Information processing system, information processing method, and information processing program

Abstract

This information processing system comprises at least one processor. The at least one processor acquires one or more target images indicating one or more target pelvic organs of a subject, and inputs each of the one or more target images to an image analysis model, which is trained so as to identify the pelvic organ and at least one lesion related to the pelvic organ from the input image, to generate one or more analysis result images indicating the result of processing by the image analysis model in a format in which each of the one or more target pelvic organs and a target lesion that is the lesion of the subject are distinguished from each other.

Description

Information processing system, information processing method, and information processing program

[0001] One aspect of the present disclosure relates to an information processing system, an information processing method, and an information processing program.

[0002] Conventionally, there is known a technique for obtaining various pieces of information from input images such as medical images using a machine learning model.

[0003] Endometriosis, an example of a lesion, is diagnosed by laparoscopy or by macroscopic findings during laparotomy. A definitive diagnosis of endometriosis is made by demonstrating "glandular structures and stroma similar to endometrium" in a pathological tissue specimen. Direct diagnosis is not performed in all cases, and in routine clinical practice, cases diagnosed comprehensively based on subjective symptoms, examination, and test findings are treated as clinical endometriosis. The accuracy rate of clinical endometriosis diagnosed by obstetrician-gynecologists is estimated to be approximately 80%. Ovarian chocolate cysts and deep endometriosis can be diagnosed by examination and imaging, but the diagnosis of subtle peritoneal lesions and mild adhesions is generally performed by direct vision.

[0004] Patent Literature 1 describes a system for assessing the severity of a vascular occlusion, which includes segmenting at least a portion of a volumetric image dataset into data segments corresponding to wall regions of a target organ, analyzing the data segments to extract features indicative of the amount of perfusion experienced by the wall regions of the target organ, obtaining a feature-perfusion classification model derived from a training set of perfused organs, classifying the data segments based on the extracted features and the FPC model, and providing as output a prediction indicative of the severity of the vascular occlusion based on the feature classification.

[0005] Special Publication No. 2020-511262

[0006] E. Pascoal, J. M. Wessels, M. K. Aas-Eng, M. S. Abrao, G. Condous, D. Jurkovic, M. Espada, C. Exacoustos, S. Ferrero, S. Guerriero, G. Hudelist, M. Malzoni, S. Reid, S. Tang, C. Tomassetti, S. S. Singh, T. Van den Bosch, M. Leonardi. “Strengths and limitations of diagnostic tools for endometriosis and relevance in diagnostic test accuracy research”, Ultrasound Obstet Gynecol. 2022 Sep;60(3):309-327.

[0007] There is a need for a system that allows accurate diagnosis of the pelvic organs.

[0008] An information processing system according to one aspect of the present disclosure includes at least one processor, which acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing a processing result of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is a lesion in the subject.

[0009] An information processing method according to an aspect of the present disclosure is executed by an information processing system including at least one processor, and includes the steps of acquiring one or more target images showing one or more target pelvic organs of a subject, inputting each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generating one or more analysis result images showing a processing result of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is a lesion in the subject.

[0010] An information processing program according to one aspect of the present disclosure causes a computer to perform the steps of acquiring one or more target images showing one or more target pelvic organs of a subject, inputting each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generating one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is a lesion in the subject.

[0011] In this aspect, by inputting a target image into the image analysis model obtained through learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Because the analysis result image shows the pelvic organs and lesions separately from each other, using this analysis result image enables accurate diagnosis of the pelvic organs.

[0012] According to one aspect of the present disclosure, it becomes possible to accurately diagnose pelvic organs.

[0013] FIG. 1 is a diagram illustrating an example of the functional configuration of an information processing system. FIG. 2 is a diagram illustrating an example of the hardware configuration of a computer that functions as an information processing system. FIG. 3 is a flowchart illustrating an example of processing of a target image executed by the information processing system. FIG. 4 is a diagram illustrating an example of generating an analysis result image. FIG. 5 is a diagram illustrating an example of determining adhesion. FIG. 6 is a diagram illustrating an example of displaying processing results. FIG. 7 is a flowchart illustrating an example of comparison of analysis result images executed by the information processing system.

[0014] Various examples of the present disclosure will be described in detail below with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are designated by the same reference numerals, and redundant description will be omitted.

[0015] [System Overview] The information processing system according to the present disclosure is a computer system that supports the diagnosis of a subject's pelvic organs. The information processing system is also referred to as a diagnosis support system. Pelvic organs refer to organs located in the pelvic cavity. A subject refers to a person who receives a diagnosis of their pelvic organs. In the present disclosure, the subject's pelvic organs are also referred to as "target pelvic organs."

[0016] In one example, the information processing system inputs a target image showing one or more target pelvic organs of a subject into an image analysis model and generates an analysis result image showing the processing result of the image analysis model. The image analysis model is a trained model trained by machine learning to identify pelvic organs and at least one lesion associated with the pelvic organs from the input image. In one example, the image analysis model is a trained model trained by machine learning to output segmentation information for each organ in an input image including multiple organs and identify the lesion, if present. Alternatively, the image analysis model may be a trained model trained by machine learning to simultaneously output segmentation information and the lesion for each organ in an input image including multiple organs. The image analysis model may include a first trained model that simultaneously outputs segmentation information for each organ and, if present, a first lesion in an input image including multiple organs, and a second trained model that identifies, if present, a second lesion based on the segmentation information for each organ.

[0017] Machine learning is a method of autonomously discovering laws or rules by repeatedly learning based on given information. The analysis result image shows the processing results of the image analysis model in a format that distinguishes one or more target pelvic organs from the target lesion, which is the lesion in the subject. Because the analysis result image shows the pelvic organs and the lesion separately, accurate diagnosis of the pelvic organs is possible by using this analysis result image.

[0018] In one example, the information processing system performs further processing based on the analysis result image. For example, the information processing system may process the analysis result image to determine whether adhesions exist between target pelvic organs. Alternatively, the information processing system may estimate a shape physical quantity, which is a physical quantity related to the shape of the target lesion, based on the analysis result image. Alternatively, the information processing system may estimate the severity of disease in the target lesion. Alternatively, the information processing system may compare analysis result images at two different time points and generate a comparison result. The information processing system may use a predetermined trained model trained by machine learning for at least part of these processes.

[0019] The pelvic organ to be diagnosed may be at least one of the uterus, ovaries, rectum, and bladder. The lesion identified by the image analysis model may be at least one of adhesions between organs, fibrous plaques, endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on the surface of the uterus other than the posterior surface of the uterus, rectovaginal septum lesions, rectal lesions, bladder lesions, ovarian cysts, endometriotic cysts, and adenomyosis lesions. The information processing system may process adhesions between organs as lesions or as a symptom different from lesions.

[0020] The information processing system may treat adhesions between organs as a symptom separate from lesions. For example, the image analysis model is trained to identify, from an input image, pelvic organs and at least one lesion associated with the pelvic organs, including fibrous plaques, ovarian cysts, rectovaginal septum lesions, rectal lesions, bladder lesions, endometriotic lesions, and adenomyosis lesions. When the image analysis model is used to assist in the diagnosis of endometriosis, for example, the image analysis model is trained to identify, from the input image, pelvic organs including the uterus and ovaries and at least one lesion, including endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on uterine surfaces other than the posterior surface of the uterus, and endometriotic cysts. The information processing system generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion, which is a lesion in the subject. The information processing system then processes the one or more analysis result images using an adhesion determination model to determine whether adhesions exist between the target pelvic organs.

[0021] The diagnosis support system according to the present disclosure may be used to support a diagnosis related to the uterus or ovaries. For example, the diagnosis support system may be used to support a diagnosis related to endometriosis, particularly deep endometriosis.

[0022] Typically, lesions or diseases related to the pelvic organs are diagnosed through surgery, such as laparoscopic surgery or laparotomy. However, such surgery is invasive and may increase the risk of complications. Furthermore, there is also deep endometriosis that cannot be detected through surgery. By using the information processing system (diagnostic support system) disclosed herein, it becomes possible to accurately diagnose pelvic organs without the need for such surgery. In addition, it is possible to reduce the physical burden on the subject, the burden on the evaluator (e.g., a physician), and the variability in diagnoses between evaluators.

[0023] A device corresponding to the information processing system according to the present disclosure can be referred to as an information processing device. In this case, the information processing system and the information processing device can be referred to as a diagnosis support device or a pathological condition evaluation device. The diagnosis support device may further include an imaging device that generates medical images. A doctor, medical professional, or other worker can operate the information processing device and refer to the generated evaluation information to diagnose a patient's pathological condition.

[0024] 1 is a diagram showing the functional configuration of an information processing system 10 according to an example. In this example, the information processing system 10 accesses a database 20 and a user terminal 30 via a communication network. The database 20 and the user terminal 30 may both be provided in a computer system separate from the information processing system 10, or may be components of the information processing system 10. The communication network may be configured by at least one of the Internet and an intranet.

[0025] The database 20 is a device that non-temporarily stores various data related to processing in the information processing system 10. For example, the database 20 may store target images, or may store processing results such as analysis result images, determination results regarding adhesions, estimated shape physical quantities, and disease severity in target lesions.

[0026] The user terminal 30 is a computer operated by a user such as an evaluator. The user terminal 30 may be any of various computers, such as a personal computer, a workstation, a tablet terminal, a smartphone, or a wearable terminal.

[0027] The information processing system 10 includes functional modules: an image acquisition unit 11, an image analysis unit 12, an adhesion determination unit 13, a lesion estimation unit 14, a result output unit 15, and a comparison unit 16. The image acquisition unit 11 is a functional module that acquires one or more target images showing one or more target pelvic organs of a subject. The image analysis unit 12 is a functional module that inputs each of the one or more target images into an image analysis model and generates one or more analysis result images that show the processing results of the image analysis model. The adhesion determination unit 13 is a functional module that processes one or more analysis result images to determine whether or not adhesions exist between target pelvic organs. The lesion estimation unit 14 is a functional module that performs estimation regarding the target lesion. The result output unit 15 is a functional module that outputs processing results including at least the analysis result image. The comparison unit 16 is a functional module that compares analysis result images at two different points in time and generates and outputs the comparison results.

[0028] FIG. 2 is a diagram showing an example of the hardware configuration of a computer 100 functioning as the information processing system 10. For example, the computer 100 includes a processor 101, a main memory unit 102, an auxiliary memory unit 103, a communication control unit 104, an input device 105, and an output device 106. The processor 101 executes an operating system and application programs. The main memory unit 102 is composed of, for example, ROM and RAM. The auxiliary memory unit 103 is composed of, for example, a hard disk or flash memory, and generally stores larger amounts of data than the main memory unit 102. The communication control unit 104 is composed of, for example, a network card or a wireless communication module. The input device 105 is composed of, for example, a keyboard, a mouse, a touch panel, etc. The output device 106 is composed of, for example, a monitor and speakers.

[0029] Each functional module of the information processing system 10 is implemented by an information processing program 110 pre-stored in the auxiliary storage unit 103. The information processing program 110 can also be considered a diagnostic assistance program or computer program code. Specifically, each functional module is implemented by loading the information processing program 110 onto the processor 101 or the main storage unit 102 and causing the processor 101 to execute the information processing program 110. The processor 101 operates the communication control unit 104, the input device 105, or the output device 106 in accordance with the information processing program 110, and reads and writes data from and to the main storage unit 102 or the auxiliary storage unit 103. Data, trained models, or databases required for processing may be stored in the main storage unit 102 or the auxiliary storage unit 103.

[0030] The information processing program 110 may be provided after being recorded on a non-transitory computer-readable storage medium such as a CD-ROM, a DVD-ROM, or a semiconductor memory. Alternatively, the information processing program 110 may be provided via a communication network as a data signal superimposed on a carrier wave. The provided information processing program 110 is stored in a memory such as the auxiliary storage unit 103.

[0031] The information processing system 10 may be configured with one computer 100 or multiple computers 100. When multiple computers 100 are used, these computers 100 are connected via a communication network such as the Internet or an intranet, thereby logically constructing a single information processing system 10.

[0032] [System Operation] (Processing of Target Image) As an example of an information processing method or a diagnostic support method according to the present disclosure, an example of processing of a target image executed by the information processing system 10 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing this example as processing flow S1.

[0033] In step S11, the image acquisition unit 11 acquires one or more target images showing one or more target pelvic organs of the subject. The target images are slice images showing cross sections of the target pelvic organs. For example, the image acquisition unit 11 acquires one or more target images showing at least one of the subject's uterus, ovaries, rectum, and bladder. In one example, the target images are medical images generated by any imaging device. The target images are obtained by photographing or measuring the human body and performing imaging. In one example, the image acquisition unit 11 acquires one or more MRI images obtained by magnetic resonance imaging (MRI) as the one or more target images. The MRI images may be sagittal images, coronal images, or axial images. Generally, sagittal images are used in MRI diagnosis of pelvic organs. The image acquisition unit 11 may read out one or more target images from the database 20 in response to an instruction signal from the user terminal 30. Alternatively, the image acquisition unit 11 may receive one or more target images from a predetermined imaging device such as an MRI device in response to the instruction signal. Alternatively, the image acquisition unit 11 may receive one or more target images transmitted from the user terminal 30.

[0034] In step S12, the image analysis unit 12 selects one of the one or more target images. For example, the image analysis unit 12 selects one target image in accordance with the order of the one or more target images arranged according to the imaging location on the subject's body.

[0035] In step S13, the image analysis unit 12 inputs the selected target image into an image analysis model to generate an analysis result image. The image analysis unit 12 uses one or more image analysis models. As described above, each image analysis model is a trained model that has been trained to identify pelvic organs and at least one lesion associated with the pelvic organs from the input image.

[0036] In one example, each image analysis model is trained to identify at least one lesion among a plurality of lesions associated with pelvic organs. When multiple image analysis models are used, the multiple image analysis models are trained independently of each other, i.e., the machine learning of each of the multiple image analysis models is performed separately from the other machine learning, e.g., without being influenced by the other machine learning.

[0037] When a single image analysis model is used, the image analysis unit 12 inputs the selected target image into the image analysis model and generates an analysis result image that shows the processing result of the image analysis model. As described above, the analysis result image shows the processing result in a format that distinguishes, for example, one or more target pelvic organs from the target lesion. As an example, when semantic segmentation is used, the image analysis model associates a label with each pixel of the target image, and the image analysis unit 12 generates the analysis result image by referring to each label.

[0038] When using multiple image analysis models trained independently of each other, the image analysis unit 12 inputs the selected target image into each of the multiple image analysis models to generate multiple provisional images. Each of the multiple provisional images represents the processing result of the corresponding image analysis model, for example, in a format that distinguishes one or more target pelvic organs from the target lesion. For example, the image analysis model associates a label with each pixel of the target image, and the image analysis unit 12 generates the provisional image by referring to each label. The image analysis unit 12 generates an analysis result image based on the multiple provisional images.

[0039] Generation of an analysis result image using multiple image analysis models will be described with reference to FIG. 4 . FIG. 4 is a diagram showing an example of the generation process. In this example, the image analysis unit 12 uses multiple image analysis models, including image analysis models 121, 122, and 123. The image analysis unit 12 inputs a selected target image 200 into each of these image analysis models to generate multiple tentative images 210, including tentative images 211, 212, and 213. The image analysis models and the tentative images have a one-to-one relationship. The image analysis unit 12 performs statistical processing on the multiple tentative images 210 to generate an analysis result image 220. The analysis result image 220 shows a bladder 221, a uterus 222, an ovary 223, and a rectum 224. For example, the image analysis unit 12 calculates the average pixel value for each of multiple voxels in the set of multiple tentative images 210 and generates the analysis result image 220 using the individual average pixel values. A set of multiple provisional images 210 for obtaining multiple voxels is obtained by virtually stacking the multiple provisional images 210 .

[0040] When generating an analysis result image based on multiple interim images, the image analysis unit 12 may calculate uncertainty for at least a portion of the analysis result image based on the multiple interim images and the analysis result image. Uncertainty is an index that quantitatively indicates the degree of variation in multiple processing results obtained by multiple image analysis models. The image analysis unit 12 may calculate uncertainty for a region of interest in diagnosing pelvic organs. In the example of FIG. 4 , the image analysis unit 12 calculates, for a partial region 231, uncertainty for each of multiple voxels obtained by assembling the multiple interim images 210 as voxel uncertainty based on the multiple interim images 210 and the analysis result image 220. The voxel uncertainty is the variation in multiple pixel values ​​in the multiple interim images 210 at that voxel, when the pixel value (average value) of the analysis result image 220 at that voxel is used as a reference. The image analysis unit 12 may generate a reference image 230 showing the distribution of voxel uncertainty in the region 231. The image analysis unit 12 calculates a statistical value (for example, an average value) of the voxel uncertainties in the region 231 as the uncertainty Us in the region 231. In the example of Fig. 4, the uncertainty Us is 0.002.

[0041] The region of interest may be arbitrarily set according to the target organ or disease. Calculating uncertainty specific to the lesion rather than evaluating the entire pelvic cavity further improves the interpretability of the prediction results.

[0042] In particular, when diagnosing endometriosis, there is a particular issue with lesions in that the boundary with the organ contour is unclear. Therefore, by setting the area around the contour of the posterior uterus as a region of interest and calculating the uncertainty focusing on the lesion on the posterior uterus, doctors can prevent false negatives of the lesion on the posterior uterus, thereby increasing the usefulness of the diagnostic support model. The image analysis unit 12 may calculate the uncertainty of the detection of the lesion in the area around the contour of the posterior uterus. The image analysis unit 12 may also have a function to output or display an alert when the uncertainty exceeds a threshold.

[0043] Returning to FIG. 3 , as shown in step S14, if there is an unprocessed target image (NO in step S14), the process returns to step S12. In the repeated step S12, the image analysis unit 12 selects the next target image. In the repeated step S13, the image analysis unit 12 inputs the target image to one or more image analysis models, respectively, to generate an analysis result image. In this way, the image analysis unit 12 inputs one or more target images to one or more image analysis models, respectively, to generate one or more analysis result images.

[0044] If all target images have been processed (YES in step S14), the process proceeds to step S15. In step S15, the adhesion determination unit 13 processes one or more analysis result images to determine adhesions between target pelvic organs. The adhesion determination unit 13 determines whether adhesions exist between target pelvic organs, i.e., whether two or more adjacent target pelvic organs are adhered to each other. For example, the adhesion determination unit 13 may determine at least one of adhesions between the uterus and rectum, adhesions between the left ovary and rectum, adhesions between the right ovary and rectum, adhesions between the left ovary and uterus, adhesions between the right ovary and uterus, adhesions between the left ovary and right ovary, and adhesions between the uterus and bladder. The adhesion determination unit 13 may also determine adhesions when target organs are adhered via plaques (nodules). The adhesion determination unit 13 may further determine the degree of each adhesion. The degree of adhesion may be expressed as "none," "mild," "severe," or the like. In this case, the adhesion determination unit 13 may use the size of the adhesion area to determine the degree of adhesion. Regarding the determination criteria, the adhesion determination unit 13 may determine that there is mild adhesion when the target pelvic organs are in contact with each other over a long area without the presence of intervening fat tissue, ascites, or when a thin, cord-like or tent-like low-signal area is observed between the target pelvic organs. The adhesion determination unit 13 may determine that there is severe adhesion when a thick, plate-like low-signal area is observed between the target pelvic organs, or when there is obvious deformation of the target pelvic organs due to adhesion.

[0045] In one example, the adhesion assessment unit 13 analyzes one or more analysis result images to generate target shape data and determines whether or not adhesions exist based on the target shape data. The target shape data is data representing a set of one or more parameter values ​​of the three-dimensional shapes of multiple target pelvic organs. The target shape data may also be referred to as feature values ​​related to the shape of each organ. For example, the adhesion assessment unit 13 calculates multiple parameter values ​​for each of the multiple target pelvic organs based on one or more analysis result images, such as surface area, volume, sphericity, flatness, length, major / minor axis lengths, axis lengths in the direction of the largest principal component, maximum diameters in the height / width / depth directions, maximum diameters in the 3D, and surface-to-volume ratios. At least some of these parameter values ​​may correspond to radiomics features. The adhesion assessment unit 13 analyzes individual analysis result images for each target pelvic organ to calculate one or more parameter values ​​for the target pelvic organ. The adhesion determination unit then generates target shape data indicating a set of calculated parameter values. The adhesion determination unit 13 inputs the target shape data into an adhesion determination model to determine whether or not adhesions exist. The adhesion determination model is a trained model that has been trained to identify adhesions between pelvic organs from shape data that is a set of parameter values ​​for the three-dimensional shapes of each of a plurality of pelvic organs.

[0046] In one example, one adhesion determination model determines whether or not adhesion exists between two predetermined pelvic organs. Therefore, an adhesion determination model is prepared for each combination of two pelvic organs to be determined. That is, the adhesion determination model determines adhesion between a first organ and a second organ among multiple pelvic organs. The adhesion determination unit 13 inputs target shape data corresponding to each combination of the first organ and the second organ into multiple adhesion determination models each having a different combination of the first organ and the second organ, and determines whether or not adhesion exists for each of the multiple combinations of the first organ and the second organ.

[0047] The adhesion determination will be described with reference to FIG. 5 . FIG. 5 is a diagram showing an example of the determination process. In this example, the adhesion determination unit 13 analyzes multiple analysis result images to generate object shape data related to the uterus, left ovary, right ovary, rectum, and bladder. The adhesion determination unit 13 then uses seven adhesion determination models 131 to 137 to determine adhesions for each of seven combinations of a first organ and a second organ. The adhesion determination model 131 receives object shape data corresponding to the uterus and rectum and determines adhesions between these organs. The adhesion determination model 132 receives object shape data corresponding to the left ovary and rectum and determines adhesions between these organs. The adhesion determination model 133 receives object shape data corresponding to the right ovary and rectum and determines adhesions between these organs. The adhesion determination model 134 receives object shape data corresponding to the left ovary and uterus and determines adhesions between these organs. The adhesion determination model 135 receives object shape data corresponding to the right ovary and uterus and determines adhesions between these organs. The adhesion determination model 136 receives object shape data corresponding to the left ovary and the right ovary and determines adhesions between these organs. The adhesion determination model 137 receives object shape data corresponding to the uterus and the bladder and determines adhesions between these organs. Each of the adhesion determination models 131 to 137 outputs adhesion data indicating the determination result.

[0048] Returning to FIG. 3 , in step S16, the lesion estimation unit 14 determines whether a target lesion exists. The lesion estimation unit 14 references each of the one or more analysis result images to determine whether an analysis result image showing a target lesion exists. The existence of at least one analysis result image showing a target lesion means that a target lesion has been identified in at least one target image corresponding to the at least one analysis result image. If a target lesion has been identified in at least one of the one or more target images (YES in step S16), the process proceeds to step S17. On the other hand, if a target lesion has not been identified in any of the one or more target images (NO in step S16), the process skips steps S17 and S18 and proceeds to step S19.

[0049] In step S17, the lesion estimation unit 14 analyzes one or more analysis result images to estimate a shape physical quantity of the target lesion. The shape physical quantity refers to a physical quantity related to the shape of the target lesion. The lesion estimation unit 14 may estimate at least one of the dimensions and volume of the target lesion as the shape physical quantity. The lesion estimation unit 14 may estimate at least one of the major axis and minor axis of the target lesion as the dimension. The major axis refers to the longest length among the lengths of imaginary axes crossing the target lesion. The minor axis refers to the shortest length among the lengths of imaginary axes perpendicular to the imaginary axis indicating the major axis. The major axis may be referred to as length, and the minor axis may be referred to as width or thickness. For example, if the target lesion is a fibrous plaque, the lesion estimation unit 14 may estimate the thickness of the fibrous plaque as the shape physical quantity. If the target lesion is an ovarian cyst, the lesion estimation unit 14 may estimate at least one of the volume, major axis, and minor axis of the ovarian cyst as the shape physical quantity.

[0050] In one example, the lesion estimation unit 14 measures the shape physical quantities of the target lesion for each of one or more analysis result images showing the target lesion using a predetermined measurement algorithm. Alternatively, the lesion estimation unit 14 may estimate the shape physical quantities for each of one or more analysis result images showing the target lesion by inputting the analysis result image into a physical quantity estimation model that has been trained to estimate physical quantities related to the shape of the lesion from an image showing the lesion. A physical quantity estimation model may be prepared for each lesion. After determining the shape physical quantities for each analysis result image, the lesion estimation unit 14 estimates the statistical values ​​of these shape physical quantities as the final shape physical quantities. Examples of such statistical values ​​include the maximum value, the average value, and the median.

[0051] In step S18, the lesion estimation unit 14 estimates the severity of the disease at the target lesion. The severity of the disease may be expressed as mild, severe, or the like. This allows a physician, who is a user of the system, to easily and comprehensively assess the disease. In one example, the lesion estimation unit 14 estimates the severity based on the estimated shape physical quantity. The lesion estimation unit 14 may refer to a predetermined correspondence table showing the correspondence between the shape physical quantity and the severity of the disease, and acquire the severity corresponding to the estimated shape physical quantity as the estimation result. As another example, the lesion estimation unit 14 may estimate the severity of the disease at the target lesion in response to the determination that adhesions exist between the target pelvic organs. For example, the lesion estimation unit 14 may refer to a predetermined correspondence table showing the correspondence between the degree of adhesion and the severity of the disease, and acquire the severity corresponding to the determined degree of adhesion as the estimation result. Alternatively, the lesion estimation unit 14 may refer to a predetermined correspondence table showing the correspondence relationships between shape physical quantities, the degree of adhesions, and the severity of disease, and acquire the severity corresponding to both the estimated shape physical quantities and the determined degree of adhesions as an estimation result. Alternatively, the lesion estimation unit 14 may estimate the severity using a severity estimation model trained to estimate the severity of disease at the lesion site from at least one of the shape physical quantities and the degree of adhesions. A severity estimation model may be prepared for each disease.

[0052] In step S19, the result output unit 15 generates and outputs processing results. In one example, the result output unit 15 generates and outputs processing results including one or more analysis result images, a determination result regarding adhesions, a shape physical quantity, and / or a disease severity level at the lesion. When outputting processing results including an analysis result image, the result output unit 15 may edit the analysis result image to highlight the target lesion and not highlight at least one of the one or more target pelvic organs, and generate processing results including the edited analysis result image. The result output unit 15 may generate processing results including uncertainty in at least a portion of the analysis result image. For example, the result output unit 15 may generate processing results including at least one of a reference image (e.g., reference image 230 shown in FIG. 4 ) showing the distribution of voxel uncertainty and an uncertainty value (e.g., uncertainty Us shown in FIG. 4 ) for each of the one or more analysis result images. Alternatively, the result output unit 15 may generate processing results including an alert regarding uncertainty in response to the uncertainty for at least one analysis result image exceeding a predetermined threshold. This alert is intended to inform the user that the uncertainty in the analysis result image is relatively high.

[0053] The result output unit 15 transmits the generated processing result to the user terminal 30. The user terminal 30 receives and displays the processing result. Therefore, the transmission of the processing result by the result output unit 15 is an example of outputting the processing result, an example of displaying the processing result on a display device, and an example of outputting an alert.

[0054] 6 is a diagram showing an example of displaying the processing results. The screen 300 shown in this example includes an analysis result image 310 edited to highlight the target lesion 311 and not highlight the target pelvic organ, a slide bar 320 that is a user interface for switching between the multiple analysis result images 310 to be displayed, and a determination result 330 regarding adhesions between the target pelvic organs. The user can operate the slide bar 320 to manually or automatically switch between the multiple analysis result images 310.

[0055] Furthermore, the result output unit 15 stores the generated processing results in the database 20. For example, the result output unit 15 generates a data record in which a subject ID, which is an identifier that uniquely identifies the subject, a shooting date and time indicating the time when one or more target images were taken, and the generated processing results are associated with each other, and stores this data record in the database 20.

[0056] The information processing system 10 may execute the process flow S1 for one or more target images for each of multiple subjects. As a result, the database 20 may store processing results for the multiple subjects. When one or more target images are acquired for a certain subject at each of multiple time points, the information processing system 10 may execute the process flow S1 for one or more target images at each time point. For example, the information processing system 10 may execute the process flow S1 for one or more target images at a time point before a therapeutic drug is administered to the subject and one or more target images at a time point after the therapeutic drug is administered to the subject. The information processing system 10 may execute the process flow S1 for one or more target images at each of multiple time points after the therapeutic drug is administered to the subject. As a result of such processing, the database 20 may store a history of processing results for a certain subject.

[0057] (Comparison of Analysis Result Images) As an example of an information processing method or a diagnostic support method according to the present disclosure, an example of comparison of analysis result images executed by the information processing system 10 will be described with reference to Fig. 7. Fig. 7 is a flowchart showing this example as processing flow S2.

[0058] In step S21, the comparison unit 16 acquires a comparison instruction from the user terminal 30. The comparison instruction is a data signal for requesting the information processing system 10 to compare analysis result images at two different points in time. Based on a user operation, the user terminal 30 generates a comparison instruction indicating the subject ID, a first point in time, and a second point in time different from the first point in time, and transmits this comparison instruction to the information processing system 10. The comparison unit 16 receives the comparison instruction.

[0059] In step S22, the comparison unit 16 acquires one or more first analysis result images generated based on one or more first target images at a first time point. The comparison unit 16 reads out one or more analysis result images corresponding to the subject ID and the first time point indicated by the comparison instruction from the database 20 as one or more first analysis result images. As described above, the first analysis result images are images obtained by inputting the first target image into an image analysis model.

[0060] In step S23, the comparison unit 16 acquires one or more second analysis result images generated based on one or more second target images at the second time point. The comparison unit 16 reads out one or more analysis result images corresponding to the subject ID and the second time point indicated by the comparison instruction from the database 20 as one or more second analysis result images. As described above, the second analysis result images are images obtained by inputting the second target image into an image analysis model.

[0061] In step S24, the comparison unit 16 compares one or more first analysis result images with one or more second analysis result images. In one example, the comparison unit 16 generates one or more pairs of first analysis result images and second analysis result images to be compared, and performs the following processing on the one or more pairs. That is, the comparison unit 16 aligns the positions of the two images so that the positions of one or more target pelvic organs match between the first analysis result image and the second analysis result image. The comparison unit 16 then compares the two images to determine whether the target lesion has changed. For example, the comparison unit 16 determines whether the size or position of the target lesion has changed.

[0062] In step S25, the comparison unit 16 generates and outputs the comparison result. The comparison unit 16 may generate the comparison result including at least one of a comparison image, which is an image showing the location of the change in the target lesion, and a numerical value indicating the amount of change in the dimension or position of the target lesion. The comparison unit 16 transmits the generated comparison result to the user terminal 30. The user terminal 30 receives and displays the comparison result. Therefore, the transmission of the comparison result by the comparison unit 16 is an example of a process of displaying the comparison result on a display device.

[0063] Furthermore, the comparison unit 16 stores the generated comparison result in the database 20. For example, the comparison unit 16 generates a data record in which the subject ID, a comparison ID which is an identifier that uniquely identifies the comparison process, and the generated comparison result are associated with each other, and stores this data record in the database 20.

[0064] As an application example of process flow S2, the comparison unit 16 may compare analysis result images at three or more time points in chronological order to generate a comparison result showing changes in the target lesion over the three or more time points. In this case, the comparison unit 16 treats two of the three or more time points as a first time point and a second time point and generates a comparison result corresponding to the two time points. The comparison unit 16 compares the analysis result images in chronological order while changing the combination of the first time point and the second time point.

[0065] [Machine Learning] As described above, the information processing system 10 can perform processing using various trained models. Each trained model may be generated by the information processing system 10. In this case, the information processing system 10 further includes a learning unit that generates the various trained models. Alternatively, each trained model may be generated by a computer system separate from the information processing system 10. A trained model generated by another computer system can be ported to the information processing system 10. The generation of a trained model corresponds to the learning phase of machine learning. The processing flow S1 executed using the generated trained model corresponds to the operation phase or estimation phase of machine learning.

[0066] In one example, various trained models used in the information processing system 10 are generated by supervised learning. In supervised learning, training data (sample data) including multiple data records indicating combinations of input data to be processed by a machine learning model and correct answers in output data from the machine learning model is used. A specific computer system performs the following processing for each data record of the training data. That is, the computer system inputs the input data indicated by the data record to the machine learning model. The computer system performs backpropagation (error backpropagation) based on the error between the output data estimated by the machine learning model and the correct answer indicated by the data record to update a set of parameters in the machine learning model. The computer system generates a trained model by repeating the processing for each data record until a predetermined termination condition is met. The termination condition may be that all data records of the training data have been processed. It should be noted that the generated trained model is a computational model estimated to be optimal, and is not necessarily a "computational model that is actually optimal."

[0067] Each of the one or more image analysis models is trained to identify pelvic organs and at least one lesion associated with the pelvic organs from an input image. As described above, when multiple image analysis models are generated and used, the multiple image analysis models are trained independently of each other. In the machine learning, training data including multiple data records representing pairs of sample images showing at least one or more pelvic organs and annotations corresponding to the sample images is used. The sample images may show at least one lesion in addition to one or more pelvic organs. The annotations indicate the correct identification of the one or more pelvic organs shown in the sample images. The annotations may further indicate the correct identification of at least one lesion shown in the sample images. In one example, the image analysis models are realized by a machine learning model that performs semantic segmentation, such as 3D U-Net or SwinUNETR. SwinUNETR is a Transformer-based machine learning model that performs self-supervised learning without annotation costs.

[0068] Each of the one or more adhesion determination models is trained to identify adhesions between pelvic organs from shape data, which is a set of parameter values ​​for the three-dimensional shapes of each of the multiple pelvic organs. When multiple adhesion determination models are generated and used, the multiple adhesion determination models are trained independently of each other. The adhesion determination model may be configured using a first determination model that determines the presence or absence of adhesions and a second determination model that determines the degree of adhesions. In machine learning, training data including multiple data records indicating pairs of shape data for two pelvic organs and adhesion data regarding adhesions between the two pelvic organs is used. The adhesion data indicates at least a correct answer regarding the presence or absence of adhesions, and this correct answer can be used in machine learning for the first determination model. The adhesion data may further indicate a correct answer regarding the degree of adhesions, and this correct answer can be used in machine learning for the second determination model. The adhesion determination model may be implemented, for example, by a decision tree such as LightGBM, a neural network, deep learning such as DenseNet, or multi-task learning.

[0069] Each of the one or more physical quantity estimation models is trained to estimate physical quantities related to the shape of a lesion from an image of the lesion. When multiple physical quantity estimation models are generated and used, the multiple physical quantity estimation models are trained independently of each other. In machine learning, training data including multiple data records representing pairs of sample images of the lesion and physical quantities related to the shape of the lesion is used. The physical quantity estimation models can be realized using techniques such as deep learning.

[0070] Each of the one or more severity estimation models is trained to estimate the severity of disease at the lesion from at least one of the shape physical quantity and the degree of adhesion. When multiple severity estimation models are generated and used, the multiple severity estimation models are trained independently of each other. In machine learning, training data including multiple data records indicating pairs of at least one of the shape physical quantity and the degree of adhesion and the severity of disease at the lesion is used. The severity estimation model can be realized by a technique such as deep learning.

[0071] [Modifications] The technology according to the present disclosure has been described in detail above based on various examples. However, the present disclosure is not limited to the above examples. The technology according to the present disclosure can be modified in various ways without departing from the spirit of the present disclosure.

[0072] In the above example, the information processing system 10 includes the adhesion determination unit 13, the lesion estimation unit 14, and the comparison unit 16, but the information processing system does not necessarily need to include at least one of these functional modules.

[0073] In the above example, the information processing system 10 is configured as a server in a client-server system. As another example, the information processing system (diagnosis support system) may be implemented in a stand-alone computer. Alternatively, the information processing system (diagnosis support system) may be implemented in a user terminal that can access a predetermined database via a communication network.

[0074] The processing procedure of the method executed by at least one processor is not limited to the above example. For example, some of the above-described steps or processes may be omitted, or the steps may be performed in a different order. For example, the adhesion determination (step S15) or the target lesion determination (step S16) may be performed first. Alternatively, the adhesion determination (step S15) or the series of processes related to the target lesion (steps S16 to S18) may be performed first. Furthermore, any two or more of the above-described steps may be combined, or some of the steps may be modified or deleted. Alternatively, other steps may be performed in addition to the above-described steps.

[0075] In the present disclosure, when comparing the magnitude of two numerical values, either of the two criteria "greater than or equal to" and "greater than" may be used, or either of the two criteria "less than or equal to" and "less than" may be used.

[0076] In the present disclosure, the expression "at least one processor executes a first process, executes a second process, ... executes an nth process" or an expression corresponding thereto indicates a concept including a case where the processor that executes n processes from the first process to the nth process changes midway. In other words, this expression indicates a concept including both a case where all n processes are executed by the same processor and a case where the processor changes among the n processes according to an arbitrary policy.

[0077] In the present disclosure, the term "to" indicating a range is an inclusive expression. For example, "A to B" means a range equal to or greater than A and equal to or less than B.

[0078] In this disclosure, the term "about" when used in conjunction with a numerical value means a range of plus and minus 10% of that numerical value.

[0079] The term "and / or" is used herein to refer to each of the objects listed before and after "and / or" or any combination thereof. For example, "A, B and / or C" includes each of the objects "A," "B," and "C," as well as the combinations "A and B," "A and C," "B and C," and "A and B and C."

[0080] [Supplementary Notes] As can be seen from the various examples above, the present disclosure includes the following aspects. (Supplementary Note 1) An information processing system including at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from input images, and generates one or more analysis result images showing processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is the lesion in the subject. (Supplementary Note 2) The information processing system described in Supplementary Note 1, wherein the image analysis model is trained to identify the at least one lesion among a plurality of lesions associated with the pelvic organs. (Supplementary Note 3) The information processing system described in Supplementary Note 1 or 2, wherein the at least one processor acquires the one or more target images showing at least one of a uterus, an ovary, a rectum, and a bladder as the one or more target pelvic organs. (Supplementary Note 4) The information processing system according to any one of Supplements 1 to 3, wherein the one or more target pelvic organs are a plurality of target pelvic organs, and the at least one processor processes the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs. (Supplementary Note 5) The information processing system according to Supplementary Note 4, wherein the at least one processor, as the processing of the one or more analysis result images, generates a set of parameter values ​​of three-dimensional shapes for each of the plurality of target pelvic organs as target shape data based on the one or more analysis result images, and determines whether or not adhesions exist between the target pelvic organs based on the target shape data. (Supplementary Note 6) The information processing system according to Supplementary Note 5, wherein the one or more target images are a plurality of target images, and the at least one processor inputs each of the plurality of target images into the image analysis model to generate a plurality of the analysis result images, and generates the target shape data based on the plurality of analysis result images.(Supplementary Note 7) The information processing system according to Supplementary Note 5 or 6, wherein the at least one processor inputs the target shape data to an adhesion determination model trained to identify adhesions between pelvic organs from shape data, which is a set of parameter values ​​of three-dimensional shapes for each of the plurality of pelvic organs, and determines whether or not adhesions exist between the target pelvic organs. (Supplementary Note 8) The information processing system according to Supplementary Note 7, wherein the adhesion determination model determines adhesions between a first organ and a second organ among the plurality of pelvic organs, and the at least one processor inputs the target shape data corresponding to each combination of the first organ and the second organ to each of a plurality of adhesion determination models, each combination of the first organ and the second organ being different from each other, and determines whether or not adhesions exist for each of a plurality of combinations of the first organ and the second organ. (Supplementary Note 9) The information processing system according to any one of Supplements 4 to 8, wherein the at least one processor processes the one or more analysis result images to further determine the degree of adhesion between the target pelvic organs. (Supplementary Note 10) The information processing system according to any one of Supplements 4 to 9, wherein the at least one processor estimates the severity of disease at the target lesion in response to determining that adhesions are present. (Supplementary Note 11) The information processing system according to any one of Supplementary Notes 4 to 10, wherein the at least one processor determines whether or not at least one of adhesions between the uterus and rectum, adhesions between the left ovary and rectum, adhesions between the right ovary and rectum, adhesions between the left ovary and uterus, adhesions between the right ovary and uterus, adhesions between the left ovary and right ovary, and adhesions between the uterus and bladder is present as the adhesions. (Supplementary Note 12) The information processing system according to Supplementary Note 11, wherein the at least one processor determines whether or not the adhesions are present as the adhesions, adhesions between the uterus and rectum, adhesions between the left ovary and rectum, adhesions between the right ovary and rectum, adhesions between the left ovary and uterus, adhesions between the right ovary and uterus, adhesions between the left ovary and right ovary, and adhesions between the uterus and bladder.(Supplementary Note 13) The information processing system according to any one of Supplements 4 to 12, wherein the image analysis model has been trained to identify at least one of endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on surfaces of the uterus other than the posterior surface of the uterus, and endometriotic cysts. (Supplementary Note 14) The information processing system according to any one of Supplements 1 to 13, wherein, in response to the target lesion being identified in at least one of the one or more target images by the image analysis model, the at least one processor estimates a shape physical quantity that is a physical quantity related to the shape of the target lesion based on the one or more analysis result images. (Supplementary Note 15) The information processing system according to Supplementary Note 14, wherein the at least one processor inputs the analysis result image to a physical quantity estimation model that has been trained to estimate a physical quantity related to the shape of the lesion from an image showing the lesion, thereby estimating the shape physical quantity of the target lesion. (Supplementary Note 16) The information processing system according to Supplementary Note 14 or 15, wherein the at least one processor estimates at least one of a dimension and a volume of the target lesion as the shape physical quantity. (Supplementary Note 17) The information processing system according to Supplementary Note 16, wherein the at least one processor estimates at least one of a major axis and a minor axis of the target lesion as the dimension. (Supplementary Note 18) The information processing system according to any one of Supplements 14 to 16, wherein the target lesion is a fibrous plaque, and the at least one processor estimates a thickness of the fibrous plaque as the shape physical quantity. (Supplementary Note 19) The information processing system according to any one of Supplements 14 to 16, wherein the target lesion is an ovarian cyst, and the at least one processor estimates at least one of a volume, a major axis, and a minor axis of the ovarian cyst as the shape physical quantity. (Supplementary Note 20) The information processing system according to any one of Supplementary Notes 14 to 19, wherein the at least one processor estimates a disease severity in the target lesion based on the estimated shape physical quantity.(Supplementary Note 21) The information processing system according to any one of Supplements 1 to 20, wherein the at least one processor, for each of the one or more target images, inputs the target image into each of a plurality of the image analysis models that have been trained independently of one another to generate a plurality of interim images, wherein each of the plurality of interim images indicates a processing result of the corresponding image analysis model in a format that distinguishes between each of the one or more target pelvic organs and the target lesion, and generates the analysis result image based on the plurality of interim images. (Supplementary Note 22) The information processing system according to Supplementary Note 21, wherein the at least one processor performs statistical processing on the plurality of interim images to generate the analysis result image. (Supplementary Note 23) The information processing system according to Supplementary Note 22, wherein the at least one processor calculates, for each of the one or more target images, uncertainty in at least a partial region of the analysis result image based on the plurality of interim images and the analysis result image. (Supplementary Note 24) The information processing system according to Supplementary Note 23, wherein the at least one processor, for each of the one or more target images, calculates, for at least a portion of the region, uncertainty for each of a plurality of voxels in the set of the plurality of provisional images as voxel uncertainty based on the plurality of provisional images and the analysis result image, and calculates a statistical value of the voxel uncertainty for each of the plurality of voxels as the uncertainty in the at least a portion of the region of the analysis result image. (Supplementary Note 25) The information processing system according to Supplementary Note 23 or 24, wherein the at least one processor calculates the uncertainty for a region surrounding the contour of the posterior surface of the uterus for each of the one or more target images. (Supplementary Note 26) The information processing system according to any one of Supplements 23 to 25, wherein the at least one processor outputs an alert in response to the calculated uncertainty exceeding a predetermined threshold. (Supplementary Note 27) The information processing system according to any one of Supplements 1 to 26, wherein the at least one processor acquires, as the one or more target images, one or more MRI images obtained by nuclear magnetic resonance imaging.(Supplementary Note 28) The information processing system according to any one of Supplements 1 to 27, wherein the at least one lesion includes at least one of inter-organ adhesions, endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on a surface of the uterus other than the posterior surface of the uterus, rectovaginal septum lesions, rectal lesions, bladder lesions, endometriotic cysts, fibrous plaques, ovarian cysts, and adenomyosis lesions. (Supplementary Note 29) The information processing system according to any one of Supplements 1 to 28, wherein the one or more target images include one or more first target images acquired at a first time point and one or more second target images acquired at a second time point different from the first time point, the one or more analysis result images include one or more first analysis result images generated by inputting each of the one or more first target images into the image analysis model and one or more second analysis result images generated by inputting each of the one or more second target images into the image analysis model, and the at least one processor generates a comparison result between the one or more first analysis result images and the one or more second analysis result images. (Supplementary Note 30) The information processing system according to any one of Supplements 1 to 29, wherein the at least one processor edits the analysis result image into a format that highlights the target lesion and does not highlight at least one of the one or more target pelvic organs, and displays the edited analysis result image on a display device. (Supplementary Note 31) The information processing system according to any one of Supplements 1 to 30, wherein the one or more analysis result images are a plurality of analysis result images corresponding to a plurality of the target images, and the at least one processor displays on a display device a screen including a user interface for switching the analysis result image to be displayed between the plurality of analysis result images.(Supplementary Note 32) An information processing system comprising at least one processor, wherein the at least one processor acquires one or more target images showing a plurality of target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the plurality of target pelvic organs from a target lesion that is the lesion in the subject, and processes the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs. (Supplementary Note 33) An information processing system comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject, and, in response to the target lesion being identified in at least one of the one or more target images by the image analysis model, estimates a shape physical quantity that is a physical quantity related to the shape of the target lesion based on the one or more analysis result images.(Supplementary Note 34) An information processing system comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject, and in generating the one or more analysis result images, the target image is input into a plurality of image analysis models trained independently of one another to generate a plurality of provisional images for each of the one or more target images, wherein each of the plurality of provisional images shows the processing result of the corresponding image analysis model in a format that distinguishes between each of the one or more target pelvic organs and the target lesion, and generates the analysis result image based on the plurality of provisional images. (Supplementary Note 35) A system comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing a processing result of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is the lesion in the subject, the one or more target images including one or more first target images acquired at a first time point and one or more second target images acquired at a second time point different from the first time point, the one or more analysis result images including one or more first analysis result images generated by inputting each of the one or more first target images into the image analysis model and one or more second analysis result images generated by inputting each of the one or more second target images into the image analysis model, The at least one processor compares the one or more first analysis result images with the one or more second analysis result images to generate a comparison image indicative of a result of the comparison.(Supplementary Note 36) An information processing method executed by an information processing system having at least one processor, comprising the steps of: acquiring one or more target images showing one or more target pelvic organs of a subject; inputting each of the one or more target images into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from input images, and generating one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject. (Supplementary Note 37) An information processing program that causes a computer to execute the steps of: acquiring one or more target images showing one or more target pelvic organs of a subject; inputting each of the one or more target images into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from input images, and generating one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject. (Supplementary Note 38) A diagnostic support system for assisting in diagnosis related to the uterus or ovaries, comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is the lesion in the subject.(Supplementary Note 39) A diagnostic support system for supporting the diagnosis of endometriosis, comprising at least one processor, wherein the at least one processor: acquires one or more target images showing a plurality of target pelvic organs of a subject, including a uterus or an ovary; inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images; generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the plurality of target pelvic organs from a target lesion that is the lesion in the subject; evaluates the pathological condition of endometriosis based on the one or more analysis result images; and the evaluation of the pathological condition of endometriosis includes: (i) processing the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs, and estimating and outputting the severity of disease at the target lesion in response to determining that adhesions exist; and generating one or more analysis result images showing a processing result of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and the target lesion, which is the lesion in the subject, in a format that distinguishes between each of the one or more target pelvic organs and the target lesion, which is the lesion in the subject, in an input image.(Supplementary Note 41) A diagnostic assistance program that causes a computer to function as a diagnostic assistance system for assisting in diagnosis related to the uterus or ovaries, the diagnostic assistance program causing the computer to execute the steps of: acquiring one or more target images showing target pelvic organs including the uterus or ovaries of a subject; and inputting each of the one or more target images into an image analysis model that has been trained to identify, from the input images, the pelvic organs including the uterus or ovaries and at least one lesion associated with the pelvic organs, and generating one or more analysis result images that show the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from the target lesion that is the lesion in the subject. (Supplementary Note 42) A pathology evaluation device for evaluating a pathology related to the uterus or ovaries, comprising at least one processor, wherein the at least one processor: acquires one or more target images showing one or more target pelvic organs of a subject including the uterus or ovaries; inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images; generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject; evaluates the pathology of endometriosis based on the one or more analysis result images; and the evaluation of the pathology of endometriosis includes: processing the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs; and estimating a shape physical quantity that is a physical quantity related to the shape of the target lesion based on the one or more analysis result images.(Supplementary Note 43) An information processing device comprising at least one processor and at least one memory storing computer program code, wherein the at least one memory and the computer program code, together with the at least one processor, acquire one or more target images showing one or more target pelvic organs of a subject, input each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generate one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is the lesion in the subject. (Supplementary Note 44) An information processing method executed by an information processing system having at least one processor, comprising: inputting one or more target images showing one or more target pelvic organs of a subject into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from an input image, and generating one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is the lesion in the subject; and processing the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs. (Supplementary Note 45) An information processing method executed by an information processing system having at least one processor, comprising: a step of inputting one or more target images showing one or more target pelvic organs of a subject into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from an input image, and generating one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject; and a step of estimating a shape physical quantity that is a physical quantity related to the shape of the target lesion based on the one or more analysis result images in response to the target lesion being identified in at least one of the one or more target images by the image analysis model.(Supplementary Note 46) An information processing method executed by an information processing system having at least one processor, comprising the steps of inputting one or more target images showing one or more target pelvic organs of a subject into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from an input image, and generating one or more analysis result images showing a processing result of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject, wherein the step of generating one or more analysis result images in response to the target lesion being identified in at least one of the one or more target images by the image analysis model comprises: for each of the one or more target images, inputting the target image into a plurality of image analysis models trained independently of each other to generate a plurality of provisional images, each of the plurality of provisional images showing a processing result of the corresponding image analysis model in a format that distinguishes between each of the one or more target pelvic organs and the target lesion; and generating the analysis result image for each of the one or more target images based on the plurality of provisional images. Information processing method. (Supplementary Note 47) An information processing method executed by an information processing system having at least one processor, comprising: inputting one or more target images showing one or more target pelvic organs of a subject into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from input images, and generating one or more analysis result images showing a processing result of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject; processing the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs; and estimating a shape physical quantity that is a physical quantity related to the shape of the target lesion, based on the one or more analysis result images, in response to the target lesion being identified in at least one of the one or more target images by the image analysis model.

[0081] According to Supplements 1, 36-38, and 43, by inputting a target image into an image analysis model obtained by learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Because the analysis result image distinguishes between the pelvic organs and lesions, accurate diagnosis of the pelvic organs is possible using this analysis result image. In one example, visualizing the predicted position of the pelvic organs makes it easier to determine the pathology of the pelvic organs, and visualizing the predicted position of the lesions prevents the physician who uses the system from overlooking the lesions and reduces variability in diagnosis. In image diagnosis, machine learning technology enables automatic detection of lesions present in the pelvic cavity, which contains mobile organs.

[0082] According to Supplementary Note 2, an analysis result image can be obtained that shows a specific lesion area separately from the pelvic organs and other lesion areas.

[0083] According to Supplementary Note 3, an analysis result image is obtained that clearly shows at least one of the uterus, ovaries, rectum, and bladder, allowing for accurate diagnosis of the specific pelvic organs.

[0084] According to Supplementary Note 4, information regarding adhesions between target pelvic organs is automatically obtained from the analysis result image. Adhesion can be an important factor for diagnosing pelvic organs. Therefore, by obtaining a determination regarding adhesions, pelvic organs can be diagnosed in more detail.

[0085] According to Supplementary Note 5, the determination of adhesions is performed based on the three-dimensional shape of the target pelvic organ, making it possible to perform the determination with greater accuracy.

[0086] According to Supplementary Note 6, by using multiple analysis result images, shape data of the target pelvic organ can be generated with higher accuracy.

[0087] According to Supplementary Note 7, adhesion determination is performed by inputting target shape data into an adhesion determination model obtained by learning. By using this adhesion determination model, adhesions, which may occur in various cases, can be determined with higher accuracy.

[0088] According to Supplementary Note 8, an adhesion determination model is prepared for each combination of two adjacent target internal organs, so that adhesion determination can be performed more accurately according to the combination.

[0089] According to Appendix 9, the degree of adhesion is also determined, so that more detailed information regarding adhesion can be provided.

[0090] According to Appendix 10, when adhesions are present, the severity of the disease at the target lesion site can be estimated, allowing for a more detailed diagnosis of the pelvic organs.

[0091] According to Appendix 11, information regarding at least one of adhesions between the uterus and the rectum, adhesions between the left ovary and the rectum, adhesions between the right ovary and the rectum, adhesions between the left ovary and the uterus, adhesions between the right ovary and the uterus, adhesions between the left ovary and the right ovary, and adhesions between the uterus and the bladder can be automatically obtained.

[0092] According to Appendix 12, information about major adhesions related to pelvic organs can be obtained automatically.

[0093] According to Supplementary Note 13, at least one of endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on surfaces of the uterus other than the posterior surface of the uterus, and endometriotic cysts is distinguished in the analysis result image. Furthermore, a determination regarding adhesions is made based on this analysis result image. As a result, an accurate diagnosis of the lesion can be made.

[0094] According to Supplementary Note 14, the shape physical quantity of the target lesion is estimated based on one or more analysis result images. The physical quantity related to the shape of the lesion can be an important factor for diagnosing the pelvic organs. Therefore, by obtaining the shape physical quantity, the pelvic organs can be diagnosed in more detail.

[0095] According to Supplementary Note 15, the shape physical quantities are estimated by inputting the analysis result image into a physical quantity estimation model obtained by learning. By using this physical quantity estimation model, the physical quantities related to the shape of the lesion can be estimated with higher accuracy.

[0096] According to Supplementary Note 16, the dimensions or volume, which are a type of physical quantity that directly indicates the state of the lesion, can be obtained, allowing for more detailed diagnosis of the pelvic organs.

[0097] According to Supplementary Note 17, the major axis or minor axis, which is a type of physical quantity that directly indicates the state of the lesion, can be obtained, allowing for more detailed diagnosis of the pelvic organs.

[0098] According to Supplementary Note 18, the thickness of the fibrous plaque is estimated, allowing for a more detailed diagnosis of this lesion.

[0099] According to Supplementary Note 19, at least one of the volume, major axis, and minor axis of an ovarian cyst is estimated, enabling a more detailed diagnosis of the lesion.

[0100] According to Supplementary Note 20, the severity of the disease at the target lesion site is estimated based on the shape physical quantity, allowing for more detailed diagnosis of the pelvic organs.

[0101] According to Supplementary Note 21, multiple image analysis models are used for each target image, and a final analysis result image is generated based on the results (provisional images) of each image analysis model. This mechanism allows for more accurate generation of analysis result images, thereby enabling more accurate diagnosis of pelvic organs.

[0102] According to Supplementary Note 22, a statistical result of a plurality of provisional images is obtained as an analysis result image, so that an analysis result image that is expected to be accurate can be generated.

[0103] According to Supplementary Note 23, uncertainty, which is information indicating how likely at least a part of the region of the analysis result image is, can be presented to a user such as an evaluator. The uncertainty can also be useful in diagnosing the pelvic organs.

[0104] According to Supplementary Note 24, a voxel uncertainty is calculated for each of a plurality of voxels in a set of a plurality of provisional images, and a statistical value of the voxel uncertainties is obtained as a final uncertainty. By such calculation, it is possible to accurately determine how likely at least a portion of the region of the analysis result image is.

[0105] According to Appendix 25, the uncertainty about the area around the contour of the posterior surface of the uterus can be presented to the user.

[0106] According to Supplementary Note 26, an alert can be sent to the user when the reliability of the analysis result image falls below a predetermined standard. This alert can be useful for interpreting the analysis result image.

[0107] According to Supplementary Note 27, an analysis result image can be generated from an MRI image, which is often used in diagnosing pelvic organs.

[0108] According to Supplementary Note 28, the analysis result image distinguishes and shows at least one of interorgan adhesions, endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on the surface of the uterus other than the posterior surface of the uterus, rectovaginal septum lesions, rectal lesions, bladder lesions, endometriotic cysts, fibrous plaques, ovarian cysts, and adenomyosis lesions. Use of this analysis result image enables accurate diagnosis of these lesions.

[0109] According to Appendix 29, a result is generated by comparing analysis result images at two different points in time, making it possible to accurately diagnose changes over time in pelvic organs or lesions.

[0110] According to Supplementary Note 30, the target lesion is displayed prominently, so that the location or condition of the lesion can be clearly presented to a user such as an evaluator.

[0111] According to Supplementary Note 31, a user interface for switching the analysis result image to be displayed is provided, so that a plurality of analysis result images can be easily handled on the display device.

[0112] According to Supplements 32 and 44, by inputting a target image into an image analysis model obtained by learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Since the analysis result image shows the pelvic organs and lesions separately from each other, accurate diagnosis of the pelvic organs is possible by using this analysis result image. In addition, information regarding adhesions between the target pelvic organs is automatically obtained from the analysis result image. Adhesion can be an important factor for diagnosing pelvic organs. Since a determination regarding adhesions is also obtained, a more detailed diagnosis of the pelvic organs can be made.

[0113] According to Supplements 33 and 45, by inputting a target image into an image analysis model obtained by learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Because the analysis result image distinguishes between the pelvic organs and the lesions, accurate diagnosis of the pelvic organs is possible by using this analysis result image. Furthermore, the shape physical quantities of the target lesion are estimated based on one or more analysis result images. The physical quantities related to the shape of the lesions can be important factors for diagnosing the pelvic organs. Therefore, by obtaining these shape physical quantities, the pelvic organs can be diagnosed in more detail.

[0114] According to Supplements 34 and 46, by inputting a target image into an image analysis model obtained by learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Because the analysis result image distinguishes between the pelvic organs and lesions, accurate diagnosis of the pelvic organs is possible using this analysis result image. Furthermore, multiple image analysis models are used for each target image, and a final analysis result image is generated based on the results (provisional images) of each image analysis model. This mechanism allows for more accurate generation of analysis result images. Therefore, more accurate diagnosis of the pelvic organs is possible.

[0115] According to Supplementary Note 35, by inputting a target image into an image analysis model obtained by learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Since the analysis result image shows the pelvic organs and lesions separately from each other, accurate diagnosis of the pelvic organs is possible by using this analysis result image. Furthermore, a comparison image showing the results of comparing the analysis result images at two different points in time is generated, making it possible to accurately diagnose changes in the pelvic organs or lesions over time.

[0116] According to Supplements 39 to 41, by inputting a target image into an image analysis model obtained through learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Because the analysis result image distinguishes between the pelvic organs and lesions, accurate diagnosis of the pelvic organs is possible using this analysis result image. In one example, visualizing the predicted position of the pelvic organs facilitates the assessment of pathological conditions in the pelvic organs, and visualizing the predicted position of the lesions prevents the physician using the system from overlooking lesions and reduces variability in assessment. For example, in image diagnosis, machine learning technology can automatically detect lesions related to uterine or ovarian conditions, such as ovarian cysts and deep-seated endometriotic lesions (e.g., plaques and adhesions) present in the pelvic cavity, which contains mobile organs. This reduces the burden on the physician using the system, prevents variability and oversights in assessment, and enables rapid evaluation, improving diagnostic efficiency and paving the way for optimal treatment and surgical planning for uterine or ovarian-related diseases, such as endometriosis.

[0117] According to Supplementary Note 42, by inputting a target image into an image analysis model obtained by learning, the pelvic organs and lesions of the subject are identified, and an analysis result image showing the processing results is generated. Since the analysis result image distinguishes between the pelvic organs and the lesions, accurate diagnosis of the pelvic organs is possible by using this analysis result image. Furthermore, information regarding adhesions between the target pelvic organs is automatically obtained from the analysis result image. Adhesion can be an important factor for diagnosing the pelvic organs. Since a determination regarding adhesions can also be obtained, the pelvic organs can be diagnosed in more detail. Furthermore, the shape physical quantities of the target lesion are estimated based on one or more analysis result images. The physical quantities related to the shape of the lesion can be an important factor for diagnosing the pelvic organs. Therefore, by obtaining these shape physical quantities, the pelvic organs can be diagnosed in more detail.

[0118] According to Supplementary Note 47, by inputting a target image into an image analysis model obtained by learning, the subject's pelvic organs and lesions are identified, and an analysis result image showing the processing results is generated. Since the analysis result image shows the pelvic organs and lesions separately from each other, using this analysis result image enables accurate evaluation of the pathology of the pelvic organs. Furthermore, information on adhesions between the target pelvic organs and the shape and physical quantities of the target lesions can be automatically obtained from the analysis result image, allowing for more detailed evaluation of the pathology of the pelvic organs.

[0119] 10...information processing system, 11...image acquisition unit, 12...image analysis unit, 13...adhesion determination unit, 14...lesion estimation unit, 15...output unit, 16...comparison unit, 20...database, 30...user terminal, 110...information processing program, 121-123...image analysis model, 131-137...adhesion determination model, 200...target image, 210...provisional image, 220...analysis result image, 230...reference image, 300...screen, 310...analysis result image, 320...slide bar, 330...determination result regarding adhesion.

Claims

1. An information processing system comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from the target lesion that is the lesion in the subject.

2. The information processing system of claim 1, wherein the image analysis model is trained to identify the at least one lesion among a plurality of lesions associated with the pelvic organs.

3. The information processing system of claim 1 or 2, wherein the at least one processor acquires the one or more target images showing at least one of a uterus, an ovary, a rectum, and a bladder as the one or more target pelvic organs.

4. An information processing system according to any one of claims 1 to 3, wherein the one or more target pelvic organs are a plurality of target pelvic organs, and the at least one processor processes the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs.

5. The information processing system of claim 4, wherein the at least one processor, as part of the processing of the one or more analysis result images, generates a set of parameter values ​​of the three-dimensional shape of each of the plurality of target pelvic organs as target shape data based on the one or more analysis result images, and determines whether or not adhesions exist between the target pelvic organs based on the target shape data.

6. The information processing system of claim 5, wherein the one or more target images are a plurality of target images, and the at least one processor inputs each of the plurality of target images into the image analysis model to generate a plurality of analysis result images, and generates the target shape data based on the plurality of analysis result images.

7. An information processing system as described in claim 5 or 6, wherein the at least one processor inputs the target shape data into an adhesion determination model trained to identify adhesions between pelvic organs from shape data, which is a collection of parameter values ​​of the three-dimensional shapes of each of the plurality of pelvic organs, and determines whether or not adhesions exist between the target pelvic organs.

8. The information processing system of claim 7, wherein the adhesion determination model determines adhesion between a first organ and a second organ among the plurality of pelvic organs, and the at least one processor inputs the target shape data corresponding to the combination of the first organ and the second organ into each of a plurality of adhesion determination models each having a different combination of the first organ and the second organ, and determines whether or not adhesion exists for each of a plurality of combinations of the first organ and the second organ.

9. The information processing system according to any one of claims 4 to 8, wherein the at least one processor processes the one or more analysis result images to further determine the degree of adhesion between the target pelvic organs.

10. The information processing system according to any one of claims 4 to 9, wherein the at least one processor estimates the severity of disease at the target lesion in response to determining that adhesions are present.

11. An information processing system according to any one of claims 4 to 10, wherein the image analysis model is trained to identify at least one of endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on surfaces of the uterus other than the posterior surface of the uterus, and endometriotic cysts.

12. An information processing system according to any one of claims 1 to 11, wherein the at least one processor, in response to the target lesion being identified in at least one of the one or more target images by the image analysis model, estimates a shape physical quantity, which is a physical quantity related to the shape of the target lesion, based on the one or more analysis result images.

13. The information processing system according to claim 12, wherein the at least one processor inputs the analysis result image into a physical quantity estimation model that has been trained to estimate physical quantities related to the shape of the lesion from an image showing the lesion, thereby estimating the shape physical quantities of the target lesion.

14. The information processing system according to claim 12 or 13, wherein the at least one processor estimates at least one of the dimensions and volume of the target lesion as the geometric physical quantity.

15. The information processing system according to claim 14, wherein the at least one processor estimates at least one of the major axis and the minor axis of the target lesion as the dimension.

16. The information processing system according to any one of claims 12 to 15, wherein the at least one processor estimates the severity of a disease in the target lesion based on the estimated shape physical quantity.

17. An information processing system according to any one of claims 1 to 16, wherein the at least one processor, for each of the one or more target images, inputs the target image into each of a plurality of image analysis models trained independently of one another to generate a plurality of provisional images, wherein each of the plurality of provisional images represents the processing result of the corresponding image analysis model in a format that distinguishes each of the one or more target pelvic organs from the target lesion, and generates the analysis result image based on the plurality of provisional images.

18. The information processing system according to claim 17, wherein the at least one processor performs statistical processing on the plurality of provisional images to generate the analysis result image.

19. The information processing system of claim 18, wherein the at least one processor calculates, for each of the one or more target images, uncertainty in at least some regions of the analysis result image based on the plurality of interim images and the analysis result image.

20. The information processing system of claim 19, wherein the at least one processor, for each of the one or more target images, calculates, based on the plurality of provisional images and the analysis result image, the uncertainty for each of the plurality of voxels in the set of the plurality of provisional images for at least a portion of the region as voxel uncertainty, and calculates a statistical value of the voxel uncertainty for each of the plurality of voxels as the uncertainty in the at least a portion of the region of the analysis result image.

21. The information processing system according to claim 19 or 20, wherein the at least one processor calculates the uncertainty for a region around the contour of the posterior surface of the uterus for each of the one or more target images.

22. The information processing system according to any one of claims 1 to 21, wherein the at least one processor acquires one or more MRI images obtained by nuclear magnetic resonance imaging as the one or more target images.

23. The information processing system of any one of claims 1 to 22, wherein the at least one lesion includes at least one of interorgan adhesions, endometriotic nodules on the posterior surface of the uterus, endometriotic nodules on the surface of the uterus other than the posterior surface of the uterus, rectovaginal septum lesions, rectal lesions, bladder lesions, endometriotic cysts, fibrous plaques, ovarian cysts, and adenomyosis lesions.

24. An information processing system according to any one of claims 1 to 23, wherein the one or more target images include one or more first target images acquired at a first time point and one or more second target images acquired at a second time point different from the first time point, the one or more analysis result images include one or more first analysis result images generated by inputting each of the one or more first target images into the image analysis model and one or more second analysis result images generated by inputting each of the one or more second target images into the image analysis model, and the at least one processor generates a comparison result between the one or more first analysis result images and the one or more second analysis result images.

25. An information processing system according to any one of claims 1 to 24, wherein the at least one processor edits the analysis result image into a format that highlights the target lesion and does not highlight at least one of the one or more target pelvic organs, and displays the edited analysis result image on a display device.

26. An information processing system comprising at least one processor, wherein the at least one processor: acquires one or more target images showing a plurality of target pelvic organs of a subject; inputs each of the one or more target images into an image analysis model trained to identify the pelvic organs and at least one lesion associated with the pelvic organs from the input images; generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the plurality of target pelvic organs from a target lesion that is the lesion in the subject; and processes the one or more analysis result images to determine whether or not adhesions exist between the target pelvic organs.

27. An information processing system comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject, and, in response to the target lesion being identified in at least one of the one or more target images by the image analysis model, estimates a shape physical quantity that is a physical quantity related to the shape of the target lesion based on the one or more analysis result images.

28. An information processing system comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes between each of the one or more target pelvic organs and a target lesion that is the lesion in the subject, and in generating the one or more analysis result images, the target image is input into a plurality of image analysis models trained independently of one another to generate a plurality of provisional images for each of the one or more target images, wherein each of the plurality of provisional images shows the processing result of the corresponding image analysis model in a format that distinguishes between each of the one or more target pelvic organs and the target lesion, and generates the analysis result image based on the plurality of provisional images.

29. A system comprising at least one processor, wherein the at least one processor acquires one or more target images showing one or more target pelvic organs of a subject, inputs each of the one or more target images into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generates one or more analysis result images showing a processing result of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from a target lesion that is the lesion in the subject, the one or more target images including one or more first target images acquired at a first time point and one or more second target images acquired at a second time point different from the first time point, the one or more analysis result images including one or more first analysis result images generated by inputting each of the one or more first target images into the image analysis model and one or more second analysis result images generated by inputting each of the one or more second target images into the image analysis model, and the at least one processor compares the one or more first analysis result images with the one or more second analysis result images to generate a comparison image showing a result of the comparison. Information processing system.

30. An information processing method executed by an information processing system having at least one processor, comprising: steps of acquiring one or more target images showing one or more target pelvic organs of a subject; and inputting each of the one or more target images into an image analysis model trained to identify pelvic organs and at least one lesion associated with the pelvic organs from the input images, and generating one or more analysis result images showing the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from the target lesion, which is the lesion in the subject.

31. An information processing program that causes a computer to execute the following steps: acquiring one or more target images showing one or more target pelvic organs of a subject; inputting each of the one or more target images into an image analysis model that has been trained to identify pelvic organs and at least one lesion associated with the pelvic organs from the input images; and generating one or more analysis result images that show the processing results of the image analysis model in a format that distinguishes each of the one or more target pelvic organs from the target lesion, which is the lesion in the subject.

Images