 4B, i.e., for producing the first trained judgment model. This processing flow is mainly performed by the processor 111 of the processing device 100 reading and executing a program stored in the memory 112.

[0084] 5, the processor 111 generates a simulation image as a first training medical image using the method described in FIG. 3B and other figures based on CT images acquired by using a CT scanner to capture a subject other than the patient being treated (S211). Such a simulation image is typically a DRR image, which reproduces a two-dimensional X-ray image by integrating the CT values ​​of multiple CT images captured consecutively along the body axis along a projection line connecting the X-ray tube 200-1b and the flat panel detector 200-1a. However, in addition to this image, it is also possible to use an image generated using a generative trained model, an image drawn by input from an operator, or a simulated two-dimensional X-ray image.

[0085] Next, the processor 111 executes a process of storing object label information indicating the disease location where a predetermined disease exists in association with each of the generated simulation images (S212). The object label information may be associated by any method, such as by receiving input from a medical professional such as a doctor, or by matching with an image in which the disease location has been specified in advance.

[0086] Once multiple simulation images (first training medical images) and the object label information associated with each simulation image are obtained, the processor 111 performs machine learning of a disease location determination pattern (S213). For example, the machine learning is performed by providing pairs of the simulation images (first training medical images) and the object label information to a learning device including a neural network configured by combining neurons, and repeating learning while adjusting the parameters of each neuron so that the output of the neural network is the same as the label information. The processor 111 then acquires a first trained determination model (e.g., a neural network and its parameters) (S214). This completes the generation of the first trained determination model.

[0087] In the above, the first learned judgment model is generated by a neural network. However, instead of or in combination with a neural network, it is also possible to generate the first learned judgment model using machine learning such as a convolutional neural network, a multilayer Hercepton (MLP), a long short term memory (LSTM), a gated recurrent unit (GRU), a graph neural network (GNN), a transformer, or the like, a gradient boosting decision tree (GBDT) such as a light gradient boosting machine (LightGBM), XGBoost, or CatBoost, a ridge regression, a logistic regression, a support vector regression (SVR), a nearest neighbor method, a decision tree, a regression tree, or a random forest.

[0088] In addition, in Figure 5, a case has been described in which the first learned judgment model is generated by processing by the processor 111 of the processing device 100, but the first learned judgment model may also be generated in a similar manner by processing by a server device or model generation device other than the processing device 100.

[0089] (D) Processing flow for generating a second trained judgment model Figure 6 is a diagram showing a processing flow executed in the processing device 100 according to an embodiment of the present disclosure. Specifically, Figure 6 is a diagram showing a processing flow for generating the second trained judgment model in S114 of Figure 4B, i.e., for producing the second trained judgment model. This processing flow is mainly performed by the processor 111 of the processing device 100 reading and executing a program stored in the memory 112.

[0090] 6 , the processor 111 acquires a training medical image (e.g., an X-ray image) acquired by imaging the subject to be treated using the imaging device 200-1 of the medical device 200 as a second training medical image by receiving it from the medical device (S311). The processor 111 then associates and stores subject label information indicating the location of a predetermined disease with the acquired training medical image (S312). The subject label information is acquired as output from a first trained judgment model by the processor 111 inputting the second training medical image into a first trained judgment model generated in advance. Using this method, subject label information can be acquired more efficiently. However, the method is not limited to this. Information acquired from the first trained judgment model may be acquired by a medical professional, such as a doctor, making corrections to the information, by the medical professional identifying the location of the disease, or based on the degree of match with an image in which the location of the disease has been identified in advance.

[0091] Once a training medical image (second training medical image) containing the subject as a subject and subject label information associated with the training medical image are obtained, the processor 111 performs additional learning of a disease location determination pattern (S313). As an example, this additional learning involves providing a pair of the training medical image (second training medical image) and the subject label information to the first trained determination model generated as shown in FIG. 5 , and adjusting the parameters set in the first trained determination model so that the training medical image outputs the same results as the subject label information. In other words, by modifying the parameters of the first trained determination model generated as a general-purpose trained model based on the training medical image containing the subject as a subject and the subject label information, it is possible to obtain a trained model personalized to the subject.

[0092] Then, the processor 111 obtains a second learned judgment model by adjusting the parameters of the first learned judgment model as described above (S314).

[0093] Note that the generation of the second trained judgment model by the above-described additional learning is merely an example, and it may, of course, be obtained by other methods. For example, another neural network (second trained judgment model) that receives the output from the first trained judgment model as input is coupled to the first trained judgment model, and only a portion of the other neural network (second trained judgment model) is additionally trained using training medical images (second training medical images) and subject label information. This allows the coupled other neural network (second trained judgment model) to correct the output of the first trained judgment model and output a more accurate disease location. Although the coupled second trained judgment model is composed of a neural network, other machine learning methods can be used instead of a neural network, as exemplified in the first trained judgment model.

[0094] In addition, in Figure 6 etc., it is stated that the second trained judgment model is generated from the first trained judgment model, but this includes both cases where the second trained judgment model is essentially the first trained judgment model itself, such as when simply modifying the parameters of the first trained judgment model, and cases where the model configuration is the same as the first trained judgment model.

[0095] Typically, generating a personalized trained judgment model requires a large amount of training medical images and label information of the subject himself / herself, and the training itself takes time. However, in Figure 6, the processor 111 of the processing device 100 performs additional training on a general-purpose first trained judgment model using the subject's own training medical images, etc., to obtain a personalized second trained judgment model. Therefore, it is possible to generate a second trained judgment model personalized to the subject himself / herself more quickly, and it can be processed sufficiently even within a limited time, such as immediately before treatment.

[0096] As described above, in this embodiment, it is possible to provide a processing device, a processing program, a processing method, and a processing system that are capable of controlling a treatment beam more efficiently.

[0097] 6. Others FIG. 4B illustrates a case in which additional learning is performed on the first trained judgment model using training medical images captured by the imaging device 200-1. However, additional learning may also be performed on the first trained judgment model using simulation images generated from CT images of the subject taken during the treatment planning stage. In this case, the second trained judgment model is preferably obtained by processing in the following order: generation of the first trained judgment model; additional learning using simulation images generated from CT images of the subject; and additional learning using second training medical images (e.g., X-ray images) of the subject. In this way, more accurate disease location determination is possible by further performing additional learning using simulation images generated from CT images of the subject.

[0098] Furthermore, the first training medical images for the first trained judgment model were made using a plurality of simulation images containing at least a portion of a person other than the treatment subject as a subject. However, instead of this, the first trained judgment model may be generated using simulation images generated from CT images of the subject himself / herself as the first training medical images, and the first trained judgment model thus generated may be further trained using second training medical images (e.g., X-ray images) of the subject himself / herself to generate a second trained judgment model.

[0099] 1A to 7, the information on the disease location determined by the second trained determination model is used to generate control information for the treatment beam, but it may also be used for other purposes. For example, the information on the disease location can be accumulated as a medical history and used as information on the treatment progress.

[0100] Furthermore, various images such as CT images, simulation images, and X-ray images may not be image data itself, but may be image data that has been subjected to various pre-processing such as high definition, area extraction, noise removal, edge enhancement, image correction, and image conversion, using filter processing such as band pass filters including high pass filters and low pass filters, averaging filters, Gaussian filters, Gabor filters, Canny filters, Sobel filters, Laplacian filters, median filters, and bilateral filters, blood vessel extraction processing using Hessian matrices, segmentation processing of specific areas (e.g., disease areas) using machine learning, trimming processing of the segmented areas, haze removal processing, super-resolution processing, and combinations thereof.

[0101] 1A to 7, the imaging device 200-1 is used to determine the location of the disease, but it is also possible to use data captured or detected by other devices in addition to the imaging device 200-1.

[0102] The embodiments and modifications of the present disclosure are presented as examples and are not intended to limit the scope of the present disclosure. The embodiments and modifications can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the present disclosure. These embodiments and modifications are included within the scope and spirit of the invention, and are also included in the inventions and their equivalents as set forth in the claims.

[0103] The processes and procedures described in this disclosure can be realized not only by those explicitly described in the embodiments, but also by software, hardware, or a combination thereof. Specifically, the processes and procedures described in this disclosure can be realized by implementing logic corresponding to the processes in a medium such as an integrated circuit, volatile memory, non-volatile memory, magnetic disk, or optical storage. Furthermore, the processes and procedures described in this disclosure can be implemented as a computer program and executed by various computers, including processing devices and server devices.

[0104] Although processes and procedures described in this disclosure are described as being performed by a single device, software, component, or module, such processes or procedures may be performed by multiple devices, multiple software, multiple components, and / or multiple modules. Furthermore, although various information described in this disclosure is described as being stored in a single memory or storage unit, such information may be stored in multiple memories within a single device or multiple memories distributed across multiple devices. Furthermore, software and hardware elements described in this disclosure may be realized by integrating them into fewer components or by decomposing them into more components.

[0105] 1 Processing system 100 Processing device 200 Medical device 200-1 Imaging device 200-2 Irradiation device

Claims

1. A processing device having at least one processor, wherein the at least one processor is configured to perform processing to: obtain a second trained judgment model by additionally learning a first trained judgment model generated by learning based on a plurality of first training medical images that include at least some parts of a human body as subjects and subject label information indicating the location of a disease in each of the first training medical images, based on one or more second training medical images obtained by photographing a subject to be treated and subject label information indicating the location of the disease in the second training medical images; and generate control information for a treatment beam to be irradiated to the subject during the treatment based on the location of the disease determined by inputting the judgment medical images obtained by photographing the subject into the second trained judgment model.

2. The processing device according to claim 1, wherein the first training medical image is generated based on a CT image taken by a CT device.

3. The processing device according to claim 1, wherein the treatment is to treat the disease present at the location of the disease by irradiating the subject with a treatment beam.

4. The processing device according to claim 1, wherein the second training medical image is taken immediately before the treatment or on the day the treatment is performed.

5. The processing device of claim 1, wherein the second trained decision model is obtained immediately before the treatment or on the day the treatment is administered.

6. The processing device according to claim 1, wherein the control information is information for instructing at least one of turning on and off irradiation of the treatment beam.

7. The processing device according to claim 1, wherein the control information is information for instructing at least one of the irradiation direction and intensity of the treatment beam.

8. The processing device according to claim 7, wherein at least one of the irradiation direction and intensity is preset when planning the treatment and then controlled based on the location of the disease.

9. The processing device described in claim 1, wherein the control information is generated by overlapping at least a portion of the position of the disease determined in the second trained determination model with the predetermined irradiation position of the treatment beam.

10. The processing device described in claim 1, wherein the treatment beam is irradiated when at least a portion of the location of the disease determined in the second trained determination model overlaps with the predetermined irradiation position of the treatment beam.

11. The processing device according to claim 1, wherein the disease in the first training medical image and the disease in the second training medical image are the same disease.

12. The processing device according to claim 1, wherein the subject label information is obtained by inputting the second training medical image into the first trained judgment model.

13. A processing program that, when executed by at least one processor, causes the at least one processor to function as follows: a first trained judgment model generated by learning based on a plurality of first training medical images that include at least some parts of a human body as subjects and subject label information indicating the location of a disease in each of the first training medical images; a second trained judgment model obtained by additionally learning the first trained judgment model based on one or more second training medical images obtained by photographing a subject to be treated and subject label information indicating the location of the disease in the second training medical images; and a processing program that causes the at least one processor to function as follows: a first trained judgment model generated by learning based on a plurality of first training medical images that include at least some parts of a human body as subjects and subject label information indicating the location of a disease in each of the first training medical images; a second trained judgment model obtained by additionally learning the first trained judgment model based on one or more second training medical images obtained by photographing a subject to be treated and subject label information indicating the location of the disease in the second training medical images; and a second trained judgment model obtained by additionally learning the first trained judgment model based on the location of the disease determined by inputting the judgment medical images obtained by photographing the subject into the second trained judgment model.

14. A processing method executed by at least one processor, comprising: a step of acquiring a second trained judgment model by additionally learning a first trained judgment model generated by learning based on a plurality of first training medical images containing at least some parts of a human body as subjects and subject label information indicating the location of a disease in each of the first training medical images, based on one or more second training medical images obtained by photographing a subject to be treated and subject label information indicating the location of the disease in the second training medical images; and a step of generating control information for a treatment beam to be irradiated to the subject during the treatment, based on the location of the disease determined by inputting the judgment medical images obtained by photographing the subject into the second trained judgment model.

15. A processing system having at least one processor, wherein the at least one processor is configured to perform processing to: obtain a second trained judgment model by additionally learning a first trained judgment model generated by learning based on a plurality of first training medical images that include at least some parts of a human body as subjects and subject label information indicating the location of a disease in each of the first training medical images, based on one or more second training medical images obtained by photographing a subject to be treated and subject label information indicating the location of the disease in the second training medical images; and generate control information for a treatment beam to be irradiated to the subject during the treatment based on the location of the disease determined by inputting the judgment medical images obtained by photographing the subject into the second trained judgment model.

16. A processing method executed by at least one processor for generating a second trained judgment model used to generate control information for a treatment beam to be irradiated to a subject in treating the subject from a judgment image obtained by photographing the subject, the processing method comprising: a step of additionally training a first trained judgment model generated by the at least one processor by learning based on a plurality of first training medical images containing at least some parts of a human body as subjects and subject label information indicating the location of a disease in each of the first training medical images, by inputting one or more second training medical images obtained by photographing the subject to be treated and subject label information indicating the location of the disease in the second training medical images; and a step of generating a second trained judgment model by additionally training the first trained judgment model.

Images