Method for estimating wall thickness, computer program, learning method, model making method, wall thickness estimation device, and wall thickness estimation system.

Abstract

This wall thickness estimation method includes: a first acquisition step (S201) for acquiring behavior information, which is numerical information pertaining to temporal changes of an organ wall or a vascular wall at multiple predetermined points therein, on the basis of moving images in which the organ wall or the vascular wall were captured and which are obtained using four-dimensional angiography; a generation step (S202) for generating estimation information in which the thicknesses of the organ wall or the vascular wall at multiple predetermined points therein are visualized, the visualization being carried out by a trained model that takes, as an input, images representing physical parameters which are based on the behavior information acquired in the acquisition step (S201), and that produces, as an output, an index representing the thicknesses; and an outputting step (S203) for outputting the estimation information generated in the generation step (S202).

Description

[[TECHNICAL FIELD]]

[0001] The present invention relates to a wall thickness estimation method for estimating the thickness of an organ wall or a blood vessel wall, and the like. [[BACKGROUND ART]]

[0002] Cerebral aneurysm, which is an example of vascular diseases, is an extremely high-risk disease with a fatality rate exceeding about 50% once it ruptures, and is also a socially influential disease that leaves sequelae at a high rate. For this reason, preventive treatment (preemptive medicine) to prevent rupture of a cerebral aneurysm is highly important, and appropriate therapeutic intervention is essential.

[0003] For appropriate treatment, it is effective to know information (e.g., thickness) of the aneurysm wall of a cerebral aneurysm. This is because it is known that rupture of a cerebral aneurysm is more likely to occur in a thin portion of the aneurysm wall than in a thick portion thereof. However, even in a single cerebral aneurysm, the shape such as the thickness of the aneurysm wall varies from one cerebral aneurysm to another.

[0004] Therefore, it is difficult even for an expert to infer information about the shape such as the thickness of the aneurysm wall only from the morphology such as the lumen of the aneurysm wall obtained by CT (Computed Tomography), MRI (Magnetic Resonance Imaging) and MRA (Magnetic Resonance Angiography).

[0005] For example, as a method for predicting the thickness of the aneurysm wall of a cerebral aneurysm, imaging or visual observation by craniotomy performed by a physician is known. However, this method is a highly invasive method that imposes a heavy burden on a patient, and is not a method that can easily predict the thickness of the aneurysm wall of a cerebral aneurysm.

[0006] Furthermore, for example, an ultrasound diagnostic device disclosed in Patent Document 1 is known as a minimally invasive method for predicting the thickness of blood vessel walls, such as the aneurysm walls of cerebral aneurysms. Patent Document 1 discloses an ultrasound diagnostic device that generates image data using ultrasound signals and displays information regarding the thickness of the blood vessel walls of a subject based on said image data. [Prior art documents] [Patent Documents]

[0007] [Patent Document 1] Japanese Patent Publication No. 2013-118932 [Overview of the Initiative] [Problems that the invention aims to solve]

[0008] However, the image data obtained by the prior art disclosed in Patent Document 1 is of low precision, making it difficult to obtain highly accurate information regarding blood vessel walls. Furthermore, the prior art does not only fail to provide highly accurate information regarding blood vessel walls, but also regarding the organ walls of other organs in the human body, making it difficult to propose information for specific treatments for organ and blood vessel diseases.

[0009] Therefore, the present invention aims to provide a wall thickness estimation method and the like that can generate highly accurate information about organ walls or blood vessel walls using a minimally invasive method, thereby enabling the proposal of useful information for implementing specific treatments for organ or blood vessel diseases. [Means for solving the problem]

[0010] A wall thickness estimation method according to one aspect of the present invention includes: a first acquisition step of acquiring behavior information, which is numerical information relating to the time change of each of the positions of a plurality of predetermined points on an organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography; a generation step of generating estimation information, which is information in which the thickness is visualized, using a trained model that takes an image showing physical parameters based on the behavior information acquired in the first acquisition step as input and outputs an index showing the thickness of each of the plurality of predetermined points on the organ wall or blood vessel wall; and an output step of outputting the estimation information generated in the generation step.

[0011] Furthermore, a computer program according to one aspect of the present invention causes a computer to execute the wall thickness estimation method described above.

[0012] Furthermore, a learning method according to one aspect of the present invention includes a second acquisition step of acquiring behavior information, which is numerical information relating to the time change of the position of each of a plurality of predetermined points on an organ wall or a blood vessel wall, based on a moving image including the organ wall or the blood vessel wall, and a second learning step of training a model using one or more datasets as training data, which are composed of the behavior information acquired in the second acquisition step, and which consist of an image showing physical parameters based on the behavior information of one of the plurality of predetermined points and an index showing the thickness of one of the plurality of predetermined points.

[0013] Furthermore, a model manufacturing method according to one aspect of the present invention includes a third acquisition step of acquiring the estimated information generated by the generation step described above, and a first manufacturing step of manufacturing a vascular model including the vascular walls described above, wherein the vascular model is manufactured such that the vascular walls included in the vascular model exhibit different characteristics for each thickness, based on the thickness visualized by the estimated information acquired by the third acquisition step.

[0014] Furthermore, a wall thickness estimation device according to one aspect of the present invention includes: an acquisition unit that acquires behavioral information, which is numerical information relating to the time change of each of the positions of a plurality of predetermined points on the organ wall or the blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography; a generation unit that takes an image showing physical parameters based on the behavioral information acquired by the acquisition unit as input and generates estimation information, which is information in which the thickness is visualized, using a trained model that outputs an index showing the thickness of each of the plurality of predetermined points on the organ wall or the blood vessel wall; and an output unit that outputs the estimation information generated by the generation unit.

[0015] Furthermore, a wall thickness estimation system according to one aspect of the present invention comprises the wall thickness estimation device described above, a video information processing device that acquires the video image, generates the behavior information and outputs it to the acquisition unit, and a display device that displays the estimation information output by the output unit. [Effects of the Invention]

[0016] According to the wall thickness estimation method of the present invention, highly accurate information regarding organ walls or blood vessel walls can be generated using a minimally invasive method, thereby providing useful information for implementing specific treatments for organ or blood vessel diseases. [Brief explanation of the drawing]

[0017] [Figure 1] Figure 1 shows the configuration of a wall thickness estimation system according to an embodiment. [Figure 2] Figure 2 is a block diagram showing the characteristic functional configuration of the wall thickness estimation device according to the embodiment. [Figure 3] Figure 3 is a perspective view showing a cerebral aneurysm according to an embodiment. [Figure 4] Figure 4 is a cross-sectional view of the cerebral aneurysm according to this embodiment, shown along line IV-IV in Figure 3. [Figure 5] Figure 5 is a cross-sectional view of the cerebral aneurysm according to this embodiment, shown in line VV of Figure 4. [Figure 6]FIG. 6 is a flowchart showing a processing procedure in which the wall thickness estimation apparatus according to the embodiment trains a machine learning model. [Figure 7] FIG. 7 is an explanatory diagram showing teacher data according to the embodiment. [Figure 8] FIG. 8 is a flowchart showing a processing procedure in which the wall thickness estimation apparatus according to the embodiment estimates the thickness of an aneurysm wall of a cerebral aneurysm. [Figure 9] FIG. 9 is a diagram showing an example of estimation information according to the embodiment. [Figure 10A] FIG. 10A is a diagram showing a still image of a cerebral aneurysm according to the embodiment. [Figure 10B] FIG. 10B is a block diagram showing a characteristic functional configuration of a model production system according to Modification 1. [Figure 10C] FIG. 10C is a flowchart showing a processing procedure in which the model production system according to Modification 1 produces a blood vessel model. [Figure 10D] FIG. 10D is a schematic diagram showing an example of estimation information according to Modification 1. [Figure 10E] FIG. 10E is a blood vessel model including a blood vessel wall (aneurysm wall) according to Modification 1. [Figure 10F] FIG. 10F is an overall cerebral blood vessel model according to Modification 1. [Figure 10G] FIG. 10G is a cerebral model according to Modification 1. [Figure 10H] FIG. 10H is a skull model according to Modification 1. [Figure 10I] FIG. 10I is a blood vessel model including a blood vessel wall (aneurysm wall) of a subject other than the target subject. [Figure 11] FIG. 11 is a block diagram showing a characteristic functional configuration of a wall thickness estimation system according to Modification 2. [Figure 12] FIG. 12 is a flowchart showing a processing procedure in which a learning apparatus according to Modification 2 trains a machine learning model. [Figure 13] FIG. 13 shows one still image (one frame) included in a two-dimensional moving image according to Modification 2 and an image showing a depth estimated for said one still image. [Modes for carrying out the invention]

[0018] The embodiments will be described below with reference to the drawings. The embodiments described below are all comprehensive or specific examples. The numerical values, shapes, materials, components, arrangement positions and connection configurations of components, processes, and process sequences shown in the following embodiments are examples only and are not intended to limit the present invention. Furthermore, components in the following embodiments that are not described in an independent claim will be described as optional components.

[0019] Please note that each figure is a schematic diagram and not necessarily a strictly accurate representation. Furthermore, the same reference numerals are used for substantially identical components in each figure, and redundant explanations may be omitted or simplified.

[0020] (Embodiment) [Configuration of the wall thickness estimation system] The configuration of the wall thickness estimation system 1000 according to this embodiment will be described. Figure 1 is a diagram showing the configuration of the wall thickness estimation system 1000 according to this embodiment.

[0021] The wall thickness estimation system 1000 uses four-dimensional angiography to acquire behavioral information, which is numerical information about the time change of each predetermined point's position, from a moving image including the organ wall or blood vessel wall of subject P. Furthermore, the wall thickness estimation system 1000 is a system that generates estimation information for estimating the thickness of the organ wall or blood vessel wall based on the acquired behavioral information. For example, the wall thickness estimation system 1000 estimates the thickness of a cerebral aneurysm, which is an example of a blood vessel wall of subject P.

[0022] Four-dimensional angiography is a technique that adds a time axis to three-dimensional angiography. Three-dimensional angiography is a technique that collects three-dimensional data of blood vessels using X-ray CT and MRI scanners and extracts vascular information. Four-dimensional angiography using an X-ray CT scanner is also called 4DCTA (4-Dimensional Computed Tomography Angiography).

[0023] A moving image is obtained using four-dimensional angiography. This moving image may be a time series of three or more still images, for example, a moving image taken over the time it takes for the heart to pulsate n times (where n is a natural number). Alternatively, for example, this moving image may be a moving image taken within a predetermined time period. This predetermined time period may be, for example, m seconds (where m is a natural number).

[0024] Here, "organ wall" refers to the wall possessed by an organ, and "organ" includes both thoracic and abdominal organs. For example, thoracic organs include the heart and lungs, and abdominal organs include the stomach, intestines, liver, kidneys, and pancreas, but are not limited to these. Furthermore, "organ" may also include thoracic organs with internal cavities and abdominal organs with internal cavities.

[0025] An organ wall is, for example, a wall that separates an organ from another organ or tissue. For example, if the organ is the heart, the organ wall is the wall made up of muscle (myocardium) that separates the heart from other tissues. An organ wall is also, for example, a wall that separates a region within an organ. For example, if the organ is the heart, the organ wall is the interventricular wall that separates the left ventricle and the right ventricle, which is an example of a region within the heart.

[0026] Furthermore, the blood vessel wall may be the wall of a blood vessel including an artery or vein, or it may be the wall of an aneurysm or varicose vein, for example, the wall of a cerebral aneurysm, an aortic aneurysm, or a visceral aneurysm.

[0027] As shown in Figure 1, the wall thickness estimation system 1000 comprises a wall thickness estimation device 100, a display device 200, a video information processing device 300, and a video image capturing device 400.

[0028] The motion image acquisition device 400 is a device that generates motion images including organ walls or blood vessel walls using four-dimensional angiography. The motion image acquisition device 400 is, for example, an X-ray CT scanner or an MRI scanner. In this embodiment, the motion image acquisition device 400 is an X-ray CT scanner, and the motion image acquisition device 400 comprises an X-ray tube that emits X-rays, a detector that receives signals, and a computer.

[0029] The detector is located opposite the X-ray tube and detects the X-rays after they have passed through the body of subject P. At this time, the computer generates a moving image that includes organ walls or blood vessel walls at specific parts of subject P's body, taking advantage of the fact that X-ray absorption differs depending on the body part of subject P. The moving image acquisition device 400 also has the function of measuring and acquiring the electrocardiogram waveform of subject P.

[0030] The technique using an X-ray CT or MRI scanner and 4D angiography is a minimally invasive method because, unlike open abdominal surgery, open heart surgery, or open craniotomy, it does not require incisions that place a heavy burden on the patient's body. Furthermore, the technique using an X-ray CT or MRI scanner and 4D angiography can generate highly precise moving images.

[0031] The video information processing device 300 acquires video images including organ walls or blood vessel walls generated by the video image acquisition device 400 using four-dimensional angiography, and generates behavioral information, which is numerical information about the time change of the position of each of several predetermined points on the organ wall or blood vessel wall. In other words, the behavioral information is information based on video images including organ walls or blood vessel walls obtained using four-dimensional angiography.

[0032] Here, behavioral information refers to numerical information where a specific time in a video and the three-dimensional coordinate positions of multiple predetermined points on the organ wall or blood vessel wall at that specific time are treated as a single set, and multiple sets are arranged according to the time it takes for one heartbeat in the video. Note that these multiple predetermined points represent extremely small regions.

[0033] The video information processing device 300 outputs behavioral information to the wall thickness estimation device 100. The video information processing device 300 is, for example, a personal computer, but it may also be a high-performance server connected to a network.

[0034] The wall thickness estimation device 100 acquires behavioral information generated by the motion image information processing device 300, generates estimation information for estimating the thickness of organ walls or blood vessel walls based on the acquired behavioral information, and outputs the generated estimation information to the display device 200. The wall thickness estimation device 100 is, for example, a personal computer, but may also be a server device with high computing power connected to a network.

[0035] The display device 200 displays the estimated information output from the wall thickness estimation device 100. Specifically, the display device 200 is a monitor device composed of a liquid crystal panel or an organic EL (Electro-Luminescence) panel, etc. A television, smartphone, or tablet terminal may be used as the display device 200.

[0036] The wall thickness estimation device 100, the display device 200, and the video information processing device 300 only need to be capable of sending and receiving behavior information or estimation information, and may be connected by wire or wirelessly.

[0037] In this manner, the video information processing device 300 acquires video images including organ walls or blood vessel walls and generates behavioral information, which is numerical information relating to the time changes in the positions of multiple predetermined points on the organ wall or blood vessel wall.

[0038] The wall thickness estimation device 100 acquires behavioral information generated by the motion image information processing device 300, and generates estimation information for estimating the thickness of organ walls or blood vessel walls based on the acquired behavioral information. Furthermore, the wall thickness estimation device 100 outputs the generated estimation information to the display device 200.

[0039] Thus, the wall thickness estimation system 1000 uses a video information processing device 300 and a video image capture device 400 to obtain video images including organ walls or blood vessel walls using a minimally invasive method. Furthermore, the wall thickness estimation system 1000 can generate estimation information for estimating the thickness of organ walls or blood vessel walls by utilizing behavioral information related to the video images. Therefore, the wall thickness estimation system 1000 can generate highly accurate information about the wall thickness near each of several predetermined points in the organ wall or blood vessel wall.

[0040] Next, the functional configuration of the wall thickness estimation device 100 according to this embodiment will be described in detail.

[0041] Figure 2 is a block diagram showing the characteristic functional configuration of the wall thickness estimation device 100 according to this embodiment. The wall thickness estimation device 100 includes a first acquisition unit 110, a generation unit 120, an output unit 130, and a first learning unit 140 as acquisition units.

[0042] The first acquisition unit 110 acquires behavioral information, which is numerical information relating to the time change of the position of each of a plurality of predetermined points on the organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography. Specifically, the first acquisition unit 110 acquires behavioral information generated by the moving image information processing device 300. The first acquisition unit 110 is, for example, a communication interface that performs wired communication or wireless communication.

[0043] The generation unit 120 generates estimation information for estimating the thickness of an organ wall or blood vessel wall based on the behavioral information acquired by the first acquisition unit 110. More specifically, the generation unit 120 has a trained model (in this case, a machine learning model 121), and uses this model to generate estimation information, which is information in which the thickness of each of several predetermined points in the organ wall or blood vessel wall is visualized.

[0044] The trained model takes an image showing physical parameters based on behavioral information acquired by the first acquisition unit 110 as input, and outputs an index indicating the thickness of each of several predetermined points on an organ wall or blood vessel wall. In this embodiment, the first learning unit 140 trains the model. In the following, the above image for generating estimation information may be described as the first input image.

[0045] The estimated information is, for example, image information in which the thickness of each of several predetermined points is visualized. The method for generating the estimated information will be described later using Figure 8. Specifically, the generation unit 120 is realized by a processor that executes a program, a microcomputer, or a dedicated circuit.

[0046] The output unit 130 outputs the estimated information generated by the generation unit 120. The output unit 130 may also output the estimated information generated by the generation unit 120 to the display device 200. The output unit 130 is, for example, a communication interface for wired or wireless communication.

[0047] The first learning unit 140 trains the model using training data. Specifically, the first learning unit 140 is implemented by a processor, microcomputer, or dedicated circuit that executes a program.

[0048] The first learning unit 140 trains and constructs a model. The first learning unit 140 provides the constructed model to the generation unit 120. Note that the first learning unit 140 is not an essential component and does not need to be provided in the wall thickness estimation device 100.

[0049] This model is designed to generate estimated information.

[0050] In this embodiment, the model is a model constructed by machine learning using one or more datasets as training data. One dataset consists of a combination of an image showing physical parameters based on the behavior information of a predetermined point among a plurality of predetermined points in an organ wall or blood vessel wall, and an index showing the thickness of said predetermined point.

[0051] In other words, the model is a recognition model constructed by machine learning using one or more datasets as training data, each of which consists of one or more datasets containing images representing the physical parameters of a given point and an index representing the thickness of that point.

[0052] More specifically, the model is a recognition model constructed using machine learning, with input data being images representing physical parameters belonging to one or more datasets that serve as the training data, and output data being an index indicating the thickness of a given point belonging to that dataset.

[0053] The first learning unit 140 trains a model using machine learning as an example, as described above. Therefore, in this embodiment, the model is the machine learning model 121.

[0054] Furthermore, the first learning unit 140 may train a model using, for example, a neural network, more specifically, a convolutional neural network (CNN). If the model is a convolutional neural network model, the first learning unit 140 determines the coefficients (weights) of the convolutional layer filters, etc., by machine learning based on the training data.

[0055] Furthermore, the first learning unit 140 may train the model using algorithms other than neural networks.

[0056] In the following examples, the above image included in the training data may be referred to as the second input image.

[0057] Here, several predetermined points regarding behavioral information will be explained using Figures 3 to 5. In this embodiment, the explanation will be based on the blood vessel wall, but the same applies to organ walls. Furthermore, in this context, the blood vessel wall refers to the aneurysm wall 11 of the cerebral aneurysm 10.

[0058] In Figures 3 to 5, for example, the positive x-axis represents the direction in which the cerebral aneurysm 10 extends from the parent vessel 20, the z-axis represents the direction in which the parent vessel 20 extends, and the y-axis represents the direction perpendicular to the x-axis and z-axis. Furthermore, Figures 3 to 5, which show the parent vessel 20, the cerebral aneurysm 10, the aneurysm wall 11, and several predetermined points, are general schematic diagrams that can be used not only to describe the brain of subject P, but also to describe the brains of other subjects.

[0059] Figure 3 is a perspective view showing a cerebral aneurysm 10 according to this embodiment. Figure 4 is a cross-sectional view of the cerebral aneurysm 10 according to this embodiment along line IV-IV in Figure 3. The parent vessel 20 is an example of a blood vessel that makes up an artery in the brain. The cerebral aneurysm 10 is an aneurysm that is a bulge in a part of the parent vessel 20 and is an aneurysm that originates extending in the x-axis direction from the parent vessel 20.

[0060] Figure 5 is a cross-sectional view of the cerebral aneurysm 10 according to this embodiment, along the VV line in Figure 4.

[0061] As shown in Figure 5, in the cross-sectional view of the cerebral aneurysm 10, several predetermined points are provided in the direction from 12 o'clock to 11 o'clock, corresponding to the time indicated on the clock face. Point p0 is provided in the direction of 12 o'clock, and points p1 to p11 are provided in the directions from 1 o'clock to 11 o'clock, respectively. In other words, 12 predetermined points are provided on the outer circumference of the cerebral aneurysm 10 in the cross-sectional view of the cerebral aneurysm 10.

[0062] The number of predetermined points is not limited to this, and for example, 10 to 1000 predetermined points may be provided on the outer circumference of the cerebral aneurysm 10 in a single cross-sectional view. Also, although one cross-sectional view is used in this embodiment, it is not limited to this, and multiple cross-sectional views (for example, 10 to 1000 cross-sectional views) may be used.

[0063] Furthermore, for each of these multiple cross-sectional views, for example, 10 to 1,000 predetermined points may be provided on the outer circumference of the cerebral aneurysm 10. In this case, 30,000 to 300,000 predetermined points may be provided for one cerebral aneurysm 10.

[0064] Furthermore, the multiple predetermined points on the blood vessel wall are not limited to those described above, and two or more points can be selected from the blood vessel wall. The number of predetermined points is not limited to 30,000 to 300,000; a number smaller than 30,000 may be selected, and a number larger than 300,000 may also be selected.

[0065] As described above, the multiple predetermined points in the blood vessel wall (aneurysm wall 11) according to this embodiment are points p0 to p11. In other words, there are a total of 12 multiple predetermined points in the aneurysm wall 11.

[0066] The first acquisition unit 110 acquires behavioral information, which is numerical information relating to the time change of position, at each of the 12 predetermined points. Based on this behavioral information, the generation unit 120 generates estimation information for estimating the thickness of the nodule wall 11 near the predetermined points.

[0067] In this embodiment, the behavioral information is numerical information relating to the time change of position over a certain period of time. For example, a certain period of time is the time it takes for one heartbeat. Furthermore, the time it takes for one heartbeat is divided equally into, for example, 100 steps.

[0068] In this case, the time when the pulse begins is defined as step 0, and the time when the pulse ends is defined as step 100. Furthermore, the duration of one heartbeat is not limited to the above and can be set arbitrarily.

[0069] Therefore, the behavioral information includes information about the x, y, and z axis positions of 12 predetermined points at each step from step 0 to step 100. In other words, the behavioral information is data where, for each of the 12 predetermined points, the time and the coordinate position (x, y, and z axis position) at that time are combined into one data set. To put it another way, the behavioral information includes data that evolves over time.

[0070] The specified time period may be a specific number of seconds, for example, 1 second, 5 seconds, or 10 seconds. Furthermore, this specified time period may be subdivided in any way, as long as it is divided into three or more parts. For example, contrary to the above, this specified time period may be divided into different numbers of steps, rather than 100 steps. Moreover, this specified time period does not have to be divided equally.

[0071] The time it takes for one heartbeat may be evenly divided by any number of steps selected from, for example, 10 to 1,000,000 steps. The number of steps is not limited to 10 to 1,000,000 steps; a number smaller than 10 steps may be selected, and a number larger than 1,000,000 steps may also be selected.

[0072] [Processing procedure for estimating wall thickness] Next, we will explain the specific processing steps in the wall thickness estimation method performed by the wall thickness estimation device 100. Again, we will use a blood vessel wall as an example, but the same applies to organ walls.

[0073] Figure 6 is a flowchart showing the processing procedure for training the model (machine learning model 121) in the wall thickness estimation device 100 according to this embodiment. Note that the processing procedure shown in the flowchart in Figure 6 is performed before the estimation of the thickness of the aneurysm wall 11 of the cerebral aneurysm 10 of subject P.

[0074] In this case, the first learning unit 140 trains the machine learning model 121 using training data obtained from one or more other subjects besides subject P. Therefore, for one or more other subjects, the video imaging device 400 generates video images including the blood vessel walls using four-dimensional angiography, and the wall thickness estimation device 100 acquires behavioral information based on these video images. For simplicity, this explanation will mainly use one other subject B as an example.

[0075] First, the first learning unit 140 acquires training data for the model (machine learning model 121) (step S101). While the training data may be generated by the generation unit 120, it is not limited to this and may be generated by other processing units or devices.

[0076] The training data will be explained below.

[0077] Figure 7 is an explanatory diagram showing the training data according to this embodiment. The training data consists of one or more datasets. Figure 7 shows one dataset (in this case, dataset D1) containing the training data as an example of training data.

[0078] A single dataset consists of a combination of an image showing physical parameters based on the behavioral information of a predetermined point (i.e., a second input image) and an index indicating the thickness of that predetermined point (i.e., a thickness index). As an example, dataset D1 shown in Figure 7 consists of a second input image I1 and a thickness index T1.

[0079] In other words, a dataset is a set of data consisting of a second input image and a thickness index for a given point. The training data consists of one or more datasets, for example, a dataset of 100 to 1,000,000 points is preferable, a dataset of 1,000 to 1,000,000 points is preferable, and a dataset of 10,000 to 1,000,000 points is even preferable. The more datasets that make up the training data, the better.

[0080] For example, the predetermined point corresponding to one dataset is point p0 shown in Figures 3-5 for another subject B. Similarly, the predetermined point corresponding to another dataset is point p1 shown in Figures 3-5 for another subject B.

[0081] Ideally, one other subject B should yield a dataset of, for example, between 100 and 150,000 data points. Furthermore, if there are multiple other subjects B, it is desirable to obtain a dataset of between 100 and 150,000 data points from each of them. Note that the number of datasets obtained from one other subject B may be less than 100 or more than 150,000.

[0082] Here, we will describe the second input image, which is an image showing physical parameters based on the behavior information of the predetermined point.

[0083] The physical parameters may, for example, be parameters relating to the time evolution of the displacement of each of several predetermined points. In other words, the physical parameters may be values ​​calculated from the time evolution of the displacement of each of several predetermined points.

[0084] More specifically, the physical parameters include the time evolution of displacement, velocity, acceleration, and strain at multiple predetermined points. Displacement is the change in position at each step, with the position at step 0 (the time when pulsation begins) being set as 0 (the origin).

[0085] Distortion is calculated using time and position data contained in the behavioral information. The method for calculating distortion is not particularly limited and known methods can be used. For example, the distortion of one point (e.g., point p1) and another point adjacent to it (e.g., point p2) may be calculated from the change in position of the two points at a certain time (e.g., step 10) and the next time (step 11).

[0086] Furthermore, we will explain the second input image, which is an image showing physical parameters.

[0087] The second input image is often a two-dimensional image composed of graphs showing physical parameters for one predetermined point out of a plurality of predetermined points, as follows: The second input image is composed of multiple graphs, and in the second input image, the multiple graphs are arranged in a k × l matrix (where k and l are natural numbers). The multiple graphs may be arranged in one column or in two columns. Here, the second input image is composed of nine graphs, and in the image, the multiple graphs are arranged in a 3 × 3 matrix as an example.

[0088] The nine graphs are as follows:

[0089] The three graphs in the first column represent the physical parameters in the x-axis direction of a given point. The three graphs in the second column represent the physical parameters in the y-axis direction of a given point. The three graphs in the third column represent the physical parameters in the z-axis direction of a given point.

[0090] The three graphs in the first row have the horizontal axis representing displacement in the x, y, and z directions, and the vertical axis representing acceleration in the x, y, and z directions. The three graphs in the second row have the horizontal axis representing displacement in the x, y, and z directions, and the vertical axis representing velocity in the x, y, and z directions. The three graphs in the third row have the horizontal axis representing velocity in the x, y, and z directions, and the vertical axis representing acceleration in the x, y, and z directions.

[0091] Here, displacement, velocity, and acceleration are used for the horizontal and vertical axes of the graph, but this is not limited to them. As described above, one of the parameters relating to the time change of displacement at each of several predetermined points, such as displacement, velocity, acceleration, and strain, may be used for the horizontal axis and another for the vertical axis.

[0092] Next, we will explain the thickness index, which is an indicator that shows the thickness of a particular point among several predetermined points. In Figure 7, the thickness index T1 is shown as the numerical value "1".

[0093] In this embodiment, the thickness index of the predetermined point is an index based on the color tone of the cerebral aneurysm 10 in the brain shown in the moving image (more specifically, the still image included in the moving image) obtained by the moving image acquisition device 400 and the surgical image obtained by craniotomy. This thickness index will be described in more detail below.

[0094] As described above, in this embodiment, behavioral information about other subjects B is also acquired. Similarly, the wall thickness estimation device 100 acquires moving images of other subjects B, which include the blood vessel wall generated by the X-ray CT device, the moving image acquisition device 400. Here, one or more still images based on the moving images acquired by the wall thickness estimation device 100 are used. Each of the one or more still images is an image cropped from the moving image, for example, one frame of the moving image.

[0095] Since the motion imaging device 400 is an X-ray CT scanner, each of the one or more still images is each one or more CT images. Each of the one or more CT images does not contain information indicating the color tone of the cerebral aneurysm 10 in the captured brain, and is shown in shades of black and white, in other words, in achromatic colors. Furthermore, each of the one or more CT images contains information indicating the location of several predetermined points within that CT image.

[0096] Furthermore, as described above, the motion image acquisition device 400 may be an MRI device. In this case, each of the one or more still images is each one or more MRI images. Each of the one or more MRI images does not contain information indicating the color tone of the cerebral aneurysm 10 in the captured brain, and is shown in shades of black and white, in other words, in achromatic colors. In addition, each of the one or more MRI images contains information indicating which of the MRI images corresponds to a plurality of predetermined points.

[0097] Furthermore, a craniotomy will be performed on the other subject, B.

[0098] The wall thickness estimation device 100 acquires surgical images taken when a craniotomy was performed on another subject B. These surgical images may be two-dimensional or three-dimensional, but in this case, they are three-dimensional. These surgical images contain information indicating the color tone of the cerebral aneurysm 10 in the brain that was captured, and are displayed in chromatic colors.

[0099] Next, the wall thickness estimation device 100 superimposes one of the CT images onto the surgical image. Furthermore, the wall thickness estimation device 100 determines the region in the surgical image corresponding to each of several predetermined points in the single CT image. It is preferable that the wall thickness estimation device 100 is equipped with an operation reception unit such as a keyboard, mouse, and touch panel, and that the above region is determined by the operation reception unit receiving the operation of the wall thickness estimation device 100 by the user. The same process is performed when the video imaging device 400 is an MRI device.

[0100] This links a predetermined point to the region in the surgical image that corresponds to that predetermined point. Furthermore, as described above, the surgical image contains information indicating the color tone of the cerebral aneurysm 10 captured in the brain. Therefore, a predetermined point is linked to the information indicating the color tone of the region that corresponds to that predetermined point.

[0101] Here, we will explain the relationship between the color and thickness of the cerebral aneurysm.

[0102] Generally, in cerebral aneurysms 10 taken during craniotomy, areas with a weak white hue and a strong red hue correspond to areas with fragile or thin blood vessel walls. Conversely, in cerebral aneurysms 10 taken during craniotomy, areas with a strong white hue and a weak red hue correspond to areas with thickened blood vessel walls.

[0103] Therefore, in this embodiment, if the region in the surgical image corresponding to the predetermined point is a region with a weak white tone and a strong red tone, the thickness index, which is an indicator of the thickness of the predetermined point, is set to "1", a "numerical value" indicating that the predetermined point is thin. Similarly, if the region in the surgical image corresponding to the predetermined point is a region with a strong white tone and a weak red tone, the thickness index, which is an indicator of the thickness of the predetermined point, is set to "0", a "numerical value" indicating that the predetermined point is thick.

[0104] A method for determining whether a region in a surgical image has weak white tones and strong red tones, or strong white tones and weak red tones, is preferable to use a method that makes this determination based on the pixel values ​​such as RGB of the region in the surgical image.

[0105] As shown in Figure 7, dataset D1 includes "1" as the thickness index T1.

[0106] While it is possible to obtain 300,000 predetermined points from one other subject B, as described above, it is sufficient to obtain a dataset of between 100 and 150,000 points from that same subject B. In other words, it is not necessary to use each of the obtained predetermined points as training data. As mentioned above, the more datasets that make up the training data, the better.

[0107] As shown in Figure 6, after process S101 is performed, the first learning unit 140 trains the model using the training data acquired in process S101 (first learning process S102). More specifically, the first learning unit 140 trains the machine learning model 121 using machine learning. Furthermore, the first learning unit 140 outputs the trained machine learning model 121 to the generation unit 120.

[0108] Figure 8 is a flowchart showing the processing procedure by which the wall thickness estimation device 100 according to this embodiment estimates the thickness of the aneurysm wall 11 of the cerebral aneurysm 10.

[0109] The first acquisition unit 110 acquires behavioral information, which is numerical information regarding the time change of each predetermined point on the aneurysm wall 11 of the cerebral aneurysm 10 of subject P, via the moving image information processing device 300 (first acquisition step S201).

[0110] Next, the generation unit 120 uses the trained machine learning model 121 to generate estimated information, which is information that visualizes the thickness of each of several predetermined points in the blood vessel wall (generation step S202).

[0111] In the machine learning model 121, when an image showing physical parameters based on behavioral information acquired in the first acquisition step S201 (i.e., the first input image) is input, an index indicating the thickness of each of several predetermined points in the blood vessel wall (i.e., a thickness index) is output.

[0112] More specifically, when a first input image relating to one of several predetermined points is input, a thickness index for that point is output. In this case, a first input image relating to each of several predetermined points on subject P is input, and a thickness index for each of the multiple predetermined points on subject P is output.

[0113] The first input image used in generation step S202 to generate estimated information is an image showing the same physical parameters as the second input image. Therefore, the first input image is as follows:

[0114] The first input image is preferably a two-dimensional image composed of graphs showing physical parameters related to one predetermined point among a plurality of predetermined points. The first input image is composed of multiple graphs, and in the first input image, the multiple graphs are arranged in a k × l matrix (where k and l are natural numbers). Here, the first input image is composed of nine graphs, and in the image, the multiple graphs are arranged in a 3 × 3 matrix.

[0115] The nine graphs are as follows:

[0116] The three graphs in the first column represent the physical parameters in the x-axis direction of a given point. The three graphs in the second column represent the physical parameters in the y-axis direction of a given point. The three graphs in the third column represent the physical parameters in the z-axis direction of a given point.

[0117] The three graphs in the first row have the horizontal axis representing displacement in the x, y, and z directions, and the vertical axis representing acceleration in the x, y, and z directions. The three graphs in the second row have the horizontal axis representing displacement in the x, y, and z directions, and the vertical axis representing velocity in the x, y, and z directions. The three graphs in the third row have the horizontal axis representing velocity in the x, y, and z directions, and the vertical axis representing acceleration in the x, y, and z directions.

[0118] In the first input image, displacement, velocity, and acceleration were used for the horizontal and vertical axes of the graph, similar to the second input image, but this is not limited to this. As described above, one of the parameters relating to the time change of displacement at each of several predetermined points, such as displacement, velocity, acceleration, and strain, may be used for the horizontal axis and another for the vertical axis.

[0119] Using this first input image as input, the generation unit 120 obtains thickness indices for each of a plurality of predetermined points in the blood vessel wall.

[0120] In this case, each of the obtained thickness indices is a "numerical value," as shown in Figure 7. The larger the "numerical value" of one thickness indice, the thinner the thickness at the corresponding predetermined point, and the smaller the "numerical value" of one thickness indice, the thicker the thickness at the corresponding predetermined point. For example, the "numerical value" is preferably between 0 and 1, but is not limited to this. If the "numerical value" is between 0 and 1, the closer the "numerical value" of one thickness indice is to 1, the thinner the thickness at the corresponding predetermined point, and the closer the "numerical value" of one thickness indice is to 0, the thicker the thickness at the corresponding predetermined point.

[0121] The generation unit 120 generates estimation information using the thickness index corresponding to each of the multiple predetermined points output as described above. Here, estimation information is, for example, image information in which the thickness of each of the multiple predetermined points is visualized, but is not limited to this. For example, the estimation information may be a table in which each of the multiple predetermined points corresponds to the thickness index of each of the multiple predetermined points.

[0122] Next, the output unit 130 outputs the estimated information generated by the generation unit 120 (output process S203). In output process S203, the output unit 130 transmits, for example, image data corresponding to the image information generated by the generation unit 120 in generation process S202 to the display device 200.

[0123] The display device 200 acquires the image data output by the output unit 130 and displays an image based on that image data.

[0124] Alternatively, the wall thickness estimation device 100 may perform the wall thickness estimation method by reading a computer program recorded on a computer-readable recording medium such as a CD-ROM.

[0125] [Relationship between estimated information and blood vessel wall thickness] Next, we will explain the relationship between estimated information and the thickness of the blood vessel wall. Here, we will explain the cerebral aneurysm 10 in subject P using the estimated information obtained from the flowchart shown in Figure 8 and the craniotomy procedure.

[0126] Figure 9 is a schematic diagram showing an example of estimated information according to this embodiment.

[0127] More specifically, Figure 9 is an image showing a schematic diagram of the relationship between the shape of the cerebral aneurysm 10, as shown by the image information which is an example of the estimated information output in output process S203 shown in Figure 8, and the thickness index of each of several predetermined points in the cerebral aneurysm 10.

[0128] In the cerebral aneurysm 10 shown in Figure 9, a thickness index is indicated by a color corresponding to each of several predetermined points. The darker the color, the higher the numerical value indicated by the thickness index, and the lighter the color, the lower the numerical value indicated by the thickness index. In Figure 9, the cerebral aneurysm 10 is represented in black and white, but when the output unit 130 actually outputs, the schematic diagram of the cerebral aneurysm 10 may be represented in color.

[0129] Furthermore, a craniotomy was performed on subject P for a brain aneurysm 10.

[0130] Figure 10A is a still image showing a cerebral aneurysm 10 according to this embodiment.

[0131] In Figure 10A, the cerebral aneurysm 10 is represented in black and white, but in actual craniotomy, the still image of the cerebral aneurysm 10 is represented in color. Therefore, the areas of the cerebral aneurysm 10 that are dark in color in Figure 10A correspond to areas with weak white tones and strong red tones in actual craniotomy. Conversely, the areas of the cerebral aneurysm 10 that are light in color in Figure 10A correspond to areas with strong white tones and weak red tones in actual craniotomy.

[0132] Craniotomy revealed the shape of the cerebral aneurysm and the areas of thinness in the vessel wall within that shape.

[0133] Here, we compare the estimated information shown in Figure 9 with the shape of the cerebral aneurysm 10 and the color of the vessel wall revealed by the craniotomy surgery shown in Figure 10A.

[0134] As shown in Figures 9 and 10A, the cerebral aneurysms 10 shown in each image have similar shapes. Furthermore, a circular region A is shown in both Figure 9 and Figure 10A. Region A shown in Figure 9 and region A shown in Figure 10A represent the same corresponding region.

[0135] Within region A shown in Figure 9, there is an area where the color indicating the thickness index is darker, meaning the "numerical value" of the thickness index is high. Therefore, based on the estimated information, the area inside region A shown in Figure 9 is estimated to have a thin thickness.

[0136] Furthermore, the area inside region A shown in Figure 10A is darker in color, meaning that in actual craniotomy, there is an area with a weaker white tone and a stronger red tone. Therefore, it can be estimated from craniotomy that the area inside region A shown in Figure 10A is thin. Note that the area inside region A shown in Figure 10A appears white due to the lighting at the time of imaging, but in actual cerebral aneurysm 10, the area inside region A has a weaker white tone and a stronger red tone.

[0137] In other words, the thickness of the cerebral aneurysm 10 estimated based on estimated information and the thickness of the cerebral aneurysm 10 obtained by actual craniotomy surgery are in good agreement.

[0138] Therefore, the estimated information shown in Figure 9 can be used as highly accurate information about the thickness of the blood vessel wall.

[0139] This type of information is useful for differentiating between cerebral aneurysms that are prone to growth and rupture and those that are not, and for making an appropriate decision on whether or not treatment is necessary.

[0140] In other words, the wall thickness estimation method according to this embodiment generates highly accurate information about blood vessel walls using a minimally invasive technique, thereby providing useful information for implementing specific treatments for vascular diseases. Furthermore, the wall thickness estimation method according to this embodiment is not limited to blood vessel walls, but can also be used to estimate the thickness of organ walls.

[0141] In other words, the wall thickness estimation method according to this embodiment generates highly accurate information about organ walls using a minimally invasive method that does not involve abdominal surgery, open-heart surgery, or craniotomy, thereby providing useful information for taking specific measures against organ diseases.

[0142] [Example 1] The configuration of the model manufacturing system 2000 according to Modification 1 of this embodiment will be described below.

[0143] Figure 10B is a block diagram showing the characteristic functional configuration of the model-making system 2000 related to this modified example.

[0144] The model fabrication system 2000 is a system that fabricates a vascular model, including a vascular wall obtained using, for example, the four-dimensional angiography method described above, based on the estimated information output from the wall thickness estimation system 1000 (more specifically, the output unit 130). If the vascular wall included in the vascular model is the aneurysm wall 11 of a cerebral aneurysm 10, the model fabrication system 2000 further fabricates a brain model into which the vascular model is embedded, and also fabricates a skull model to enclose this brain model.

[0145] Before performing surgery on subject P (i.e., the patient)'s cerebral aneurysm 10, the physician explains the surgery to subject P. The fabricated vascular models, brain models, and skull models are used when the physician explains the surgery to subject P. Following the procedure shown in Figure 8, the thickness of the aneurysm wall 11 of subject P's cerebral aneurysm 10 is estimated, and a vascular model of subject P is fabricated based on the estimated information output in output step S203. When subject P receives the above explanation, a vascular model of subject P themselves is used instead of a general commercially available model, allowing subject P to deepen their understanding of the surgery and undergo the surgery with confidence.

[0146] Next, the functional configuration of the model-making system 2000 related to this modified example will be described in detail.

[0147] As shown in Figure 10B, the model making system 2000 comprises a third acquisition unit 610 and a manufacturing unit 620.

[0148] The third acquisition unit 610 acquires the estimated information generated in the generation process S202. More specifically, the third acquisition unit 610 acquires the estimated information generated in the generation process S202 and further output in the output process S203. The third acquisition unit 610 is, for example, a communication interface that performs wired or wireless communication.

[0149] The fabrication unit 620 fabricates a vascular model including the blood vessel walls. The fabrication unit 620 fabricates the vascular model based on the thickness visualized by the estimated information acquired by the third acquisition unit 610. More specifically, the fabrication unit 620 fabricates the vascular model such that the blood vessel walls included in the vascular model exhibit different characteristics for each thickness. The fabrication unit 620 is, for example, a 3D printer.

[0150] Next, we will explain the specific processing steps in the model-making method performed by the model-making system 2000. Here, we will explain that the blood vessel wall included in the blood vessel model is the aneurysm wall 11 of the cerebral aneurysm 10, but the same applies to blood vessel walls other than the aneurysm wall 11.

[0151] Figure 10C is a flowchart showing the processing procedure for creating a vascular model using the model-making system 2000 related to this modified example.

[0152] First, the third acquisition unit 610 acquires the estimated information generated in the generation step S202 (third acquisition step S401). The estimated information acquired by the third acquisition unit 610 is, for example, image information in which the thickness of each of a plurality of predetermined points is visualized, but is not limited to this. For example, the estimated information may be a table in which each of the plurality of predetermined points corresponds to the thickness index of each of the plurality of predetermined points.

[0153] Figure 10D is a schematic diagram showing an example of estimated information related to this modified example. More specifically, Figure 10D is an image that, like Figure 9, shows a schematic diagram of the relationship between the shape of the cerebral aneurysm 10 shown by the image information, which is an example of estimated information output in output process S203 shown in Figure 8, and the thickness index of several predetermined points in the cerebral aneurysm 10.

[0154] In the cerebral aneurysm 10 shown in Figure 10D, a color indicating thickness is applied to each of several predetermined points, meaning that the blood vessel wall (more specifically, the aneurysm wall 11) shows different colors according to its thickness. Although the cerebral aneurysm 10 shown in Figure 10D is shown in white and black, in reality, the colors are shown in the order of darkest black to red, yellow, green, light blue, and blue. Note that the colors indicating thickness are not limited to the five colors of red, yellow, green, light blue, and blue, but may also be reddish-brown, an intermediate color between reddish-brown and white, and white, for example.

[0155] Furthermore, the darker the black (closer to red), the higher the numerical value of the thickness index, and the lighter the black (closer to blue), the lower the numerical value of the thickness index. As mentioned above, the higher the numerical value of the thickness index, the thinner the thickness of the corresponding point, and the lower the numerical value of the thickness index, the thicker the thickness of the corresponding point.

[0156] Furthermore, the fabrication unit 620 fabricates a vascular model including the blood vessel wall (aneurysm wall 11) (first fabrication step S402). Based on the thickness visualized by the estimated information acquired by the third acquisition unit 610, the fabrication unit 620 fabricates the vascular model such that the blood vessel walls included in the vascular model show different characteristics for each thickness.

[0157] Figure 10E shows a vascular model 30 including a vascular wall (aneurysm wall 11) according to this modified example. The vascular model 30 including a vascular wall (aneurysm wall 11) is also a vascular model 30 including a cerebral aneurysm 10. The fabrication unit 620 fabricates the vascular model 30 based on the model diagram (i.e., estimated information) shown in Figure 10D.

[0158] In Figure 10E, as in Figure 10D, the blood vessel model 30 is shown in white and black. However, in reality, the colors are shown in the order of darkest black to red, yellow, green, light blue, and blue. Furthermore, the darker the black (closer to red), the thinner the blood vessel wall, and the lighter the black (closer to blue), the thicker the blood vessel wall.

[0159] In other words, the fabrication unit 620 fabricates the blood vessel model 30 so that the blood vessel walls included in the blood vessel model 30 show different characteristics depending on their thickness, or more specifically, so that the blood vessel walls (aneurysm walls 11) show different colors depending on their thickness.

[0160] The fabrication unit 620 is fabricated in the first fabrication step S402 when fabricating the blood vessel model 30, for example, by the following first fabrication method or second fabrication method.

[0161] In the first manufacturing method, the manufacturing unit 620 first manufactures a model showing the external shape corresponding to the blood vessel model 30, and then manufactures the blood vessel model 30 by coloring or staining the surface of the manufactured model with red, yellow, green, light blue, and blue. The surface color of the model is not limited to the five colors of red, yellow, green, light blue, and blue, but may also be reddish-brown, an intermediate color between reddish-brown and white, and white, for example.

[0162] In this case, for example, when the fabrication unit 620, which is a 3D printer, fabricates a model that shows the external shape corresponding to the blood vessel model 30, it uses a white or transparent material (e.g., filament or UV resin) to fabricate the model. The model is white, and the blood vessel model 30 is fabricated by coloring or staining this white model.

[0163] In the second manufacturing method, the manufacturing unit 620 may manufacture the blood vessel model 30 as follows. The manufacturing unit 620, which is a 3D printer, manufactures the blood vessel model 30 using red, yellow, green, light blue, and blue materials (for example, filament or UV resin). The manufacturing unit 620 is not limited to the five colors of red, yellow, green, light blue, and blue, but may also use, for example, reddish-brown, an intermediate color between reddish-brown and white, and white materials. In this case, for example, the manufacturing unit 620 may manufacture the blood vessel model 30 according to the colors shown in the model diagram (i.e., estimated information) shown in Figure 10D. According to this manufacturing method, the coloring or staining step can be omitted.

[0164] In the first manufacturing step S402, as shown in Figure 10E, the vascular model 30 was manufactured such that the blood vessel wall (aneurysm wall 11) showed different colors for each thickness at multiple predetermined points, but it is not limited to this. In the first manufacturing step S402, the vascular model 30 should be manufactured such that the blood vessel wall contained in the vascular model 30 shows different characteristics for each thickness, based on the thickness visualized by the estimated information. For example, in the first manufacturing step S402, the vascular model 30 may be manufactured such that the blood vessel wall contained in the vascular model 30 shows different surface textures for each thickness. More specifically, the thicker the surface, the rougher it may be (i.e., the greater the surface irregularities), and the thinner the surface, the smoother it may be (i.e., the smaller the surface irregularities).

[0165] Furthermore, in the first manufacturing step S402, the manufacturing unit 620 manufactured a vascular model 30 including the vascular wall (aneurysm wall 11), but it is not limited to this. The manufacturing unit 620 may also manufacture vascular models that do not include the vascular wall (aneurysm wall 11). In other words, since the vascular model 30 including the vascular wall (aneurysm wall 11) manufactured in the first manufacturing step S402 is a model of only a part of the blood vessels of the brain, it is also desirable to manufacture vascular models that do not include the vascular wall (aneurysm wall 11) (i.e., models of other parts of the blood vessels of the brain that are different from the above-mentioned part).

[0166] A vascular model 30 including the blood vessel wall (aneurysm wall 11) and a vascular model without the blood vessel wall (aneurysm wall 11) are manufactured separately by the manufacturing unit 620. Subsequently, the vascular model 30 including the blood vessel wall and the vascular model without the blood vessel wall are combined to obtain a model of the entire blood vessels of the brain (hereinafter referred to as the whole blood vessel model). Figure 10F shows a whole blood vessel model 31 of the brain according to this modified example.

[0167] Furthermore, it is preferable that the vascular model 30 including the vascular wall (aneurysm wall 11) and the vascular model without the vascular wall (aneurysm wall 11) each have magnets. The magnets of the vascular model 30 including the vascular wall (aneurysm wall 11) and the magnet of the vascular model without the vascular wall (aneurysm wall 11) connect and combine the vascular model 30 including the vascular wall (aneurysm wall 11) and the vascular model without the vascular wall (aneurysm wall 11). In addition, since two magnets are provided, the vascular model 30 including the vascular wall (aneurysm wall 11) and the vascular model without the vascular wall (aneurysm wall 11) are detachable.

[0168] As an example, a vascular model that does not include the blood vessel wall (aneurysm wall 11) is prepared as follows. As described above, 4D angiography is a method that adds a time axis to 3D angiography, and 3D angiography is a method that collects three-dimensional data of blood vessels using an X-ray CT device and an MRI device and extracts vascular information. In order to prepare a vascular model that does not include the blood vessel wall (aneurysm wall 11), for example, in the third acquisition step S401, the third acquisition unit 610 acquires the three-dimensional data of the blood vessel from the moving image capture device 400 or moving image information processing device 300 of the wall thickness estimation system 1000. Subsequently, in the first preparation step S402, the preparation unit 620 prepares a vascular model that does not include the blood vessel wall (aneurysm wall 11) based on the acquired three-dimensional data of the blood vessel.

[0169] Furthermore, while the acquired 3D data of blood vessels contains information indicating the external shape of the vessels, it does not contain information regarding the thickness of the vessel walls. Therefore, a vascular model that does not include the vessel walls (aneurysm walls 11) is a model that shows the external shape of the vessels, but not a model that shows the thickness of the vessel walls.

[0170] Furthermore, the fabrication unit 620 fabricates a brain model into which the blood vessel model 30 fabricated in the first fabrication step S402 is embedded (second fabrication step S403). Figure 10G shows a brain model 40 according to this modified example. In this modified example, the brain model 40 is a model into which a model of the entire blood vessels of the brain, including the blood vessel model 30, is embedded. The brain model 40 has a right brain model 42 and a left brain model 41, and the right brain model 42 and the left brain model 41 are configured to be separable. With the right brain model 42 and the left brain model 41 separated, the model of the entire blood vessels of the brain 31 is placed between the right brain model 42 and the left brain model 41, and the model of the entire blood vessels of the brain 31 is embedded in the brain model 40 by sandwiching it between the right brain model 42 and the left brain model 41.

[0171] In this modified example, the brain model 40 includes a right brain model 42 and a left brain model 41, but is not limited to this. In other examples, the brain model may include a right brain model and a left brain model, and each of the right and left brain models may include the cerebrum, midbrain, cerebellum, and brainstem.

[0172] In the second manufacturing process S403, a brain model 40 is prepared as an example, as follows.

[0173] As described above, three-dimensional data of blood vessels is collected by an X-ray CT scanner and an MRI scanner, etc. When this three-dimensional data of blood vessels is collected, three-dimensional data of the brain and three-dimensional data of the skull are also collected in addition to the three-dimensional data of blood vessels. In this modified example, for example, in the third acquisition step S401, the third acquisition unit 610 also acquires the above-mentioned three-dimensional data of the brain from the video imaging device 400 or video information processing device 300 of the wall thickness estimation system 1000. After the first manufacturing step S402, in the second manufacturing step S403, the manufacturing unit 620 manufactures the brain model 40 based on the acquired three-dimensional data of the brain.

[0174] Next, the fabrication unit 620 fabricates a skull model to enclose the brain model 40 fabricated in the second fabrication step S403 (third fabrication step S404). Figure 10H shows the skull model 50 according to this modified example. In this modified example, the skull model 50 encloses a model 31 of the entire vascular system of the brain, including a vascular model 30, and the brain model 40.

[0175] In Figure 10H, only a model of a portion of the skull is shown to illustrate the space containing the brain model 40, and models of other parts of the skull are not shown. However, in the actual third manufacturing step S404, a model representing the entire skull (i.e., the skull model 50) is manufactured.

[0176] Furthermore, the skull model 50 includes the neurocranium (neurocranium), which consists of the occipital bone, temporal bone, parietal bone, frontal bone, and sphenoid bone, and the visceral cranial (visceral craniium), which consists of the ethmoid bone, lacrimal bone, nasal bone, maxilla, mandible, palatine bone, inferior nasal concha, zygomatic bone, vomer, and hyoid bone.

[0177] In the third manufacturing process S404, a skull model 50 is produced as an example, as follows.

[0178] As described above, three-dimensional data of the skull is also collected by an X-ray CT scanner and an MRI scanner. In this modified example, for example, in the third acquisition step S401, the third acquisition unit 610 also acquires the three-dimensional data of the skull from the video imaging device 400 or video information processing device 300 of the wall thickness estimation system 1000. After the first manufacturing step S402 and the second manufacturing step S403, in the third manufacturing step S404, the manufacturing unit 620 manufactures a skull model 50 based on the acquired three-dimensional data of the skull.

[0179] As described above, in this modified example, a vascular model 30 of subject P himself is created based on estimated information, and furthermore, a brain model 40 and a skull model 50 of subject P himself are created.

[0180] Furthermore, in the second manufacturing step S403, the brain model 40 may be manufactured as follows. For example, instead of using the 3D data of subject P's brain, 3D data of a commercially available brain model of a typical size may be used. Such 3D data of the brain can be obtained by measuring a commercially available brain model of a typical size using an X-ray CT scanner, an MRI scanner, and a 3D scanner capable of measuring three-dimensional shape (3D form). In this case, subject P's own brain model 40 is not manufactured.

[0181] When a doctor explains the surgery for a brain aneurysm 10 in subject P (i.e., the patient), the prepared vascular model 30, brain model 40, and skull model 50 of subject P are used. In particular, with respect to the vascular model 30, since the vascular walls in the vascular model 30 show different characteristics depending on their thickness, subject P can easily understand which parts of the vascular wall are thick and which parts are thin. Subject P can deepen their understanding of the above surgery and be able to undergo the surgery with confidence. In other words, the model-making method according to this modified example is a method that can support doctors in explaining surgery to subject P (i.e., the patient).

[0182] Furthermore, for example, the fabricated blood vessel model 30, brain model 40, and skull model 50 can be used by a physician to conduct a rehearsal of the surgery before the operation. This allows the physician to approach the surgery with confidence and peace of mind. In other words, the model fabrication method according to this modified example is a method that can support physicians in performing surgery with confidence and peace of mind.

[0183] Furthermore, in this modified example, the blood vessel model 30 and brain model 40 produced in the first production step S402 and the second production step S403 should have flexibility and elasticity. For example, when a doctor or subject P touches the blood vessel model 30 and brain model 40 with their hands, the blood vessel model 30 and brain model 40 should deform, and when they release their hands, the blood vessel model 30 and brain model 40 should return to their original shape. For example, the blood vessel model 30 and brain model 40 should be produced in the first production step S402 and the second production step S403 using a material such as silicone resin.

[0184] Furthermore, in this modified example, the manufacturing unit 620 that produced the brain model 40 and skull model 50 in the second manufacturing step S403 and the third manufacturing step S404 was a 3D printer, but is not limited to this. For example, the manufacturing unit 620 may be a mold into which the brain model 40 and skull model 50 have been molded, and the brain model 40 and skull model 50 may be produced by pouring resin into the mold.

[0185] Furthermore, Figure 10I shows a vascular model 30a that includes the vascular wall (aneurysm wall 11a) of another subject C, in addition to subject P. The vascular model 30a that includes the vascular wall (aneurysm wall 11a) is also a vascular model 30a that includes a cerebral aneurysm 10a. The vascular model 30a shown in Figure 10I was made using the same method as the vascular model 30 of subject P.

[0186] Furthermore, as shown in Figure 10I, the blood vessel model 30a is tubular, and also has holes 12a corresponding to the flow channels for blood. The blood vessel model 30 shown in Figure 10E is also tubular, similar to the blood vessel model 30a. Note that these holes 12a do not have to be provided; in other words, the blood vessel model 30 shown in Figure 10E does not have to be tubular.

[0187] [Differentiation 2] The configuration of the wall thickness estimation system 1000a according to a modified example 2 of this embodiment will be described below.

[0188] Figure 11 is a block diagram showing the characteristic functional configuration of the wall thickness estimation system 1000a according to this modified example.

[0189] The wall thickness estimation system 1000a differs from the wall thickness estimation system 1000 according to the embodiment mainly in that it includes a learning device 500, and that the wall thickness estimation system 100a does not include a first learning unit 140.

[0190] The learning device 500 acquires behavioral information generated by the video information processing device 300. The learning device 500 uses one or more datasets, each consisting of a combination of images showing physical parameters based on the acquired behavioral information and a thickness index, as training data to train a model (in this case, a machine learning model 121). The learning device 500 also outputs the trained model to the generation unit 120 of the wall thickness estimation device 100a. The learning device 500 is, for example, a personal computer, but it may also be a high-performance server connected to a network.

[0191] The learning device 500 comprises a second acquisition unit 110a and a second learning unit 140a.

[0192] The second acquisition unit 110a acquires behavioral information, which is numerical information relating to the time change in the position of each of several predetermined points on an organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall. More specifically, the second acquisition unit 110a acquires behavioral information, which is numerical information relating to the time change in the position of each of several predetermined points on an organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography. Specifically, the second acquisition unit 110a acquires behavioral information generated by the moving image information processing device 300. The second acquisition unit 110a is, for example, a communication interface that performs wired communication or wireless communication.

[0193] The second learning unit 140a trains the model using one or more datasets as training data. Specifically, the second learning unit 140a is implemented by a processor, microcomputer, or dedicated circuit that executes a program.

[0194] A single dataset consists of a combination of images showing physical parameters based on behavioral information of a predetermined point among multiple predetermined points in an organ wall or blood vessel wall, and an index indicating the thickness of that predetermined point.

[0195] The behavioral information refers to the information acquired by the second acquisition unit 110a.

[0196] In this modified example, the training data may be data generated by the second learning unit 140a.

[0197] Next, we will explain the specific processing steps in the learning method performed by the learning device 500. While we will use a blood vessel wall as an example, the same procedure applies to organ walls.

[0198] Figure 12 is a flowchart showing the processing procedure by which the learning device 500 according to this modified example trains a model (machine learning model 121).

[0199] First, the second acquisition unit 110a acquires behavioral information (second acquisition step S301).

[0200] Furthermore, the second learning unit 140a generates training data based on the acquired behavioral information and trains the model (second learning process S302). More specifically, the second learning unit 140a trains the machine learning model 121. The second learning unit 140a outputs the trained machine learning model 121 to the generation unit 120.

[0201] As this modified example shows, the wall thickness estimation device 100a that generates estimated information and the learning device 500 that trains the model (machine learning model 121) may be separate devices.

[0202] In the above description, the second acquisition unit 110a acquired behavioral information based on a moving image including an organ wall or blood vessel wall obtained using four-dimensional angiography, but it is not limited to this. The moving image may also be a moving image (two-dimensional video) obtained using a two-dimensional video imaging device. In other words, the second acquisition unit 110a may acquire behavioral information, which is numerical information regarding the time change of the position of each of a plurality of predetermined points on the organ wall or blood vessel wall, based on a moving image (two-dimensional image) including an organ wall or blood vessel wall obtained using a two-dimensional video imaging device.

[0203] The 2D video in question is, for example, a video of one or more other subjects besides subject P. For simplicity, we will mainly use one other subject D as an example.

[0204] The two-dimensional video in question is a surgical video taken when an abdominal or craniotomy was performed on another subject, D. Unlike the video obtained using four-dimensional angiography, this two-dimensional video is not three-dimensional data (three-dimensional data), meaning it does not contain information indicating the depth of the two-dimensional video. In this case, the video capture device 400 corresponds to a two-dimensional video capture device (e.g., a camera).

[0205] Furthermore, the video information processing device 300 acquires the two-dimensional video captured by the video image capture device 400. The video information processing device 300 estimates the depth of the two-dimensional video and generates depth information indicating the estimated depth. For example, the video information processing device 300 uses a depth estimation AI model to estimate the depth of the two-dimensional video, but is not limited to this, and other methods may be used.

[0206] Figure 13 shows a still image (1 frame) contained in the 2D video according to this modified example, and an image showing the depth estimated for that still image. More specifically, Figure 13(a) shows a still image (1 frame) contained in the 2D video, and Figure 13(b) shows an image showing the depth estimated for that still image. In Figure 13(b), the deeper the depth, the darker the color, and the shallower the depth, the lighter the color. The information combining the 2D video and depth information becomes 3D data (three-dimensional data), just like a moving image obtained using 4D angiography.

[0207] The video information processing device 300 generates behavior information, which is numerical information regarding the time change of the position of each of a plurality of predetermined points on the organ wall or blood vessel wall, based on the video image including the organ wall or blood vessel wall obtained using the two-dimensional video imaging device and depth information indicating the estimated depth. Subsequently, the second acquisition unit 110a acquires the generated behavior information. In other words, the behavior information acquired by the second acquisition unit 110a is information based on the video image including the organ wall or blood vessel wall obtained using the two-dimensional video imaging device.

[0208] Thus, even if the above-mentioned video is a video obtained using a two-dimensional video imaging device (two-dimensional video), behavioral information is estimated and acquired by the second acquisition unit 110a, just as if the video were a video obtained using four-dimensional angiography. The subsequent second learning step S302 is performed in the same manner.

[0209] [Effects, etc.] As described above, the wall thickness estimation method according to this embodiment includes a first acquisition step S201, a generation step S202, and an output step S203. The first acquisition step S201 acquires behavioral information, which is numerical information regarding the time change of the position of each of a plurality of predetermined points on the organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography. The generation step S202 takes an image showing physical parameters based on the behavioral information acquired in the first acquisition step S201 as input and generates estimated information, which is information in which the thickness is visualized, using a trained model that outputs an index showing the thickness of each of a plurality of predetermined points on the organ wall or blood vessel wall. The output step S203 outputs the estimated information generated in the generation step S202.

[0210] As a result, in the wall thickness estimation method, for example, a moving image including the blood vessel wall is generated using an X-ray CT or MRI device and four-dimensional angiography. Compared to methods such as craniotomy, for example, a moving image including the blood vessel wall can be obtained using a less invasive method. The wall thickness estimation method can generate estimation information, which is information that visualizes the thickness of each of several predetermined points in the blood vessel wall, using behavioral information related to the moving image. The blood vessel wall thickness estimated based on the estimation information closely matches the blood vessel wall thickness obtained by craniotomy.

[0211] In other words, the wall thickness estimation method can generate highly accurate information about the wall thickness near each of several predetermined points on the blood vessel wall. In this embodiment, for example, the thickness of the aneurysm wall 11 of a cerebral aneurysm 10 is estimated. Such information is useful for distinguishing between cerebral aneurysms that are prone to growing and rupturing and those that are not prone to growing and rupturing, and for appropriately determining whether treatment is necessary.

[0212] Furthermore, the wall thickness estimation method is not limited to blood vessel walls; it can also be used to estimate the thickness of organ walls.

[0213] In other words, the wall thickness estimation method according to this embodiment generates highly accurate information about organ walls or blood vessel walls using a minimally invasive method, thereby providing useful information for taking specific measures against organ or blood vessel diseases.

[0214] Furthermore, the wall thickness estimation method according to this embodiment further includes a first learning step S102. In the first learning step S102, a model is trained using one or more datasets as training data, each dataset consisting of a combination of an image showing physical parameters based on behavioral information of a predetermined point and an index showing the thickness of that predetermined point.

[0215] This allows the model to output an index indicating thickness based on an image showing the input physical parameters. Therefore, the wall thickness estimation method according to this embodiment can generate more accurate information about organ walls or blood vessel walls using a minimally invasive method.

[0216] Furthermore, the first learning step in this embodiment involves training the model using machine learning.

[0217] As a result, in the generation step S202, estimated information can be generated using a model trained with machine learning (machine learning model 121). Therefore, the wall thickness estimation method according to this embodiment can generate more accurate information about organ walls or blood vessel walls using a minimally invasive method.

[0218] Furthermore, the estimation information in this embodiment is image information in which the thickness is visualized.

[0219] This allows estimated information to be obtained as image information. Therefore, for example, doctors can visually obtain highly accurate information about the thickness of organ walls or blood vessel walls.

[0220] Furthermore, in the method for estimating wall thickness, the blood vessel wall may also be the aneurysm wall in an aneurysm or varicose vein.

[0221] This allows the wall thickness estimation method to estimate the thickness of the aneurysm wall of an aneurysm or varicose vein.

[0222] Furthermore, in the wall thickness estimation method according to this embodiment, the blood vessel wall is the aneurysm wall 11 in the cerebral aneurysm 10.

[0223] This allows the wall thickness estimation method to estimate the thickness of the aneurysm wall 11 of the cerebral aneurysm 10.

[0224] Furthermore, in the wall thickness estimation method, the blood vessel wall may be an arterial or vein blood vessel wall.

[0225] This allows the wall thickness estimation method to estimate the thickness of the blood vessel wall of an artery or vein.

[0226] Alternatively, the computer program may be used to have the computer execute the wall thickness estimation method described above.

[0227] This allows the above wall thickness estimation method to be executed by a computer.

[0228] Furthermore, the learning method according to the modified example 2 includes a second acquisition step S301 and a second learning step S302. The second acquisition step S301 acquires behavioral information, which is numerical information relating to the time change of the position of each of a plurality of predetermined points on an organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall. The second learning step S302 trains a model using one or more datasets obtained in the second acquisition step as training data, each dataset consisting of a combination of an image showing physical parameters based on the behavioral information of a predetermined point among a plurality of predetermined points and an index showing the thickness of a predetermined point among a plurality of predetermined points.

[0229] When such a model is used in the generation process S202 according to this embodiment, the model can output an index indicating thickness based on an image showing the input physical parameters. Therefore, the wall thickness estimation method using the learning method according to Modified Example 2 can generate more accurate information about organ walls or blood vessel walls using a minimally invasive method.

[0230] In the learning method relating to Modification 2, the video is a video obtained using 4D angiography or a 2D video imaging device.

[0231] As a result, the second acquisition step S301 can acquire behavioral information based on moving images obtained using 4D angiography or a 2D video imaging device.

[0232] The model manufacturing method according to Modification 1 includes a third acquisition step S401 for acquiring estimated information generated by the generation step S202 described above, and a first manufacturing step S402 for manufacturing a vascular model 30 including the vascular walls described above, wherein the vascular model 30 is manufactured such that the vascular walls included in the vascular model 30 show different characteristics for each thickness, based on the thickness visualized by the estimated information acquired by the third acquisition step S401.

[0233] In Modification 1, a model 30 of the subject P's own blood vessels is created. The created model 30 of the subject P's own blood vessels is used when the doctor explains the surgery to subject P (i.e., the patient). In particular, with respect to the model 30, since the blood vessel walls in the model 30 show different characteristics depending on their thickness, subject P can easily understand which parts of the blood vessel wall are thick and which parts are thin. As a result, subject P can deepen their understanding of the surgery and undergo the surgery with confidence. In short, the model creation method according to Modification 1 is a method that can support the doctor in explaining the surgery to subject P (i.e., the patient).

[0234] Furthermore, for example, the fabricated blood vessel model 30 can be used by a physician to conduct a rehearsal of the surgery before the operation. This allows the physician to approach the surgery with confidence and peace of mind. In other words, the model fabrication method according to this modified example is a method that can support physicians in performing surgery with confidence and peace of mind.

[0235] In the model manufacturing method according to Modification 1, in the first manufacturing step S402, a vascular model 30 is manufactured such that the blood vessel walls show different colors according to their thickness.

[0236] This allows subject P to more easily understand which parts of the blood vessel wall are thicker and which parts are thinner. Subject P will be able to gain a deeper understanding of the surgery and undergo the surgery with greater confidence. In other words, a model-making method is realized that can more easily assist doctors in explaining surgery to subject P (i.e., the patient).

[0237] In the model manufacturing method according to Modification 1, the blood vessel wall included in the blood vessel model 30 manufactured in the first manufacturing step S402 is the aneurysm wall 11 of a cerebral aneurysm 10. The model manufacturing method according to Modification 1 includes a second manufacturing step S403 for manufacturing a brain model 40 into which the blood vessel model 30 manufactured in the first manufacturing step S402 is embedded.

[0238] In Modification 1, a vascular model 30 and a brain model 40 of subject P are created. The created vascular model 30 and brain model 40 of subject P are used when a doctor explains surgery for subject P's (i.e., patient's) cerebral aneurysm 10. If the vascular wall included in the vascular model 30 is the aneurysm wall 11 of the cerebral aneurysm 10, subject P can easily understand where the cerebral aneurysm 10 and the aneurysm wall 11 are located in subject P's brain. As a result, subject P can deepen their understanding of the surgery and undergo the surgery with confidence. In other words, a model creation method is realized that can more easily support doctors in explaining surgery to subject P (i.e., the patient).

[0239] The model manufacturing method according to Modification 1 includes a third manufacturing step S404 for manufacturing a skull model 50 to enclose the brain model 40 manufactured in the second manufacturing step S403.

[0240] In Modification 1, a model of the blood vessels and brain of subject P is created. The created model of subject P's blood vessels 30, brain 40, and skull 50 are used by the doctor when explaining the surgery for subject P's (i.e., the patient's) cerebral aneurysm 10. Subject P can easily understand the positional relationship between the cerebral aneurysm 10, the aneurysm wall 11, the brain, and the skull. As a result, subject P can deepen their understanding of the surgery and undergo the surgery with confidence. In other words, a model creation method is realized that can more easily support the doctor in explaining the surgery to subject P (i.e., the patient).

[0241] Furthermore, the skull model 50 includes the neurocranium (neurocranium), which consists of the occipital bone, temporal bone, parietal bone, frontal bone, and sphenoid bone, and the visceral cranial (visceral craniium), which consists of the ethmoid bone, lacrimal bone, nasal bone, maxilla, mandible, palatine bone, inferior nasal concha, zygomatic bone, vomer, and hyoid bone.

[0242] This allows subject P to easily understand the positional relationship between the forehead and the back of the head, in other words, the front and back of the face. As a result, subject P can deepen their understanding of the surgery and undergo the surgery with confidence. In short, this method of creating a model makes it easier for doctors to explain surgery to subject P (i.e., the patient).

[0243] Furthermore, the wall thickness estimation device 100 according to this embodiment comprises a first acquisition unit 110, a generation unit 120, and an output unit 130. The first acquisition unit 110 acquires behavioral information, which is numerical information regarding the time change of the position of each of a plurality of predetermined points on the organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography. The generation unit 120 takes an image showing physical parameters based on the behavioral information acquired by the first acquisition unit 110 as input and generates estimation information, which is information in which the thickness is visualized, using a trained model that outputs an index showing the thickness of each of a plurality of predetermined points on the organ wall or blood vessel wall. The output unit 130 outputs the estimation information generated by the generation unit 120.

[0244] As a result, the wall thickness estimation device 100 generates, for example, a moving image including the blood vessel wall using an X-ray CT or MRI device and four-dimensional angiography. Compared to methods such as craniotomy, for example, a moving image including the blood vessel wall can be obtained using a less invasive method. The wall thickness estimation device 100 can use behavioral information related to the moving image to generate estimation information, which is information that visualizes the thickness of each of several predetermined points in the blood vessel wall. The blood vessel wall thickness estimated based on the estimation information closely matches the blood vessel wall thickness obtained by craniotomy.

[0245] In other words, the wall thickness estimation device 100 can generate highly accurate information about the wall thickness near each of several predetermined points on the blood vessel wall. In this embodiment, for example, the thickness of the aneurysm wall 11 of a cerebral aneurysm 10 is estimated. Such information is useful for distinguishing between cerebral aneurysms that are prone to growing and rupturing and those that are not prone to growing and rupturing, and for appropriately determining whether treatment is necessary.

[0246] Furthermore, the wall thickness estimation device 100 is not limited to blood vessel walls, but can also be used to estimate the thickness of organ walls.

[0247] In other words, the wall thickness estimation device 100 according to this embodiment can generate highly accurate information about organ walls or blood vessel walls using a minimally invasive method, thereby providing useful information for taking specific measures against organ or blood vessel diseases.

[0248] Furthermore, the wall thickness estimation system 1000 according to this embodiment includes the wall thickness estimation device 100 described above, a video information processing device 300 that acquires video images, generates behavior information and outputs it to the first acquisition unit 110, and a display device 200 that displays the estimation information output by the output unit 130.

[0249] As described above, the wall thickness estimation device 100 according to this embodiment can generate highly accurate information about organ walls or blood vessel walls using a minimally invasive method. Therefore, the wall thickness estimation system 1000 according to this embodiment, which includes such a wall thickness estimation device 100, can propose useful information for taking specific measures against organ or blood vessel diseases.

[0250] Furthermore, by visualizing and displaying the estimated information, for example, doctors can visually obtain highly accurate information about the thickness of organ walls or blood vessel walls.

[0251] (Other embodiments) Although the embodiments and modified versions of the wall thickness estimation method have been described above, the present invention is not limited to the embodiments described above.

[0252] Furthermore, the first learning unit 140 may update the model (machine learning model 121) using machine learning.

[0253] The update of the machine learning model 121 by the first learning unit 140 does not need to be performed in real time, but may be performed retrospectively using the second input image and the thickness index associated with the second input image as training data.

[0254] The inventor has also conducted the following verifications.

[0255] In the above embodiment, when the first learning unit 140 trained the model (machine learning model 121), it used training data obtained from multiple other subjects in addition to subject P. Here, as a verification, the inventor used training data obtained from subject P and multiple other subjects instead of this training data. In other words, the first learning unit 140 trained the machine learning model 121 using training data obtained from subject P and multiple other subjects. Even in this case, the generation unit 120 can generate estimated information of subject P using this machine learning model 121.

[0256] In this embodiment and its modified form, the thickness index of the predetermined point is an index obtained based on the color tone of the cerebral aneurysm 10 in the brain shown in the moving image including the blood vessel wall obtained by the moving image capture device 400 and the surgical image obtained by craniotomy.

[0257] However, the thickness index of the predetermined point may be obtained based on the video image and other information different from the surgical image. Other information may include, for example, information obtained by mathematical analysis of the estimated mass of each of a plurality of predetermined points.

[0258] In this case, the heavier the mass of each predetermined point, the thicker the thickness of that predetermined point, and the lighter the mass of each predetermined point, the thinner the thickness of that predetermined point. Therefore, the thickness index of the 20,000 points with the heaviest mass among the multiple predetermined points may be set to "0", and the thickness index of the 2,000 points with the lightest mass among the multiple predetermined points may be set to "1". The thickness index may be obtained in this way.

[0259] In the above embodiment, a method for obtaining behavioral information was demonstrated using actual cases and 4D angiography. However, the method for obtaining behavioral information is not limited to this. For example, behavioral information may also be obtained by the methods shown in Other Example 1 and Other Example 2 below.

[0260] In the other example method (1), behavioral information is obtained by using an artificially created artificial aneurysm, an artificial heart connected to the artificial aneurysm, and an imaging device.

[0261] An artificial aneurysm has an artificial blood vessel and an artificial aneurysm. The artificial blood vessel and artificial aneurysm are made to mimic human blood vessels and aneurysms that have developed in human blood vessels. The artificial aneurysm may be made of a rubber material, for example, and silicone rubber, fluororubber, etc. can be used.

[0262] Furthermore, the artificial tumor may be made of, for example, silicone resin. However, it is not limited to the above, as long as the artificial tumor is made of a flexible material.

[0263] Artificial aneurysms are created using image data obtained from the X-ray CT or MRI scanner described above. This image data includes data on human blood vessels and aneurysms that have developed in those blood vessels.

[0264] The artificial aneurysm is created based on DICOM (Digital Imaging and Communications in Medicine) data related to the image data obtained above.

[0265] An artificial heart is a device that takes over the pumping function of a human heart. This artificial heart is connected to an artificial aneurysm, and by activating the artificial heart's pumping function, the artificial aneurysm pulsates. This movement of the artificial aneurysm, along with imaging equipment, allows for the acquisition of behavioral information.

[0266] The imaging device is, for example, a camera capable of capturing both still images and moving images. Furthermore, the imaging device may be a device that can obtain information on the three-dimensional coordinates and displacement in three-dimensional space of the surface of the object being observed. Such an imaging device can obtain all the information on the three-dimensional coordinates, displacement in three-dimensional space, velocity in three-dimensional space, and acceleration in three-dimensional space of the surface of the object being observed by imaging for 1 second, 5 seconds, or 10 seconds.

[0267] The imaging time of the imaging device is not limited to the above and may be other times. In this case as well, an X-ray CT scanner or an MRI scanner can be used.

[0268] As described above, in the other example 1 method, the imaging device obtains information on the three-dimensional coordinates and displacement in three-dimensional space of the artificial tumor surface by imaging the pulsating artificial tumor. Behavioral information may be obtained based on any or all of this information on three-dimensional coordinates and displacement in three-dimensional space.

[0269] In the other example method (Example 1), behavioral information can be obtained more easily because it is a less invasive technique compared to the craniotomy described above.

[0270] Furthermore, in the other example method (2), behavioral information can be obtained by using a model animal in which aneurysms have developed in blood vessels and the imaging device described above. In this case as well, an X-ray CT scanner or an MRI scanner can be used.

[0271] Specifically, the imaging device images the blood vessels and aneurysms of a model animal, thereby obtaining information on the three-dimensional coordinates of the surfaces of the blood vessels and aneurysms of the model animal and their displacement in three-dimensional space. Behavioral information may be obtained based on any or all of this information.

[0272] In the other example method (Example 2), unlike the human case shown in the embodiment, consent forms from the human subject are not required. Furthermore, since the surface of the blood vessels and aneurysms of the model animal can be marked with patterns necessary for imaging (e.g., marking by spraying), precise time-evolution data of three-dimensional coordinates can be obtained.

[0273] Furthermore, data on the blood vessels and aneurysms of the model animals can be acquired at equal time intervals (e.g., once every two weeks). Therefore, behavioral information can be obtained more easily compared to the embodiment.

[0274] By using the above method, a large amount of behavior information can be easily obtained, and as a result, a large amount of estimation information can be obtained. This is expected to improve the accuracy of information related to the wall.

[0275] In the present embodiment, the case where the blood vessel wall thickness is the thickness of the aneurysm wall 11 of the cerebral aneurysm 10 is shown; however, as described above, the thickness may be the thickness of a blood vessel wall including an artery or a vein. For example, when the blood vessel wall is the thickness of a blood vessel including an artery or a vein, the degree of stenosis of the artery or vein can be estimated by using the wall thickness estimation method according to the embodiment.

[0276] In each of the above embodiments, each component may be configured by dedicated hardware, or may be realized by executing a software program suitable for each component. Each component may be realized by a program execution unit such as a CPU or a processor reading and executing a software program recorded in a recording medium such as a hard disk or a semiconductor memory.

[0277] In addition, the present invention also includes forms obtained by applying various modifications that a person skilled in the art can conceive to each embodiment, and forms realized by any combination of components and functions in each embodiment without departing from the spirit of the present invention. [Industrial Applicability]

[0278] The wall thickness estimation method according to the present invention can be used for various applications such as medical devices and medical methods. [Description of Symbols]

[0279] 10, 10a Cerebral aneurysm 11, 11a Aneurysm wall 12a Hole 20 Parent blood vessel 30, 30a Blood vessel model 31 Whole blood vessel model 40 Brain model 41 Left brain model 42 Right brain model 50 Skull Models 100, 100a Wall Thickness Estimation Device 110 First acquisition part 110a 2nd acquisition part 120 Generation part 130 Output section 140 First Learning Department 140a Second Learning Department 200 Display device 300 Motion Image Information Processing Device 400 Motion Image Capture Device 500 Learning Devices 610 Third Acquisition Department 620 Manufacturing Department 1000, 1000a Wall Thickness Estimation System 2000 Model Making System Area A B Other subjects C Other subjects D Other subjects P Subject p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, p10, p11 points S102 First Learning Process S201 1st acquisition process S202 Generation process S203 Output Process S301 2nd acquisition process S302 Second Learning Process S401 3rd acquisition process S402 First Manufacturing Process S403 Second Manufacturing Process S404 Third Manufacturing Process

Claims

1. A first acquisition step involves acquiring behavioral information, which is numerical information relating to the time change of each of the positions of multiple predetermined points on the organ wall or blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography, A generation step involves using a trained model that takes an image showing physical parameters based on the behavioral information acquired in the first acquisition step as input and outputs an index showing the thickness of each of the plurality of predetermined points in the organ wall or the blood vessel wall to generate estimated information in which the thickness is visualized; The process includes an output step that outputs the estimated information generated by the generation step, The aforementioned image is a two-dimensional image composed of multiple graphs showing the physical parameters relating to one predetermined point among the plurality of predetermined points, The aforementioned multiple graphs are arranged in a k × l matrix (where k and l are natural numbers) in the image. Method for estimating wall thickness.

2. Furthermore, the process includes a first learning step in which the model is trained using one or more datasets as training data, each dataset consisting of a combination of an image showing the physical parameters based on the behavior information of one of the plurality of predetermined points and an index showing the thickness of one of the predetermined points. The method for estimating wall thickness according to claim 1.

3. The first learning step involves training the model using machine learning. The method for estimating wall thickness according to claim 2.

4. The estimated information is image information in which the thickness is visualized. The method for estimating wall thickness according to claim 1.

5. The aforementioned blood vessel wall is the aneurysm wall in an aneurysm or varicose vein. The method for estimating wall thickness according to claim 1.

6. The aforementioned blood vessel wall is the aneurysm wall in a cerebral aneurysm. The method for estimating wall thickness according to claim 1.

7. The aforementioned blood vessel wall is the blood vessel wall of an artery or a vein. The method for estimating wall thickness according to claim 1.

8. A computer program for causing a computer to execute the wall thickness estimation method described in any one of claims 1 to 7.

9. A second acquisition step involves acquiring behavioral information, which is numerical information relating to the time change of the position of each of a plurality of predetermined points on the organ wall or the blood vessel wall, based on a moving image including the organ wall or the blood vessel wall. The process includes a second learning step in which a model is trained using one or more datasets as training data, each dataset consisting of a combination of an image showing physical parameters based on the behavior information of a predetermined point among a plurality of predetermined points, and an index showing the thickness of a predetermined point among a plurality of predetermined points, the behavior information obtained in the second acquisition step, The aforementioned image is a two-dimensional image composed of multiple graphs showing the physical parameters relating to one predetermined point among the plurality of predetermined points, The aforementioned multiple graphs are arranged in a k × l matrix (where k and l are natural numbers) in the image. Learning methods.

10. The aforementioned video is a video obtained using four-dimensional angiography or a two-dimensional video imaging device. The learning method according to claim 9.

11. A third acquisition step for acquiring the estimated information generated by the generation step described in claim 1, A step for producing a vascular model including the blood vessel wall described in claim 1, comprising: a first production step of producing the vascular model such that the blood vessel wall included in the vascular model exhibits different characteristics for each thickness, based on the thickness visualized by the estimated information obtained in the third acquisition step. Model making method.

12. In the first manufacturing step, the blood vessel model is manufactured such that the blood vessel wall exhibits different colors according to its thickness. The method for making a model according to claim 11.

13. The blood vessel wall included in the blood vessel model produced by the first manufacturing step is the aneurysm wall in a cerebral aneurysm. The aforementioned model manufacturing method is The process includes a second manufacturing step of creating a brain model into which the blood vessel model created in the first manufacturing step will be embedded. The method for making a model according to claim 12.

14. The process includes a third manufacturing step of creating a skull model to enclose the brain model created in the second manufacturing step. The method for making a model according to claim 13.

15. An acquisition unit that acquires behavioral information, which is numerical information relating to the time change of each of the positions of a plurality of predetermined points on the organ wall or the blood vessel wall, based on a moving image including the organ wall or blood vessel wall obtained using four-dimensional angiography, A generation unit generates estimated information in which the thickness is visualized, using a trained model that takes an image showing physical parameters based on the behavior information acquired by the acquisition unit as input and outputs an index showing the thickness of each of the plurality of predetermined points in the organ wall or the blood vessel wall as output. The system comprises an output unit that outputs the estimated information generated by the generation unit, The aforementioned image is a two-dimensional image composed of multiple graphs showing the physical parameters relating to one predetermined point among the plurality of predetermined points, The aforementioned multiple graphs are arranged in a k × l matrix (where k and l are natural numbers) in the image. Wall thickness estimation device.

16. The wall thickness estimation device according to claim 15, A motion image information processing device that acquires the motion image, generates the behavior information and outputs it to the acquisition unit, The system includes a display device that displays the estimated information output by the output unit. Wall thickness estimation system.

Images