Image analysis system, support orthosis, image analysis method, and program

Abstract

Provided are an image analysis system, an assistive apparatus, an image analysis method, and a program that make it possible to more easily acquire the state of a knee joint during walking.SOLUTION: An image analysis system includes an analysis processing unit that predicts a meniscus dynamic amount and a mechanical index of a knee joint part of a subject during walking from a feature amount of an image of the inside of a knee of the subject during standing or walking, the image being acquired using an ultrasonic wave.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an image analysis system, an assistive device, an image analysis method, and a program. [Background technology]

[0002] One example of knee joint inflammation is osteoarthritis (hereinafter referred to as "knee OA"), in which joint deformation develops and progresses due to cartilage lesions. Such knee joint inflammation can lead to walking difficulties, knee joint pain, and can be a factor in reduced motor and mobility function and the need for nursing care (see Figure 3).

[0003] There are two methods for measuring the condition of the knee joint: invasive, such as arthroscopic evaluation, and non-invasive. It is difficult to translate invasive procedures such as arthroscopic evaluation into proactive preventive interventions. Non-invasive methods include detecting the condition of the knee joint using X-ray images. X-ray images can only obtain still images (transmission images) in a stationary, standing position, and are therefore difficult to identify unless joint deformation has occurred following cartilage damage, limiting their ability to detect early pathology. Furthermore, it is difficult to detect the condition during movement or walking.

[0004] Similarly, there is a non-invasive method for detecting the condition of the knee joint using ultrasound diagnostic images.Using ultrasound diagnostic images can avoid some of the limitations of using X-ray images, but it is basically limited to evaluation of the condition in a static standing position.

[0005] In response to this, a method is known in which ultrasonic waves are used to sequentially detect the state of the knee joint during walking (see, for example, Patent Document 1).A method is known in which an acceleration sensor is used to detect movement around the knee (see, for example, Patent Documents 2 and 3). [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Publication No. 2022-20362 [Patent Document 2] Japanese Patent Publication No. 2022-51450 [Patent Document 3] Japanese Patent Publication No. 2022-51451 Summary of the Invention [Problem to be solved by the invention]

[0007] It is desirable to slow the progression of knee OA by taking preventative measures before the symptoms worsen and by providing appropriate treatment according to the condition of each individual's knee joint. However, even if the acceleration sensor can detect the movement around the knee, it is difficult to identify the internal state of the knee joint. Therefore, it remains difficult to easily identify the state of the knee joint during walking.

[0008] The present invention has been made to solve the above-mentioned problems, and aims to provide an image analysis system, an assistive device, an image analysis method, and a program that make it possible to more easily obtain the state of the knee joint during walking. [Means for solving the problem]

[0009] (1) An image analysis system according to one aspect of the present invention is an image analysis system that includes an analysis processing unit that predicts meniscus dynamics when a subject is walking based on features of an image of the inside of the knee obtained using ultrasound while the subject is standing or walking. (2) In the image analysis system, the analysis processing unit predicts the meniscus dynamics of the subject while the subject is walking from the feature values ​​of the image of the inside of the knee when the subject is standing still. (3) In the above-mentioned image analysis system, the analysis processing unit predicts the mechanical indices of the subject's knee joint from the features of the internal image of the knee when the subject is standing still, and predicts the amount of exercise appropriate for the subject's meniscus dynamics using the mechanical indices of the subject's knee joint. (4) In the above image analysis system, the features of the internal knee image include any of the following information: the amount of meniscus deviation, the tilt (orientation) of the meniscus, the degree of sagging (angle) of the medial collateral ligament, and the presence or absence of osteophytes. (5) In the above image analysis system, the analysis processing unit detects the tibia and meniscus related to the knee joint from the internal knee image, and predicts the amount of deviation or inclination of the meniscus based on the detection results. (6) In the image analysis system, the analysis processing unit detects the degree of slack in the medial collateral ligament from the image of the inside of the knee. (7) In the image analysis system, the analysis processing unit detects osteophytes of the tibia and femur related to the knee joint from the image of the inside of the knee. (8) In the image analysis system described above, the analysis processing unit includes a first analysis processing unit, a second analysis processing unit, and a third analysis processing unit. The first analysis processing unit divides the image of the internal knee of the subject when the subject is standing still into regions corresponding to tissues inside the knee. The second analysis processing unit extracts feature quantities of the image of the internal knee of the subject when the subject is standing still using the results of dividing the image into regions corresponding to tissues inside the knee. The third analysis processing unit predicts meniscus dynamics of the subject when the subject is walking from the feature quantities of the image of the internal knee of the subject when the subject is standing still. (9) In the above-mentioned image analysis system, the third analysis processing unit predicts the meniscus dynamics of the subject while walking and the mechanical indicators of the subject's knee joint from the features of the internal knee image of the subject when the subject is standing still, using a trained model for meniscus dynamics analysis. (10) In the above image analysis system, the analysis processing unit predicts the meniscus dynamics of the subject from the features of a specific knee internal image identified from multiple knee internal images acquired while the subject was walking. (11) The image analysis system includes an auxiliary orthosis that supports a probe, the probe having a surface that contacts the subject and is made of an elastic material, on the knee of the subject. The probe irradiates ultrasound waves in the direction of the subject and detects reflected waves of the ultrasound waves. The probe includes a specific surface that receives the reflected waves of the ultrasound waves, and the auxiliary orthosis supports the specific surface of the probe facing the knee of the subject when attached to the knee of the subject. (12) An assistive orthosis according to one aspect of the present invention is used in an image analysis system that predicts meniscus dynamics during walking of a subject based on features of an image of the inside of the knee of the subject taken using ultrasound while the subject is standing or walking. This assistive orthosis is an assistive orthosis that includes a support part that supports a probe on the knee of the subject that irradiates ultrasound and detects reflected waves of the ultrasound. (13) The above-mentioned assistive orthosis includes a first pad portion and a second pad portion. The first pad portion includes a first core portion made of an elastic material having a contact surface that contacts the knee portion of the subject, and a first belt for attaching to the knee portion. The second pad portion includes a second core portion made of an elastic material that is formed to be able to support the probe, and a second belt for attaching to the knee portion. The probe includes a specific surface that receives the reflected waves of the ultrasound. The contact surface of the first core portion is provided with an opening that allows the ultrasound of the probe to pass through. The second core portion is disposed over the first pad portion and maintains a state in which the specific surface of the probe is aligned with the position of the opening provided in the contact surface of the first pad portion. (14) In the above-described assistive orthosis, the first belt attaches the first core portion of the first pad portion to the knee of the subject. The second belt attaches the second core portion of the second pad portion, which is placed over the first pad portion, to the knee of the subject. The position of the opening of the contact surface relative to the knee of the subject is determined by the fixed position of the first pad portion. The second pad portion is disposed over the first pad portion and supports the probe at the position of the opening in the contact surface.

[0010] (15) An image analysis method according to one aspect of the present invention is an image analysis method that includes predicting meniscus dynamics when a subject is walking from features of an image of the inside of the knee of the subject while standing or walking, the image being obtained using ultrasound.

[0011] (16) A program according to one aspect of the present invention is a program for causing a computer to predict meniscus dynamics when a subject is walking from features of an internal knee image obtained using ultrasound while the subject is standing or walking. [Effects of the Invention]

[0012] According to the present invention, the state of the knee joint during walking can be more easily obtained. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a schematic configuration diagram of an image analysis system according to an embodiment. [Figure 2] 1 is a schematic configuration diagram of an image analysis device according to an embodiment. [Figure 3] FIG. 1 is a diagram for explaining needs related to an image analysis system according to an embodiment. [Figure 4] FIG. 1 is a diagram for explaining needs related to an image analysis system according to an embodiment. [Figure 5] FIG. 1 is a diagram for explaining the relationship between mechanical load on a joint and cartilage. [Figure 6] FIG. 10 is a diagram for explaining an evaluation method using knee joint varus moment as an example of a comparative example. [Figure 7] FIG. 10 is a diagram for explaining a method of a comparative example. [Figure 8] FIG. 1 is a diagram illustrating the configuration of a knee joint as seen from the front and the hoop function of the meniscus. [Figure 9] 1 is a diagram for explaining the symptoms of knee OA using longitudinal cross-sectional views of the joint and images of the inside of the joint. [Figure 10] FIG. 10 is a diagram illustrating an example of an operation of applying a load to a knee joint. [Figure 11]FIG. 10 is a diagram for explaining the movement of the meniscus when a load is applied to the knee joint. [Figure 12] 10A and 10B are diagrams for explaining the relationship between the degree of cartilage damage in the knee joint and the detection results using various measurement indexes. [Figure 13] FIG. 1 is a diagram for explaining a mechanical indicator shown in an embodiment. [Figure 14A] FIG. 10 is a first diagram for explaining an example in which an appropriate range of exercise amount is derived using a dynamics index according to the embodiment. [Figure 14B] FIG. 10 is a second diagram for explaining an example in which an appropriate range of exercise amount is derived using a dynamics index according to the embodiment. [Figure 15A] FIG. 1 is a first diagram for explaining the relationship between management of exercise amount and knee pain according to an embodiment. [Figure 15B] FIG. 2 is a second diagram for explaining the relationship between management of the amount of exercise and knee pain according to the embodiment. [Figure 16] 10 is a flowchart of an image analysis process according to an embodiment. [Figure 17] 10A and 10B are diagrams for explaining the amount of deviation of the meniscus according to the embodiment. [Figure 18] FIG. 10 is a diagram for explaining the orientation of the meniscus according to the embodiment. [Figure 19] FIG. 10 is a diagram for explaining slack in the medial collateral ligament according to the embodiment. [Figure 20] FIG. 10 is a diagram for explaining a bone spur according to an embodiment. [Figure 21] FIG. 1 is a diagram for explaining the relationship between deviation behavior and related factors in an embodiment. [Figure 22] FIG. 1 is a diagram for explaining the relationship between a mechanical index and a factor related thereto according to an embodiment. [Figure 23A] FIG. 10 is a diagram for explaining an image recognition process according to an embodiment. [Figure 23B] 10 is a first diagram for explaining the detection accuracy of the meniscus and the like according to the embodiment. FIG. [Figure 23C] FIG. 10 is a second diagram for explaining the detection accuracy of the meniscus and the like according to the embodiment. [Figure 24A] 10A and 10B are diagrams for explaining a trained model for estimating meniscus prolapse dynamics and indexes according to an embodiment. [Figure 24B] FIG. 10 is a diagram illustrating another example of a trained model for estimating meniscus prolapse dynamics and indexes according to an embodiment. [Figure 24C] FIG. 1 is a first diagram for explaining the accuracy of a trained model according to an embodiment. [Figure 24D] FIG. 2 is a second diagram for explaining the accuracy of the trained model according to the embodiment. [Figure 25] This is a diagram to explain the amount of movement of the meniscus during one step. [Figure 26] FIG. 10 is a diagram for explaining estimation of an index according to the embodiment. [Figure 27] 1 is a flowchart of an evaluation procedure according to an embodiment. [Figure 28] FIG. 1 is a diagram illustrating an example of a treatment process for knee osteoarthritis. [Figure 29] FIG. 10 is a top view of a probe attached to the knee using an auxiliary orthosis. [Figure 30] FIG. 10 is a cross-sectional view of the probe attached to the knee using an auxiliary orthosis. [Figure 31] FIG. 10 is a diagram for explaining the range inside the knee that can be detected by the attached probe. [Figure 32] FIG. [Figure 33] FIG. 1 is a diagram illustrating the configuration of an assistive device. [Figure 34] FIG. 10 is a top view of the assistive device with the first and second pad sections opened. [Figure 35] FIG. 10 is a diagram showing the configuration of the assistive device with the second pad section suspended from the first pad section. [Figure 36] FIG. 10 is a diagram illustrating the configuration of an auxiliary orthosis according to a modified example of the embodiment. [Figure 37A] FIG. 10 is an overhead view of the probe attached to the knee using the second assistive device. [Figure 37B] FIG. 10 is a schematic diagram of a second assistive device. [Figure 38] FIG. 10 is a diagram of an image taken during the wearing of the second assistive device. [Figure 39] 10A to 10C are diagrams showing examples of ultrasound images of the knee for each type of assistive device. [Figure 40] 10A to 10C are diagrams showing examples of ultrasound images of the knee for each type of assistive device. [Figure 41] FIG. 10 is a comparison diagram of dynamic quantities for different assistive devices. [Figure 42] FIG. 10 is a configuration diagram of an image analysis system according to a modified example of the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0014] The image analysis system according to the present invention will be described below with reference to the accompanying drawings, but the present invention is not limited to this embodiment.

[0015] In this embodiment, the meniscus dynamic amount refers to the distance that the meniscus moves radially from the knee joint during walking. The meniscus of the knee joint is located between the tibia and femur, and moves out of the joint due to stress from the tibia and femur, for example. The knee internal image is an image (knee joint internal image) generated based on the detection result by detecting the inside of the knee joint of the subject using the ultrasound diagnostic device 20 (ultrasound).

[0016] (Image analysis system configuration) FIG. 1 is a schematic diagram of an image analysis system 1 according to an embodiment. The image analysis system 1 shown in Fig. 1(a) uses ultrasound to acquire and analyze images of the inside of the knee of a subject when the subject is standing or walking. For example, an ultrasound diagnostic device 20 and its probe 30 may be used to acquire images of the inside of the knee of the subject when the subject is standing still. The image analysis device 10 in the image analysis system 1 predicts the dynamic amount (amount of deviation) of the subject's meniscus from the feature amounts of the images of the inside of the knee of the subject when the subject is standing still, acquired using ultrasound.

[0017] This image analysis system 1 includes, for example, an image analysis device 10 or a program (analysis processing unit) that can be executed by the image analysis device 10. The image analysis device 10 predicts the meniscus dynamics (amount of deviation) of a subject while walking based on the feature amount of an internal image of the knee of the subject when the subject is standing still, acquired using ultrasound. Figures 1(b) and 1(c) are modified examples of Figure 1(a). Details will be described later.

[0018] FIG. 2 is a schematic diagram of the configuration of the image analysis device 10 according to the embodiment. The image analyzing device 10 includes, for example, an acquisition unit 110, an analysis processing unit 120, a control unit 130, an output unit 140, a communication unit 150, and a storage unit 160.

[0019] The acquisition unit 110 acquires an image of the inside of the knee of the subject and temporarily adds it to the storage unit 160.

[0020] The control unit 130 controls the acquisition of internal knee images by the acquisition unit 110 and the display by the output unit 140 (display unit), thereby acquiring internal knee images of the subject user US and displaying them on the output unit 140 as moving or still images.

[0021] The analysis processing unit 120 performs a predetermined process on the internal knee image acquired by the acquisition unit 110. The predetermined process includes a process of predicting the meniscus dynamic amount (amount of deviation) of the subject user US while the subject is walking, based on the internal knee image of the user US. Details of this process will be described later.

[0022] The output unit 140 includes a display unit that displays various images such as the acquired internal knee image, images showing operation procedures, calculation results, etc. The display unit may be provided with an operation reception unit that can be used as a touch panel.

[0023] The communication unit 150 includes a communication interface for communicating with an external device via a network. For example, the ultrasound diagnostic device 20 may be configured as an external device of the image analyzing device 10, and data on the internal knee images may be collected by communicating with the ultrasound diagnostic device 20 via the communication unit 150.

[0024] The storage unit 160 includes, for example, a semiconductor memory element, and stores a program for operating the image analysis device 10, various variables used in the program, and calculation results. The image analysis device 10 may be configured as described above. For example, the image analysis device 10 may be an information device such as a smartphone, a tablet terminal device, or a personal computer. Each of the above functional units included in image analysis device 10 is realized by a processor such as a CPU (Central Processing Unit) executing a program. Note that some or all of the functional units such as acquisition unit 110, analysis processing unit 120, and control unit 130 may be hardware functional units such as an LSI (Large Scale Integration), an ASIC (Application Specific Integrated Circuit), or an FPGA (Field-Programmable Gate Array). Details of each unit in image analysis device 10 will be described later.

[0025] The processing by the analysis processing unit 120 is made up of a combination of several processes. The processing by the analysis processing unit 120 includes a process of predicting the meniscus dynamic amount (amount of deviation) while the subject is walking, based on the feature amount of the image of the inside of the knee when the subject is standing still. The processing by the analysis processing unit 120 includes a process of predicting a mechanical index (sometimes referred to as "index" in the following explanation) of the subject's knee joint from the features of an image of the inside of the knee when the subject is standing still, and a process of predicting the amount of exercise appropriate for the subject's meniscus dynamics (deviation amount) using the mechanical index (index) of the subject's knee joint. In this embodiment, the state of the knee joint of the subject is visualized using a predicted value of the meniscus dynamic amount (deviation amount) while the subject is walking and a mechanical index of the knee joint of the subject.

[0026] For example, the analysis processing unit 120 includes an area division processing unit 121, a feature extraction processing unit 122, and an estimation calculation processing unit 123. The area division processing unit 121, the feature extraction processing unit 122, and the estimation calculation processing unit 123 are examples of a first analysis processing unit, a second analysis processing unit, and a third analysis processing unit, respectively.

[0027] The region division processing unit 121 (first analysis processing unit) divides the image of the inside of the knee when the subject is standing still into regions corresponding to the tissues inside the knee. For example, the region division processing unit 121 includes a process of dividing the image into multiple regions (segmentation process). These regions are associated with the tissues inside the knee. Types of tissues inside the knee include the meniscus, tibia (skin), femur (skin), medial collateral ligament, osteophyte, skin, etc.

[0028] The feature extraction processing unit 122 (second analysis processing unit) extracts feature amounts of the image of the inside of the knee when the subject is standing still, using the results of dividing the image of the inside of the knee into areas associated with tissues inside the knee. The result of dividing the image into regions corresponding to the internal tissues of the knee corresponds to the result of the segmentation process described above. The feature quantities of the internal knee image may include any or all of the following: the amount of meniscus deviation, the tilt (orientation) of the meniscus, the degree (angle) of slack of the medial collateral ligament, and the presence or absence of osteophytes.

[0029] The estimation calculation processing unit 123 (third analysis processing unit) predicts the meniscus dynamic amount (deviation amount) when the subject is walking from the feature amount of the image of the inside of the knee when the subject is standing still.

[0030] Furthermore, the estimation calculation processing unit 123 (third analysis processing unit) may use a trained model to predict the meniscus dynamics (deviation amount) while the subject is walking and the mechanical index of the subject's knee joint from the features of the internal knee image when the subject is standing still.

[0031] As described above, the calculation processing of the analysis processing unit 120 can be configured to be divided into multiple stages.

[0032] For example, the feature extraction processing unit 122 of the analysis processing unit 120 may detect the tibia and meniscus related to the knee joint from the image of the inside of the knee, and predict the amount of meniscus deviation or inclination of the meniscus based on the detection results. For example, the feature extraction processing unit 122 of the analysis processing unit 120 may detect the medial collateral ligament from the image of the inside of the knee, and detect the degree of slack in the medial collateral ligament from the result. For example, the feature extraction processing unit 122 of the analysis processing unit 120 may detect osteophytes of the tibia and femur related to the knee joint from the image of the inside of the knee.

[0033] The probe 30 has a specific surface that outputs ultrasonic waves and receives the reflected waves. A typical probe 30 has a housing that also serves as a grip, and the specific surface is provided at the tip of the housing. A cord for connecting to the ultrasound diagnostic device 20 is provided at the other end of the housing of the probe 30. Note that a cordless (wireless) probe can also be used as the probe 30.

[0034] As described above, the image analysis system 1 shown in Fig. 1(a) is an example of a configuration in which the measurement position can be determined by a person DC grasping the gripping portion of the probe 30. Note that although the measurement position can be determined with practice, the configuration shown in Fig. 1(b) is recommended to make the measurement position more stable.

[0035] An image analysis system 1A shown in Fig. 1(b) is a modified example of the image analysis system 1. An assistive device 40 is added to the image analysis system 1. The assistive device 40 used in the image analysis system 1A will be described. The support device 40 has a surface formed of an elastic material that comes into contact with the subject, and is configured to support a probe 30 on the subject's knee, which emits ultrasound waves in the direction of the subject and detects the reflected waves of the ultrasound waves. The auxiliary orthosis 40 is configured to stably support the probe 30 with a specific surface facing the knee of the subject when attached to the knee of the subject. The auxiliary orthosis 40 is configured to be supported on the knee joint by multiple belts with hook-and-loop tape, but this is not shown in the figure. A more specific configuration example of the auxiliary orthosis 40 will be described later.

[0036] Examples other than the image analysis systems 1 and 1A shown in FIGS. 1(a) and 1(b) can also be selected. For example, a flat probe 30B may be used instead of the above-described probe 30. The flat probe 30B is a probe that does not have a housing that does not have a grip portion like the above-described probe 30. An example of such a flat probe is the image analysis system 1B shown in FIG. 1(c). The auxiliary orthosis 40B in the image analysis system 1B may be configured to support the flat probe 30B on the knee joint of the subject.

[0037] The inventors have found that the use of the image analysis system 1A and the like is suitable for the following cases. 3 and 4 are diagrams for explaining needs related to the image analysis system 1A of the embodiment. As shown in Figure 3, inflammation of the knee joint can cause a decline in the patient's motor function depending on its severity. For example, structural degeneration of bones, articular cartilage, muscles, etc. can occur, and if this worsens, it can lead to a decline in motor function. Decreased motor function can cause symptoms such as pain and swelling in the knee joint, and the accompanying functional decline (such as muscle weakness and loss of balance), which can lead to walking difficulties, decreased motor function, decreased mobility, and the need for nursing care. As shown in Figure 4, the type of information needed changes before and after the onset of knee OA. For example, before the onset of knee OA, it is desirable to detect the condition earlier. This can provide an opportunity for preventive treatment. Also, after the onset of knee OA, it is desirable to maintain an appropriate level of activity while preventing the progression of knee OA. The image analysis system 1A of the embodiment is applicable to both of these needs.

[0038] The estimation process will be described below with reference to FIGS.

[0039] (Principles of estimation processing) Relationship between mechanical loads on joints and cartilage: First, the relationship between the mechanical load on the joint and the cartilage will be explained. FIG. 5 is a diagram for explaining the relationship between the mechanical load on the joint and the cartilage. Each joint contains cartilage. Joints absorb or disperse mechanical stresses that occur there, reducing their impact. If the mechanical stress is to a certain extent, cartilage can be repaired through metabolism (cartilage homeostasis). However, if cartilage is subjected to repeated or excessive mechanical stresses, or if metabolism declines due to aging, the cartilage cannot keep up with the repair process (cartilage degeneration). In such a situation, the cartilage's functions cannot be fulfilled, and pain can occur. Therefore, if the effects of the above-mentioned factors that cause cartilage degeneration can be reduced, it is expected that the progression of cartilage degeneration can be inhibited.

[0040] Effects of mechanical loading on joints: An evaluation method using the knee joint varus moment will be exemplified as an example of a comparative example with reference to FIGS. 6 to 8, and an overview of the evaluation method will be described. 6 is a diagram illustrating an evaluation method using knee joint varus moment as an example of a comparative example. This evaluation method uses the magnitude of the moment acting on the knee joint (knee joint varus moment) as an index in relation to the vector of the floor reaction force acting on the foot and the distance to the center of the knee joint.

[0041] Figure 7 shows the results of evaluating the integral of the knee joint varus moment for young and elderly people with healthy knee joints, and for subjects with early symptoms of knee OA, using the method of the comparative example.

[0042] Figure 7(a) shows the evaluation results of the integral of the knee joint varus moment for healthy young people, elderly people, and subjects with early symptoms of knee OA (early stage OA). The magnitude of the integral of the knee joint varus moment varies within each group. However, the evaluation results show that there is little difference in the range of variation in the integral of the knee joint varus moment, making it difficult to distinguish between the two groups. These results demonstrate that the integral of the knee joint varus moment is not suitable for use in identifying subjects with early stage knee OA.

[0043] The following factors are presumed to be contributing to these results. For example, the force related to action and reaction shown in Figure 7(b) also occurs in the stepping force during walking shown in Figure 7(c). Therefore, if the stepping force during walking is large, the floor reaction force, which is the reaction force, will also be large. This can be presumed to be the reason why the integral of the knee joint varus moment is large in young people with strong leg strength. In addition, the varus moment is an index calculated only from extra-articular information, and has the limitation of not taking into account the individually different joint functions.

[0044] Next, the structure of the knee joint and the hoop function of the meniscus will be described with reference to FIG. Figure 8 is a diagram for explaining the structure of the knee joint as seen from the front and the hoop function of the meniscus. Figure 8(a) shows a model of the structure of the knee joint when extended as seen from the front. Figure 8(b) shows a model of a cross section of the knee joint when extended. The meniscus (MEN) surrounds the cartilage of the knee joint. When mechanical load is applied to the joint, stress is applied to the meniscus before the femoral (FEM) and tibial (TIB) cartilage come into contact, causing the meniscus to move. This function is called the "hoop function." This "hoop function" distributes the mechanical load on the knee joint, but at the same time, some of the mechanical load accumulates in the knee joint. For example, if the knee joint becomes fatigued and inflammation occurs in the area that supports the meniscus, symptoms such as excessive movement of the meniscus may occur.

[0045] With reference to FIG. 9, an example of a knee OA case will be described using a longitudinal cross-sectional view of the joint and an image of the inside of the joint. Fig. 9(a) shows an example of a healthy subject, and Fig. 9(b) shows an example of a knee OA patient. The femur is located above the joint, and the tibia is located below the joint. The femur and tibia appear dark, while the surrounding tissue appears bright. The wedge-shaped tissue sandwiched between the femur and tibia is the image of the meniscus. In the case of a healthy subject shown in Figure 9(a), the position of the meniscus does not significantly protrude from the line segment connecting the contours of the femur and tibia. In the case of a knee OA patient shown in Figure 9(b), it can be seen that the position of the meniscus significantly protrudes from the line segment connecting the contours of the femur and tibia. For comparison, Figure 9(c) shows the MRI image of the patient in Figure 9(b) above. Figure 9(c) shows that bone marrow edema has developed in the femur. It can be seen that the meniscus protrusion shown in Figure 9(b) was related to this symptom. In this way, when inflammation occurs in the knee joint, the hoop function becomes ineffective, and the meniscus (medial meniscus) deviates radially beyond the lateral edge of the tibia. It is desirable to perform the necessary treatment before the knee joint symptoms deteriorate excessively.

[0046] The movement of the meniscus will now be described with reference to Figures 10 and 11. FIG. 10 is a diagram illustrating an example of an operation in which a load is applied to the knee joint. FIG. 11 is a diagram illustrating the movement of the meniscus when a load is applied to the knee joint. As shown in FIG. 10, a load is applied to the knee joint by, for example, shifting the center of gravity or bending the knee joint. In this case, the meniscus may move radially of the knee joint. For example, when an inversion moment occurs in the knee joint as shown in FIG. 11(a), the balance between the left and right sides is disrupted and a compressive force acts on one of the menisci. At this time, as shown in FIG. 11(b), the shock-absorbing function (hoop function) of the meniscus allows the compressive force to be absorbed by the meniscus and surrounding tissues. In this case, if the shock absorption function (hoop function) of the meniscus works well, the position of the meniscus will not change significantly, as shown in Figure 11(c). On the other hand, if the shock absorption function (hoop function) of the meniscus does not work well, the position of the meniscus will change significantly, as shown in Figure 11(d). In other words, understanding not only the inversion moment, which is an extra-articular force, but also the joint function, which is intra-articular information, allows for a more detailed mechanical evaluation.

[0047] The magnitude of the load on the knee joint will now be explained. For example, three examples of task movements imposed on the knee joint of a subject will be given: standing still, walking, and climbing stairs. If the joint load during standing still is taken as the reference (reference value 1.0), the joint load during walking and climbing stairs may increase to values ​​such as 2.3 and 4.0, respectively. Therefore, the impact of walking, which imposes an intermediate joint load, will be explained below.

[0048] The relationship between the degree of cartilage damage in the knee joint and the detection results using various measurement indices will be described with reference to FIG. A method for distinguishing the condition of knee cartilage is known, which is divided into the following four stages:

[0049] Rank 1: The cartilage tissue directly above the lesion has softened Rank 2: There are some cracks in the cartilage, but the affected area is stable. Rank 3: The cartilage is cracked, but not completely detached from the base. Rank 4: The cartilage has been detached and the subchondral bone is exposed The higher the rank, the worse the cartilage condition, and the more likely joint deformity (knee OA) will develop and progress.

[0050] Invasive evaluation can identify the condition of knee cartilage and lead to early prevention of joint deformity. However, it is not realistic to apply the evaluation of invasive methods to patients with mild symptoms.

[0051] Here, we will explain a method for estimating the condition of the knee joint, which is classified as a non-invasive method that can be easily applied even to patients with mild symptoms. Various measurement indices used in existing estimation methods include (a) the femoro-tibial angle (FTA), (b) the varus moment, and (c) the amount of meniscus deviation in an unloaded static position (amount of deviation in the supine position) shown in Figure 12. The Roman numerals 1 to 3 on the horizontal axis of Figures 12(a) to 12(d) correspond to the above-mentioned ranks 1 to 3, with 0 indicating a normal stage. In the case of (a) FTA and (b) varus moment in Figure 12 above, no correlation was observed with the rank of cartilage damage in the knee joint.

[0052] In the case of the amount of deviation of the meniscus in the unloaded static position in Figure 12 (c) above, a statistical trend was confirmed in which a correlation was confirmed in which the greater the amount of deviation, the higher the rank of cartilage damage in the knee joint. However, it was confirmed that the results of the detection of the meniscus deviation amount in the static posture of subjects classified as rank 0 (normal) varied greatly when evaluated based on the (c) meniscus deviation amount in the unloaded static posture. As a result, the range of variation overlapped with the range of variation in the detection results of the meniscus deviation amount in the static posture of subjects classified as other ranks. Therefore, it can be seen that it is difficult to identify the condition of the subject's knee joint from the results of the (c) meniscus deviation amount in the unloaded static posture in Figure 12.

[0053] The evaluation results using the various measurement indexes of such comparative examples sometimes did not allow for proper identification of the rank of cartilage damage in the knee joint. If the patient's symptoms are severe, it is expected that there is a certain probability that they will be identified as being at rank 3. However, even if it is possible to identify a patient as being at rank 3, it has been difficult to use this information to determine whether to implement appropriate measures to prevent the patient from progressing to rank 3 at the rank 1 or 2 stage.

[0054] Therefore, in this embodiment, it is proposed to use the prolapse dynamics of the meniscus during walking (FIG. 12(d)) as an index of the state of the knee joint. As shown in Figure 12(d), there is a high correlation between the deviation dynamics of the meniscus during walking and the rank of cartilage damage in the knee joint. Even taking into account the variability in the results for each rank, it is possible to distinguish between three ranks, Rank 1 to Rank 3, of the deviation dynamics of the meniscus during walking, based on the magnitude of the deviation dynamics of the meniscus during walking.

[0055] This suggests that appropriate measures be taken to prevent the product from reaching rank 4 based on the three-level classification results (rank 1 to rank 3).

[0056] Next, we will explain how to use it in the early stages of knee OA. If you discover that you have developed knee OA, you will need to be more careful than when you are in the early stages of knee OA.

[0057] The right amount of exercise is: It is known that maintaining an appropriate amount of average daily physical activity can help prevent various diseases. This physical activity can be expressed in minutes of activity or steps. For example, if the goal is to avoid metabolic syndrome in healthy adults, a known guideline is "10,000 steps per day." Another known indicator for encouraging exercise is one that shows the relationship between the number of steps taken and the economic benefits of health. This indicates that walking more increases health value. For example, the cost of medical expenses saved by walking one step per day is estimated to be between 0.065 and 0.072 yen. In other words, walking 1,500 more steps is expected to reduce medical expenses by 35,000 yen per year. All of the above are indicators that strongly favor the recommendation of more exercise.

[0058] However, if people with knee joint inflammation (knee OA) put excessive strain on their knee joints, it can actually worsen the condition of the knee joint.If exercise aimed at preventing various diseases worsens the inflammation in the knee joint, it could have the opposite effect and make it impossible to exercise.

[0059] In this embodiment, the amount of exercise recommended is appropriate for the condition of the knee joint of each individual, within a range that does not worsen inflammation of the knee joint. In the following embodiment, the relationship between the condition of the knee joint and the amount of exercise will be described.

[0060] The exercise load on the knee joint is modeled using the number of steps as an index. The mechanical load caused by walking is repeatedly applied to the knee joint. Therefore, the mechanical load of each step is accumulated on the knee joint depending on the number of steps or amount of exercise. As a result, the accumulated mechanical load (accumulated load) can affect the health of the knee joint.

[0061] The shock absorbing function of a healthy knee joint will be described with reference to FIG. A healthy knee joint has the following shock absorbing function that utilizes the meniscus. The following example shows a case where an inversion moment is applied to the knee joint, causing it to bend in a bow-legged position. At this time, a load is placed on the knee joint. If the knee joint has good hoop function, the meniscus will move appropriately, distributing the load appropriately between the meniscus and cartilage, preventing excessive movement of the meniscus. On the other hand, if the hoop function of the knee joint is broken, the load is not distributed and the impact is accumulated in the knee joint.

[0062] An example of the mechanical indicator according to the embodiment will be described with reference to FIGS. FIG. 13 is a diagram for explaining the mechanical indicators shown in the embodiment. For comparison, Fig. 13(a) shows a comparative example in which a previous varus moment (knee joint varus moment) is applied. This comparative example corresponds to the example shown in Fig. 7(a) above.

[0063] FIG. 13(b) shows an example in which a mechanical index calculated using a previous varus moment is applied. We compare the cases where the same amount of varus moment is applied to the knee joint when the knee joint is healthy (normal) and when it has developed knee OA. As mentioned above, even if the varus moment is the same, the impact of the mechanical load accumulated in the knee joint in each case will be different. One of the phenomena that occurs when knee OA has developed is that the dynamic volume of the meniscus when a varus moment is applied increases. Therefore, we propose a new mechanical index that corrects the varus moment according to the health of the knee joint.

[0064] FIG. 14A is a diagram for explaining an example in which an appropriate range of exercise amount is derived using a dynamics index according to the embodiment.

[0065] The mechanical index according to the embodiment is called an "index value." The relationship between the "index value" and the "number of steps per foot" is shown in the following formula (1). The "number of steps per foot" is half the value of the so-called pedometer (the total number of steps per foot).

[0066] (Index value) x (Number of steps per foot) = Constant (1)

[0067] As shown in the above formula (1), the product of the "index value" and the number of steps for one foot is a constant of a predetermined value.

[0068] In the above formula (1), assuming that the knee joint is in a healthy (normal) state, if the "index value" is set to 0.25 and the number of steps per foot is set to 4000, the value of the constant that is the product of these will be exactly 1000.

[0069] Total load for a person with a healthy (normal) knee joint: (0.25) x (4000) = 1000

[0070] Next, we consider the case of people who have developed knee OA. The condition of people with knee OA varies from person to person. For example, let's calculate the upper limit (stepX) of exercise volume (number of steps per foot) for a person with an index value of 0.37.

[0071] Total load for people with knee OA: (0.37)x(stepX)=1000

[0072] When this equation is solved, the upper limit value (stepX) above becomes 2700. For this person, it is best to manage the amount of exercise so that the number of steps does not exceed 2,700 per leg and 5,400 for both legs. (However, as will be discussed later, it is not enough to avoid exceeding 5,400 steps; not exercising at all can also lead to worsening knee OA.)

[0073] Another example of a mechanical index is shown in Figure 14B. The vertical axis of Figure 14B(a) is the same index value as in Figure 14A. In contrast, the vertical axis of Figure 14B(a) is the value obtained by multiplying the index value by the subject's weight. Since the subject's weight is closely related to the impact on the knee when walking, the index value x weight is used as the mechanical index, and the relationship between the "index value" and the "number of steps per foot" is defined by the following equation (1').

[0074] (Index value) x (Weight) x (Number of steps per foot) = Constant (1´)

[0075] In Figure 14B(a), the "index value" of the upper limit of normal values ​​is 0.32. If the number of steps per foot is set to 7000, the value of formula (1) becomes 2240. Also, as an example, if the weight is set to 59 kg, the value of formula (1') becomes 0.32 x 59 x 3500 = 66080.

[0076] In contrast, in equation (1'), if the "index value x weight" of a knee OA patient is 24.7, the number of steps per leg for this subject will be 66080 ÷ 24.7 ≒ 2675 steps. In this way, by using the proposed dynamic index ("index value" or "index value" x "body weight"), it is possible to derive the upper limit of the amount of exercise that is appropriate for each individual.

[0077] It was confirmed that managing exercise volume using an upper limit (appropriate number of steps) based on a mechanical index ("index value") was effective in alleviating knee pain. FIG. 15A is a diagram for explaining the relationship between management of the amount of exercise and knee pain according to the embodiment. Patients with knee OA were asked to complete a questionnaire about changes in knee pain at the time of their first visit and three months later, as well as the amount of exercise they did during that period. During this period, each patient received general treatment.

[0078] The survey results were divided into patients whose actual step count exceeded the upper limit (appropriate number of steps) based on the mechanical index ("index value") (excess group, n=7) and patients whose actual step count did not reach the upper limit (appropriate number of steps) (non-excess group, n=9). Responses from the excess group indicated that the level of pain had either remained the same or had increased. On the other hand, many of the non-excess group responded that the pain had eased. These results are shown in a graph, tabulating the amount of change in pain (the product of the degree of change and the number of people). The responses from the excess group showed a trend toward worsening, while the responses from the non-excess group revealed a trend toward improvement. The specific method for deriving the above dynamic index ("index value") will be described later.

[0079] FIG. 15B is a second diagram for explaining the relationship between management of the amount of exercise and knee pain according to the embodiment. In Figure 15A, we explained how to set an upper limit on the number of steps based on the index value to prevent knee pain from increasing due to excessive walking. However, on the other hand, not walking at all is also thought to impair knee function. Therefore, we define the "excess rate," which indicates the appropriateness of the number of steps, as shown in the following formula (1A), and the number of steps that falls within a specified range is considered to be the appropriate number of steps. Excess rate = (actual number of steps ÷ upper limit of number of steps) × 100 (1A)

[0080] When conservative walking therapy was performed for three months from the initial consultation, patients with a daily step count exceeding 50% (too little walking) or 100% (too much walking) were classified as inappropriate, while those with a daily step count exceeding 50% or 100% were classified as appropriate. The results are shown in Figure 15B. Figure 15B(a) shows the change in meniscus prolapse in the inappropriate group, and Figure 15B(b) shows the change in meniscus prolapse in the appropriate group. The vertical axis of Figures 15B(a) and 15B(b) represents the meniscus prolapse, and the horizontal axis, "before" represents the time of the initial consultation and "after" represents 3 months after the initial consultation. As shown in Figure 15B(a), the inappropriate group showed an increase in prolapse after three months of conservative therapy. As shown in Figure 15B(b), the appropriate group showed a decrease in prolapse after three months of conservative therapy. Furthermore, only 13% of the people in the inappropriate group felt a reduction in pain, while 50% of the people in the appropriate group acknowledged a reduction in pain. In this way, it was confirmed that by continuing to walk while keeping the number of steps within an appropriate range, the condition of knee OA can be improved.

[0081] General health needs: In recent years, efforts aimed at improving various social issues (SDGs) have been attracting attention in order to realize a sustainable world. This embodiment is believed to contribute to the achievement of the goals set out in the SDGs, such as "3. Good health and well-being for all," "8. Decent work and economic growth," and "9. Increasing infrastructure, inclusive and sustainable industrialization."

[0082] If the function of the knee joint is maintained, it will be possible to participate in activities that involve the desired amount of exercise, which can lead to an ``extension of healthy life expectancy.'' Furthermore, if we can contribute to supporting the health of older workers, they will be able to participate in economic activities, which is expected to increase the number of healthy older workers and help to fill the labor shortage. However, in order to maintain the function of the knee joint, there has been no method provided for collecting information to identify the condition of an individual's knee, or for making the collected information available without spending time on its analysis. Therefore, this embodiment proposes a method that enables the social implementation of technology for maintaining knee joint function. A more specific method will be described below.

[0083] FIG. 16 is a flowchart of the image analysis process according to the embodiment. Pre-learning of the inference model is performed using each functional unit of the image analysis device 10 according to the following procedure. (1-0) The control unit 130 prepares and calibrates the environment for measuring ultrasound (intra-articular information) (S10). (1-1) The acquisition unit 110 acquires an image of the inside of the knee of a subject in a stationary standing position using ultrasound (S11). (1-2) The analysis processing unit 120 (region division processing unit 121) divides the image of the inside of the knee in a static standing position into regions (S12). (1-3) The analysis processing unit 120 (feature extraction processing unit 122) reads the following parameters to be used from the results of region segmentation of the image of the inside of the knee in the static standing position (S13). (1-4) The acquisition unit 110 acquires an image of the inside of the knee of a walking subject using ultrasound (S14). (1-5) The analysis processing unit 120 (estimation calculation processing unit 123) estimates (determines) the amount of meniscus movement (amount of deviation) using the group of images of the inside of the knee during walking (S15). (1-6) The analysis processing unit 120 (estimation calculation processing unit 123) performs a three-dimensional analysis of the image of the subject wearing the marker and walking on the force plate to analyze the posture of the subject while walking (such as the angle of the knee), and calculates the knee varus moment from the analyzed posture of the subject and the measurement value of the force plate (S16). (1-7) The analysis processing unit 120 (estimation calculation processing unit 123) calculates a mechanical index by multiplying the integral value of the inversion moment by the meniscus dynamic amount (S17). (1-8) The analysis processing unit 120 uses the above results to create training data (S18). (1-9) The analysis processing unit 120 uses the above training data to perform a learning process for the inference model (S19). The above learning process is carried out until a predetermined performance is obtained, and then the learning process of the inference model is completed. This inference model can be used, for example, in the "trained model for estimating meniscus prolapse dynamics and indexes" described below.

[0084] FIG. 17 is a diagram for explaining the amount of deviation of the meniscus according to the embodiment. FIG. 18 is a diagram for explaining the orientation of the meniscus according to the embodiment. FIG. 19 is a diagram for explaining slack in the medial collateral ligament according to the embodiment. FIG. 20 is a diagram illustrating a bone spur according to the embodiment.

[0085] 17 to 20, an image of the femur is on the left side of the image and an image of the tibia is on the right side, with the femur extending leftward and the tibia extending rightward. Note that an image of the subject's skin is located near the upper side of the tomographic image, and the downward direction is toward the center of the knee joint.

[0086] Deviation amount As shown in Figure 17, the amount of meniscus deviation is determined based on an extension line from the image of the tibial cortex toward the femur. Figure 17(a) shows the original image. Figure 17(b) shows the analysis results of the amount of meniscus deviation. For example, the amount of meniscus deviation is defined as the distance (unit: mm) to the innermost edge of the meniscus that appears above the extension line. In the example shown in Figure 17(b), the amount of meniscus deviation is 8.47 mm.

[0087] Meniscus orientation As shown in Figure 18, the cross section of the meniscus is considered to be a fan-shaped section that opens from the depth of the meniscus toward the outside of the knee (in the radial direction of the knee joint). The direction corresponding to half the angle (central angle) formed by this fan-shaped cross section of the meniscus is defined as the "orientation of the meniscus." FIG. 18(a) shows the reference for the "meniscus orientation." The reference direction is when moving radially from the center of the knee joint, and the direction approximately perpendicular to the extension direction of the tibia and femur is defined as "directly upward." More specific examples are shown. For example, FIGS. 18(b), 18(c), and 18(d) show the "meniscus orientation" in an example of a knee joint image that is the detection result. FIG. 18(b) shows an example of movement directly upward (upward) in the knee joint image, while FIGS. 18(c) and 18(d) show an example of movement due to the tibia side and an example of movement due to the femur side.

[0088] · Medial collateral ligament slack (angle of the medial collateral ligament): As shown in Figure 19, the medial collateral ligament may be pushed radially by the prolapsed meniscus, causing it to bend near the meniscus. The larger the angle of the medial collateral ligament at the point of flexion, the more slack there is in the medial collateral ligament. If the degree of meniscus prolapse is small, it can be assumed that the original function of the ligament is still effective. In this case, the slack (angle) of the medial collateral ligament becomes small. Figure 19(a) shows the original image. A more specific example is shown below. For example, Figures 19(b) and 19(c) show examples of the identification results of the slack (angle) of the medial collateral ligament in an example of a knee joint image from the detection results. Figure 19(b) shows an example of 155 degrees, and Figure 19(c) shows an example of 145 degrees.

[0089] ·Bone spurs: Osteophytes are bone spurs that occur on the femur and tibia. Osteophytes occur at the site of stimulation depending on the degree and frequency of stimulation that concentrates bone load. The larger the osteophyte, the more likely it is that the state in which bone load has been concentrated on the bone has continued. Figure 20 shows cross-sectional images of different symptoms. Figure 20(a) shows a case where there are no osteophytes on either the tibia or femur. Figure 20(b) shows a case where there are osteophytes on the tibia but no osteophytes on the femur. Figure 20(c) shows a case where there are no osteophytes on the tibia but there are osteophytes on the femur. Figure 20(d) shows a case where there are osteophytes on both the tibia and femur.

[0090] Below, we will summarize the relationship between each of the above factors and the deviation dynamics and index.

[0091] Referring to Figure 21, the relationship between deviation dynamics and related factors will be explained. FIG. 21 is a diagram for explaining the relationship between deviation behavior and related factors in the embodiment. As examples of the above factors, we will explain the relationship between the amount of prolapse, the direction of the meniscus, the slack (angle) of the medial collateral ligament, and the state of osteophytes. The number of specimens shown here is 100.

[0092] Figure 21(a) shows a two-dimensional scatter plot of the relationship between the amount of meniscus deviation (amount of deviation in standing position) and the dynamics of meniscus deviation. A straight line that linearly approximates the distribution above is added to this scatter plot. The coefficient of determination (R-squared value) based on the example results was 0.2646. These results indicate that there is a correlation between the amount of meniscus deviation and deviation dynamics during walking.

[0093] Figure 21(b) shows the relationship between the orientation of the meniscus and the dynamics of meniscus prolapse as a one-dimensional scatter plot, broken down by the general direction of the meniscus. The "orientation of the meniscus" refers to the general direction of the tilt of the symmetry axis along which the cross section of the meniscus is approximately symmetric. The orientation of the meniscus exemplified here is in three cases: when the distal radial end of the meniscus axis of interest of the knee is leaning toward the femur, when it is leaning toward the tibia, and when no lean toward the femur or tibia can be confirmed (hereinafter referred to as the "straight up direction"). For example, when the general direction of the meniscus is closer to the femur or directly upward, the deviation of the meniscus is concentrated within a specific range (0.6 to 2.1). In contrast, when the general direction of the meniscus is closer to the tibia, the deviation of the meniscus is dispersed over a wider range. From these results, it can be said that when the general direction of the meniscus is directly upward, there is little deviation within the specific range. These results indicate that there is a correlation between the orientation of the meniscus and the dynamics of meniscus prolapse.

[0094] Figure 21(c) shows a two-dimensional scatter plot of the relationship between the slack (angle) of the medial collateral ligament and the prolapse dynamics of the meniscus. A straight line that linearly approximates the distribution above is added to this scatter plot. The coefficient of determination (R-squared value) based on the example results was 0.1472. These results indicate that there is a negative correlation between the slack (angle) of the medial collateral ligament and the prolapse dynamics of the meniscus.

[0095] Figure 21(d) shows the relationship between the state of osteophytes and the dynamics of meniscus prolapse in a one-dimensional scatter plot for each state of osteophytes. The four states of osteophytes shown are those present in both the femur and tibia, those present only in the femur, those present only in the tibia, and those present in neither the femur nor the tibia. For example, when osteophytes are present in both the femur and tibia, the meniscus deviation kinetics is relatively widely distributed, ranging from 0.5 to over 3. When osteophytes are present only in the femur, the meniscus deviation kinetics is concentrated within a specific range (ranging from 0.7 to 2.1). When osteophytes are present only in the tibia, the meniscus deviation kinetics is concentrated within a specific range (ranging from 1.0 to 1.9). When osteophytes are absent from either the femur or tibia, the meniscus deviation kinetics is concentrated within a range of 0.6 to 1.5. These results indicate that there is a correlation between the state of osteophytes and the dynamics of meniscus prolapse.

[0096] Referring to FIG. 22, the relationship between the dynamic index and the factors related thereto will be explained. FIG. 22 is a diagram for explaining the relationship between the mechanical indices and the factors related thereto according to the embodiment. As shown in Figure 22(a) to (d), the relationship between the mechanical indices and the above factors will be explained using examples of the factors, such as the amount of deviation, the orientation of the meniscus, the slack (angle) of the medial collateral ligament, and the state of osteophytes. The number of specimens shown here is 100.

[0097] Figure 22(a) shows a two-dimensional scatter plot of the relationship between meniscus deviation and mechanical indices. A straight line that linearly approximates the distribution is added to this scatter plot. The coefficient of determination (R-squared value) based on the example results was 0.2789. These results show that there is a correlation between the amount of meniscus deviation and mechanical indicators.

[0098] Figure 22(b) shows the relationship between the orientation of the meniscus and the mechanical indices in a one-dimensional scatter plot, broken down by the general direction of the meniscus. The general directions of the meniscus shown here are three: one closer to the femur, one closer to the tibia, and one facing relatively vertically upward (hereinafter referred to as the vertically upward direction). For example, when the general direction of the meniscus is closer to the femur or directly upward, the mechanical indices are concentrated within a specific range (from near 0 to over 0.5). In contrast, when the general direction of the meniscus is closer to the tibia, the mechanical indices are dispersed over a wide range from 0.1 to over 1. These results indicate that there is a correlation between the orientation of the meniscus and mechanical indicators.

[0099] Figure 22(c) shows a two-dimensional scatter plot of the relationship between the slack (angle) of the medial collateral ligament and mechanical indices. A straight line that linearly approximates the distribution above is added to this scatter plot. The coefficient of determination (R-squared value) based on the example results was 0.1462. These results indicate that there is a weak negative correlation between the slack (angle) of the medial collateral ligament and mechanical indicators.

[0100] Figure 22(d) shows the relationship between the state of osteophytes and the mechanical indices in a one-dimensional scatter plot, broken down by osteophyte state. The osteophyte states shown here are four: presence in both the femur and tibia, presence in the femur only, presence in the tibia only, and presence in neither the femur nor the tibia. For example, when osteophytes are present in both the femur and the tibia, the mechanical index is distributed relatively widely, from near 0 to values ​​exceeding 1.3. When osteophytes are present only in the femur, the mechanical index is concentrated within a specific range (from near 0 to 0.7). When osteophytes are present only in the tibia, the mechanical index is concentrated within a specific range (from near 0 to 0.5). When osteophytes are not present in either the femur or the tibia, the mechanical index is concentrated within a range from 0.1 to 0.4. These results showed that there was a correlation between the condition of osteophytes and mechanical indicators. In this way, based on the results of the analysis of case data to date, it is possible to further categorize the condition of an individual's knee by utilizing the correlation between the condition of osteophytes and mechanical indicators.

[0101] FIG. 23A is a diagram illustrating the image recognition process according to the embodiment. Fig. 23A(a) shows an example of an image to be analyzed. Fig. 23A(b) shows an example of training data (annotation image) when image recognition processing is performed using a trained image analysis model. Fig. 23A(c) shows an example of the results of image recognition processing using the trained image analysis model. In the annotation image of FIG. 23A(b), the positions (areas) of the tibia, femur, meniscus, medial collateral ligament, and osteophytes occurring in the tibia and femur, and their corresponding labels may be added. As shown in FIG. 23A(c), the above-mentioned parts can be identified.

[0102] With reference to FIG. 23B, the detection accuracy of the meniscus and the like in the above image recognition processing will be described. FIG. 23B is a diagram for explaining the detection accuracy of the meniscus and the like according to the embodiment. FIG. 23B(a) shows the results of image recognition of the meniscus according to the embodiment. In this image, the estimation results for each part are divided into regions. The estimated area of ​​the tibia is labeled ES_TIB, the estimated area of ​​the femur is labeled ES_FEM, the estimated area of ​​the meniscus is labeled ES_MEN, the estimated area of ​​the medial collateral ligament is labeled ES_LIG, and the estimated area of ​​the tibial osteophyte is labeled ES_TIBSPUR.

[0103] The method used to evaluate the meniscus detection results used this time is the f-value and the IOU (Intersection over Union) of each part. This IOU is calculated using the following formula (2) to quantify the results of comparing the range that includes the area of ​​the detection target in the reference image used as the basis in image recognition processing with the range of the extracted detection target that is extracted based on the detection image.

[0104]

number

[0105] The IOU shown in formula (2) is an example of a performance indicator for image recognition, etc. As shown in the formula above, this IOU is the ratio of the product of the area that should be the correct answer and the predicted area to the sum of the area that should be the correct answer and the predicted area. For example, if the area that should be the correct answer and the predicted area match perfectly, the IOU will be 1, and if there is no match at all, the IOU will be 0. If there is a slight deviation, the value will be around 0.8.

[0106] The f-value and IOU of each part in FIG. 23B(a) are as follows:

[0107] Evaluation index (f value): 0.73511 IOU for the whole image: 0.58278 Femoral osteophyte IOU: 0.35282 Ligament IOU: 0.34171 Meniscus IOU: 0.75106 Epidermal IOU: 0.68781 IOU of the tibia line: 0.64786 IOU for tibial osteophytes: 0.1

[0108] FIG. 23B(b) is an example of the region segmentation result (estimation result) by image recognition repeatedly performed under different conditions. The graph shown in Figure 23B(b) shows the relationship between the accuracy rate of the classification results and the history of repeated classification attempts under different conditions. Four broken lines are drawn in this graph. The first one (f_tr) is the f value (Dice coefficient) using the training dataset (train). The second (f_val) is the f value (Dice coefficient) using the validation dataset. The third (IoU_tr) is the IoU evaluation value using the training dataset (train). The fourth (IoU_val) is the IoU evaluation value using the validation dataset.

[0109] The overall evaluation value (f-measure) in FIG. 23B(b) is 0.73511, and the IoU is 0.58278. The IoU for the femoral spur, ligament, meniscus, epidermis, tibial line, and tibial spur in Figure 23B(a) and (b) are as follows: For example, the IOU for the meniscus was 0.73511, which shows that it was generally well estimated.

[0110] Figure 23B (c) and (d) show examples of region segmentation results (estimation results) obtained by image recognition of an image of the knee joint of a patient with knee OA who has a symptom of tilted meniscus.

[0111] FIG. 23B(c) is a detection example of a subject in the early stage of knee OA, and the f-value and IOU of each part are as follows.

[0112] Femoral osteophyte IOU: 1.0 Ligament IOU: 0.47846 Meniscus IOU: 0.78814 Epidermal IOU: 0.76722 IOU of the tibia line: 0.64292 IOU for tibial osteophytes: 5.4945e-10

[0113] In FIG. 23B(c), for example, it can be seen that the IOU of the meniscus is 0.76722, which is a generally good estimate.

[0114] FIG. 23B(d) shows an example of detection of a subject in an advanced stage of knee OA, and the f-values ​​and IOU of each part are as follows:

[0115] IOU for femoral osteophytes: 0.30498 Ligament IOU: 0.19878 Meniscus IOU: 0.68764 Epidermal IOU: 0.73171 IOU of the tibia line: 0.73171 IOU for tibial osteophytes: 1.0173e-10

[0116] The IOU of the meniscus in Figure 23B(d) is 0.68764, which is somewhat lower than the value in the above example. However, it can be seen that the estimation is generally satisfactory even when the image of the meniscus is distorted.

[0117] Figure 23C(a) shows an image of the vicinity of the meniscus, and Figures 23C(b) and 23C(c) show an example of the region segmentation results using the image recognition method described above. In the example of Figure 23C, the deviation amount in the standing position was analyzed to be 1.681124, the meniscus orientation was 3 (1 is directly above, 2 is toward the tibia, and 3 is toward the femur), the ligament slack was 144.138°, the osteophyte was 0 (0 is none, 1 is toward the tibia, 2 is toward the femur, and 3 is toward both the tibia and femur), and the tibial inclination angle was analyzed to be -2.20323°.

[0118] Referring to FIG. 24A, the relationship between meniscal prolapse dynamics and indices and associated factors will be described. FIG. 24A is a diagram for explaining a trained model for estimating the prolapse dynamics and index of the meniscus according to the embodiment.

[0119] In this embodiment, the prolapse dynamics and index of the subject's meniscus are estimated, and information on factors related to these is used as input information for this estimation process. These related factors may include, for example, the amount of meniscus prolapse, the orientation of the meniscus, the slack of the medial collateral ligament, and the state of the osteophyte. As explained above, there is a correlation between these factors and the dynamics of meniscus prolapse.

[0120] As described above, since there is a correlation between the meniscus prolapse dynamics and index and the factors related thereto, it is preferable to perform calculation processing to estimate the meniscus prolapse dynamics and index using information on the above factors.

[0121] For example, the machine learning model of the estimation calculation processing unit 123 may be configured as shown in FIG. 24A. The estimation calculation processing unit 123 is configured as a neural network including an intermediate layer 123M located after the input layer 123I, and an output layer 123O. The number of intermediate layers and the number of neurons in each layer may be determined appropriately as needed. The intermediate layer 123M is configured with, for example, three layers (three stages), including an intermediate layer 1231, an intermediate layer 1232, and an intermediate layer 1233. Each layer of the intermediate layer 123M includes, for example, a fully connected layer (Dense) 123MD and an activation function processing unit 123MA at the subsequent stage. An example of the activation function shown here is ReLU (Rectified Linear Unit). The type of activation function can be selected as appropriate. The output layer 123O includes a fully connected layer 123OD and a subsequent activation function processing unit 123OA. The activation function illustrated here is ReLU. The type of activation function can be selected appropriately. The output from the activation function processing unit 123OA becomes the output of the estimation calculation processing unit 123.

[0122] An example of the neuron 123MN included in the hidden layer 123M will now be described. The neuron 123MN is formulated as the following equations (3) and (4).

[0123] y' = ΣXiWi, W0 = 1, i = integers from 0 to n (3) y=f(y') (4)

[0124] In the above formula (3), Xi is the output value of each neuron in the previous stage, and Wi is the weighting coefficient. In the above formula (4), y' is the intermediate output value, and f(*) is the activation function.

[0125] The above is just an example, and the configuration of the machine learning model of the estimation calculation processing unit 123 is not limited to this. The number of neurons in the input layer 123I, the number of stages of the intermediate layer, the number of neurons 123MN provided in each stage, the connection conditions with the previous stage, the number of neurons in the output layer 123O, and the like of the machine learning model of the estimation calculation processing unit 123 may be changed to configurations other than those shown in the figure.

[0126] Furthermore, the machine learning model of the estimation calculation processing unit 123 is trained in advance using appropriate training data, and the calculation result corresponding to the data X input to the estimation calculation processing unit 123 is output.

[0127] Although the machine learning model of the estimation calculation processing unit 123 has been described, it may include machine learning models for purposes other than those mentioned above.

[0128] FIG. 24B shows another example of the configuration of the machine learning model of the estimation calculation processing unit 123. The machine learning model shown in Fig. 24A uses meniscus orientation, osteophyte, standing deviation amount, and ligament slack as explanatory variables, and meniscus deviation dynamics and index value as objective variables. In contrast, the machine learning model shown in Fig. 24B uses gender, age, height, weight, BMI (Body Mass Index), meniscus orientation, ligament angle, osteophyte classification, tibial tilt angle, and standing deviation amount as explanatory variables, and meniscus deviation dynamics and / or index value as objective variables.

[0129] Figure 24C(a) shows a comparison of deviation dynamics estimated by trained models constructed using various machine learning methods such as decision trees and neural networks, and deviation dynamics estimated by the trained model illustrated in Figure 24B, which is an ensemble of these methods. The rightmost graph in Figure 24C(a) ("Ensemble") shows the estimation results using the trained model illustrated in Figure 24B, and the remaining graphs show the estimation results using other learning methods. The root-mean-square error (RMSE) of the deviation dynamics estimated by the trained model in Figure 24B was 0.32 mm.

[0130] Figure 24C(b) shows a comparison of the index estimated by the trained model shown in Figure 24B with the index estimated by other machine learning models. The rightmost graph in Figure 24C(b) ("Ensemble") shows the estimation results using the trained model shown in Figure 24B, and the rest show the estimation results using other learning methods. The RMSE of the index estimated by the trained model in Figure 24B was 0.09.

[0131] Figure 24D(a) shows the distribution of deviation dynamics amounts by cartilage damage rank for knee OA estimated by the trained model illustrated in Figure 24B. The vertical axis of Figure 24D(a) shows deviation dynamics amounts, and the horizontal axis shows cartilage damage rank. This graph shows deviation dynamics amounts estimated by the trained model for people diagnosed as normal (0) or ranked in Roman numerals 1 to 4, and then tabulated and organized by rank (the same applies to Figure 24D(b) shown next). As shown in the figure, although errors may affect the judgment of rank, it was confirmed that it is possible to judge a person's cartilage damage rank from the deviation dynamics amounts estimated by the trained model.

[0132] Figure 24D(b) shows the distribution of index values ​​by cartilage damage rank for knee OA estimated by the trained model illustrated in Figure 24B. The vertical axis of Figure 24D(b) shows the index value, and the horizontal axis shows the cartilage damage rank. As shown in the figure, although errors may affect the rank judgment, it was confirmed that the index value estimated by the trained model can determine the cartilage damage rank of a person.

[0133] Currently, diagnoses of cartilage damage in knee OA involve invasive arthroscopic procedures. Furthermore, calculation of deviation dynamics and / or index values ​​takes 3 to 4 hours per case. In contrast, the trained model of this embodiment can estimate deviation dynamics and / or index values ​​in 3 to 5 minutes. By comparing these estimation results with the statistical data shown in Figures 24D(a) and 24D(b), for example, it becomes possible to quickly determine whether or not a subject has knee OA and recommend an appropriate walking distance.

[0134] Next, we will explain some examples of social implementation by showing several scenarios.

[0135] (Scenario SC1): As a scenario SC1, an example in which an existing probe 30 with a connection cable and an ultrasonic image processing device are used will be described. In this scenario SC1, the above-mentioned existing equipment that is already owned or available on the market can be used to detect the amount of meniscus prolapse, which increases the immediacy of starting actual use in response to the demands for social implementation. In this scenario SC1, the probe 30 needs to be placed at a predetermined position on the subject's knee.

[0136] (Scenario SC2): As scenario SC2, a case will be described in which, in addition to using an existing probe 30 with a connection cable and an ultrasound image processing device as in scenario SC1, an auxiliary device for the probe 30 is used to support the probe 30. This scenario SC2 is a modification of the above-mentioned scenario SC1. In this scenario SC2, the use of auxiliary equipment for the probe 30 simplifies the work of placing the probe 30, improves the reproducibility of measurements, and reduces the variability in detection results. When detecting the amount of meniscus deviation, the measurement can be performed with the probe 30 released. As a result, in the case of scenario SC2, in addition to achieving the same effect as scenario SC1, it is possible to detect the amount of deviation using a simple method by placing a probe 30 at a specified position on the subject's knee, thereby increasing the immediacy until actual use can begin, in response to the demands for social implementation. In the case of this scenario SC2, the probe 30 can be placed on the knee joint of the subject using an auxiliary device for the probe 30, and can also be used to directly measure deviation dynamics during walking, etc.

[0137] (Scenario SC3): Scenario SC3 will be described in which a flat probe 30B with a connection cable is used instead of the probe 30 equivalent to that in scenario SC1. Many of the probes available on the market have a gripping portion for gripping the probe, which is provided between the ultrasonic signal transmission surface and the connection cable. The flat probe 30B with a connection cable described above has a form suitable for detecting the state of knee OA. The flat probe 30B with a connection cable proposed here does not have the above-mentioned gripping portion, and the connection cable is provided so as to extend in a direction along the transmission surface of the ultrasonic signal. In the case of this scenario SC3, when the subject moves the leg opposite to the leg to be detected, the leg can be detected without interfering with the grip portion or connection cord of the flat probe 30B. This flat probe 30B with a connection cable can also be used to detect deviations in a static standing position, and is also suitable for directly measuring deviation dynamics during walking and the like.

[0138] (Generating a trained model) The trained model used to estimate meniscus prolapse dynamics and indexes based on each factor in this embodiment is generated as follows.

[0139] First, a probe is used to detect deviations occurring during walking of a subject, and a detected value of the deviations of the subject is identified. For example, when detecting deviation behavior of the subject, the existing probe with connection cable 30 or flat probe with connection cable 30B can be used. Furthermore, during this detection, the reproducibility of the detection results can be improved by using an auxiliary device. Note that the communication between each probe and the ultrasound diagnostic device may be performed by wireless communication, in which case the connecting cables can be omitted (see Fig. 37 and Fig. 38(b)).

[0140] FIG. 25 is a diagram for explaining the amount of movement of the meniscus during one step (departure dynamics). The images shown in Figure 25(a) are extracted intermittently from images of the inside of the knee joint acquired while walking, and are arranged in chronological order starting from the left. For example, 25 frames of images are included for one step of the knee joint internal image. The amount of deviation of the meniscus at each timing is detected from each of the images of the inside of the knee joint for one step. The amount of deviation of the meniscus in each image of the inside of the knee joint is indicated by an arrow in the image.

[0141] The graph shown in Figure 25(b) shows the deviation amount of the meniscus at each timing during one walking cycle. For example, as shown in the figure, the deviation amount changes during one cycle, and by subtracting the minimum deviation amount from the instantaneous deviation amount, the dynamic amount of the meniscus at that timing can be obtained. The maximum value among the changes in the dynamic amount is defined as the "dynamic amount (deviation dynamics)" of this step.

[0142] FIG. 26 is a diagram illustrating the estimation of the index according to the embodiment. As shown in Fig. 26(a), images of the subject walking are acquired, and the posture of the subject at each timing is detected. The varus moment of the knee joint is obtained from the posture of the subject at each timing. For example, the positions of the ankles, knees, hips, etc. are obtained from the positions of markers attached to the subject's joints, etc. The floor reaction force is detected from the detection results of a pressure sensor attached to a floor mat where the subject walks, and the knee joint varus moment at each timing is obtained from these detection results.

[0143] The graph in Figure 26(b) shows the varus moment of the knee joint at each timing during one step. In this embodiment, the varus moment shown in this graph is integrated over the period of one step. The result of this integration is called the "integral value of the varus moment." This resulting value corresponds to the mechanical load of one step.

[0144] From the above results, the "index value" per step is calculated using the following formula (5).

[0145] "Index value" = "dynamic amount" x "integral value of varus moment" (5)

[0146] For example, the "index value" of a healthy adult and the "index value" of a person with knee OA are estimated using the following formulas (6) and (7).

[0147] "Index value" for healthy adults =(1.0mm)x(0.21Nms / kg)=0.210 (6)

[0148] "Index value" of patients with knee OA =(1.7mm)x(0.24Nms / kg)=0.408 (7)

[0149] As described above, it was confirmed that there are differences in meniscus deviation dynamics and indices between healthy adults and those with knee OA.

[0150] Therefore, in this embodiment, a more specific evaluation procedure will be described.

[0151] FIG. 27 is a flowchart of an evaluation procedure according to the embodiment. Following the procedure below, each functional unit of the image analysis device 10 performs processing for evaluation using an inference model according to the following evaluation procedure.

[0152] <Evaluation Procedure> (2-0) The control unit 130 prepares and calibrates the environment for measuring ultrasound (intra-articular information) (S20). (2-1) The acquisition unit 110 acquires an image of the inside of the knee of a subject in a stationary standing position (standing position) using ultrasound (S21). (2-2) The analysis processing unit 120 (estimation calculation processing unit 123) estimates the meniscus dynamic amount (deviation amount) and mechanical index using the trained inference model (S22). (2-3) The analysis processing unit 120 (estimation calculation processing unit 123) determines a guideline for the amount of exercise (S23). The above evaluation procedure makes it possible to more easily obtain the state of the knee joint during walking.

[0153] For example, by using ultrasound to obtain images of the inside of the knee of a subject in a static standing position (standing), it is possible to estimate the meniscus dynamics (amount of deviation) and mechanical indices during walking using subsequent analytical processing. In this case, it is not necessary to take the time to sequentially detect the state of the knee joint while walking.

[0154] In addition, as the internal knee image for evaluation in the above evaluation procedure, an internal knee image of the subject while walking can be used instead of an internal knee image of the subject while standing still (standing). The internal knee image of the subject while walking may be any of so-called moving images, semi-moving images with frame skipping, or still images. For example, in the case of moving images and semi-moving images, one or more specific internal knee images are identified from multiple internal knee images acquired while the subject is walking, etc. By identifying this specific internal knee image so that it is relatively close to the posture of standing still, the above-mentioned trained inference model can be used. In this way, the analysis processing unit 120 may predict the meniscus dynamic amount of the subject from the feature amount of a specific knee internal image identified from among multiple knee internal images acquired while the subject is walking.

[0155] Furthermore, in the case of an obese person, where still images are unclear, it is advisable to convert a video of the person walking into a cine image and use it. The analysis processing unit 120 automatically calculates a dynamic value based on the difference between the minimum and maximum deviation amounts of the meniscus during walking from the video of the person walking (cine image). In the case of this analysis method, it is advisable to apply the region segmentation processing by the region segmentation processing unit 121 (first analysis processing unit) described above. In this case, a trained model for extracting the meniscus region can be used. By adding an analytical method for deriving such dynamic values ​​to the algorithm for predicting the "index value," the accuracy of the analytical process can be improved.

[0156] Next, an example of application of the evaluation procedure according to this embodiment to treatment of knee osteoarthritis will be described. As shown in Figure 28, if a patient meets certain criteria, such as being over 35 years old, suffering from knee OA, and experiencing pain for more than one month, they undergo X-rays, MRI scans, and other tests and begin three months of conservative treatment. After three months of conservative treatment, a decision is made as to whether to continue conservative treatment or consider surgery. Traditionally, this decision is made based on whether the patient's pain has improved since the three-month conservative treatment. However, because the patient's pain varies depending on the day of the week and the time of day, deciding whether to continue conservative treatment or consider surgery based on the patient's pain is problematic. Therefore, after three months of conservative treatment, ultrasound images of the inside of the knee are acquired while the patient is walking using the evaluation procedure described above. One of the acquired ultrasound images is selected (or multiple images may be selected). The selected ultrasound image is then processed to calculate the meniscus orientation, ligament angle, osteophyte classification, tibial tilt angle, and deviation amount during standing. These are then input into the trained model shown in Figure 24A or 24B to obtain deviation dynamics and index values.

[0157] The subject's knee OA cartilage damage level is then assessed based on the graphs in Figures 24D(a) and 24(b). The step count, i.e., the upper limit for the number of steps required to maintain knee function, is calculated using Equation (1) and Equation (1'), and the appropriate step count range (50% ≤ exceedance rate ≤ 100%) is determined using Equation (1A). Data on the relationship between step count and potentially preventable diseases, such as those provided on the Health and Longevity Network (https: / / www.tyojyu.or.jp / net / kenkou-tyoju / rouka-yobou / haya-aruki.html), can then be used to identify the range of diseases that can be prevented by an appropriate step count based on deviation behavior and index values. Based on the preventable diseases, a decision can be made to continue conservative treatment or consider surgery. For example, if the appropriate step count for the knee condition is approximately 4,000 steps, the range of preventable diseases would be "bedridden" or "depression." In this case, because the number of diseases that can be prevented in the future by continuing conservative treatment is limited, a decision can be made to consider surgery. Alternatively, if the appropriate number of steps according to the condition of the knee is about 10,000 steps, walking (conservative treatment) can prevent many diseases, so it may be decided to continue the conservative treatment.

[0158] The analysis processing unit 120 may automatically perform the following steps: determining the rank of cartilage damage in knee OA based on the deviation behavior, index value, and the data illustrated in FIG. 24D; calculating the number of steps using Equation (1) or Equation (1'); and identifying the range of diseases that can be prevented based on the calculated number of steps. For example, even if a patient feels that their pain has significantly decreased after three months of conservative treatment, surgery may be considered if their upper step limit is lower than before the three months of conservative treatment. Conversely, even if a patient feels that their pain has not significantly improved after three months of conservative treatment, if their upper step limit is higher than before the three months of conservative treatment or if the range of diseases that can be prevented by the upper step limit is sufficient, it may be decided to continue conservative treatment. This allows for the determination of a treatment plan based not only on the patient's pain but also on objective indicators such as the patient's deviation behavior, index value, upper step limit, and the range of diseases that can be prevented by the number of steps.

[0159] (Assistive devices) Next, with reference to Figures 29-35, the auxiliary equipment used during measurement will be described.

[0160] FIG. 29 is a perspective view of the probe 30 attached to the knee using the assistive device 40. FIG. 30 is a cross-sectional view of the probe 30 attached to the knee using the assistive device 40. FIG. 31 is a diagram for explaining the range inside the knee that can be detected by the attached probe 30. FIG. 32 is a perspective view of the auxiliary orthosis 40. As shown in FIG. FIG. 33 is a diagram showing the configuration of the assistive device 40. FIG. 34 is a bird's-eye view of the assistive orthosis 40 with the first pad section 41 and the second pad section 42 opened. FIG. 35 is a diagram showing the configuration of the assistive orthosis 40 in a state where the second pad section 42 is suspended from the first pad section 41. As shown in FIG.

[0161] The probe 30 shown below includes a specific surface that receives reflected ultrasound waves. The probe 30 is supported by an auxiliary orthosis, which will be described later, so that the normal direction (Xs direction) of this particular plane faces the knee of the subject during measurement by the image analysis system 1A.

[0162] The support orthosis 40 is used during measurement by the image analysis system 1A, and is attached to the knee of the wearer together with the probe 30.

[0163] The auxiliary orthosis 40 is formed with at least a contact surface 411F that comes into contact with the knee of the wearer (subject). The auxiliary orthosis 40 functions as a flange for supporting the probe 30 of the ultrasound diagnostic device 20 on the knee of the subject. This reduces fluctuations in the angle of the probe 30 relative to the skin.

[0164] For example, the support device 40 includes a first pad portion 41 and a second pad portion 42.

[0165] The first pad portion 41 has a first core portion 411 , a first base portion 412 , a first rib portion 413 , and a first belt 51 . The first core portion 411 and the first base portion 412 are made of, for example, an elastic material. The first core portion 411 has a contact surface 411F that comes into contact with the knee of the subject. The surface of first core portion 411 opposite contact surface 411F is joined to the first surface of first base portion 412. Two surfaces of first core portion 411 excluding the surface opposite contact surface 411F are joined to the first surface of first base portion 412 via first rib portions 413. This reinforces the bond between first core portion 411 and the first surface of first base portion 412. As a result of being formed as described above, contact surface 411F of first core portion 411 protrudes from the first surface of first base portion 412.

[0166] The first core portion 411 is formed in a substantially rectangular parallelepiped shape, and its extension direction is formed so as to be substantially the same as the extension direction of the tibia when attached. The extension direction of the contact surface 411F of the first core portion 411 also becomes substantially the same as the extension direction of the tibia when attached. The extension direction of the contact surface 411F is the Zs direction. The first core portion 411 is provided near the center in the width direction of the first surface along the extension direction of the first surface of the first base portion 412. A through hole for attaching the base end portion of the first belt 51 is provided in the width direction of the first surface of the first base portion 412.

[0167] The contact surface 411F of the first core portion 411 is provided with an opening 41H that accommodates the tip of the probe 30 and allows the ultrasonic waves of the probe 30 to pass through.

[0168] As described above, the first core portion 411 of the first pad portion 41 has a contact surface 411F that comes into contact with a part of the circumferential direction of the outer surface of the knee when the assistive orthosis 40 is worn on the knee of a subject. For example, as shown in Figures 30-31, the contact surface 411F of the first core part 411 is in contact with a portion of the circumference of the knee. The position of the contact surface 411F of the first core part 411 can be adjusted to a position that is not in contact with the patella (kneecap) within the knee joint, but is aligned with the position of a depression formed in the part closer to the inside of the thigh than the patella. In this way, although the area in which the first core part 411 is in contact with the circumference of the knee is limited, the first core part 411 is supported by the knee by the first belt 51 that surrounds the circumference of the knee.

[0169] 30 and 31, a through-hole is formed that connects the contact surface 411F of the first core portion 411 to the second surface of the first base portion 412. The probe 30 of the ultrasound diagnostic device 20 is placed in this through-hole. In this embodiment, the opening 41H of the through-hole is formed, for example, in a substantially rectangular shape. This makes it possible to detect an image of the inside of the knee using the probe 30 whose tip has a substantially rectangular cross-sectional shape.

[0170] For example, when a rectangular opening 41H is provided in the contact surface 411F of the first core portion 411, the size of the opening 41H should correspond to the thickness of the probe 30 and the width of the probe 30 in the arrangement direction of the ultrasonic vibrators. In particular, by matching the circumferential width of the knee portion of the opening 41H to the thickness of the probe 30 to be used, horizontal shaking of the probe 30 can be suppressed. The center of the width of the contact surface 411F of the first core portion 411 may be aligned with the center of the opening 41H. It is preferable to provide a predetermined distance between the widthwise end of contact surface 411F of first core portion 411 and the end of opening 41H. By increasing this distance, deformation of first core portion 411 can be suppressed. The opening 41H may be disposed above the center in the extension direction (Zs direction) of the contact surface 411F of the first core portion 411. A predetermined distance may be provided between the upper end of the contact surface 411F of the first core portion 411 in the extension direction and the upper end of the opening 41H.

[0171] An outer edge portion of a predetermined width may be provided between the widthwise end of the first core portion 411 and the widthwise end of the first surface of the first base portion 412. A fastening means for the first belt 51 is provided on the outer edge of the first base portion 412. The fastening means for the belt 51 will be described later.

[0172] The second pad portion 42 has a second core portion 421 , a second base portion 422 , a second rib portion 423 , and a second belt 52 . The second core portion 421 and the second base portion 422 are made of, for example, an elastic material. The second core portion 421 has a contact surface 421F that comes into contact with the second surface of the first base portion 412 of the first pad portion 41. The surface of second core portion 421 opposite contact surface 421F is joined to the first surface of second base portion 422. Two surfaces of second core portion 421 excluding the surface opposite contact surface 421F are joined to the first surface of second base portion 422 via second rib portions 423. This reinforces the bond between second core portion 421 and the first surface of second base portion 422. As a result of being formed as described above, contact surface 421F of second core portion 421 protrudes from the first surface of second base portion 422.

[0173] The second core portion 421 is formed in a substantially rectangular parallelepiped shape, and its extension direction is formed so as to be substantially the same as the extension direction of the tibia when attached. The extension direction of the contact surface 421F of the second core portion 421 also becomes substantially the same as the extension direction of the tibia (Zs direction) when attached. The extension direction of the contact surface 421F becomes the Zs direction. The second core portion 421 is provided near the center in the width direction of the first surface along the extension direction of the first surface of the second base portion 422. A through hole for attaching the base end portion of the second belt 52 is provided in the width direction of the first surface of the second base portion 422.

[0174] When attached, the second core portion 421 is placed on top of the first pad portion 41, and keeps a specific surface of the tip of the probe 30 aligned with the position of the opening 41H provided in the first core portion 411 of the first pad portion 41.

[0175] The second core portion 421 of the second pad portion 42 has a contact surface 421F that comes into contact with the first pad portion 41 when the assistive orthosis 40 is worn on the knee of the subject. For example, as shown in FIGS. 29-30 , the second core portion 421 comes into contact with, for example, the first pad portion 41. The position at which the second core portion 421 is arranged corresponds to the rear side of the position at which the first core portion 411 of the first pad portion 41 is provided.

[0176] 30, 32-35, etc., a through-hole opening 42H is provided that connects the contact surface 421F of the second core portion 421 to the second surface of the second base portion 422. The probe 30 of the ultrasound diagnostic device 20 is placed in this through-hole. In this embodiment, the through-hole opening 42H is formed, for example, in a substantially rectangular shape. A slit 42SL is provided in part of the periphery of the through-hole opening 42H, leading from the edge of the opening 42H to the outer periphery of the second pad portion 42. This slit 42SL is used when passing the cord of the probe 30 through the opening 42H of the second core portion 421. This makes it easy to attach the corded probe 30 to the ultrasound diagnostic device 20. When using a probe 30 that communicates wirelessly with the ultrasound diagnostic device 20, this slit 42SL may be omitted.

[0177] The width of contact surface 421F of second core portion 421 may be adjusted to the width of contact surface 421F of first core portion 411, for example. The size of the square opening 42H provided in the contact surface 421F of the second core portion 421 may be adjusted to correspond to the thickness of the probe 30 to be used and the shape of the grip portion of the probe 30. In particular, by adjusting the opening of the circumferential opening 42H of the knee portion to match the shape of the grip portion of the probe 30 to be used and making it smaller than the size of the tip of the probe 30, shaking of the probe 30 can be suppressed. When the first pad portion 41 and the second pad portion 42 are used in an overlapping state, it is necessary to align the positions of the openings in the first core portion 411 and the second core portion 421. The position of the opening in the second pad portion 42 may be determined based on the position of the opening in the second pad portion 42. The spacing around the opening in the second pad portion 42 may be determined based on the spacing around the opening in the first pad portion 41.

[0178] It is preferable to provide a gap of a predetermined dimension between the widthwise end of the convex portion of the second core portion 421 and the widthwise end of the second core portion 421. A fastening means for the belt 5 is provided on the outer edge provided at this gap. The fastening means for the belt 5 will be described later.

[0179] The first core portion 411 and the second core portion 421 are formed using, for example, a foam having elasticity and cushioning properties. Examples of the material that can be used include expanded polystyrene resin (polystyrene) and expanded polystyrene resin containing resin additives (organic compounds, inorganic chemicals, etc.). For example, strength can be increased by adding carbon fiber or the like as a resin additive.

[0180] First core portion 411 and second core portion 421 may each be formed by laminating plate-like members made of foam having a predetermined thickness. For example, first core portion 411 and second core portion 421 may be molded to have approximately the same size, which may improve convenience when first core portion 411 and second core portion 421 are used in a stacked state. The detailed shapes of first core portion 411 and second core portion 421 may be different from each other as will be explained below, and an example thereof will be explained below.

[0181] (belt support) As shown in the above-mentioned FIGS. 29, 30, 32, etc., the first pad portion 41 and the second pad portion 42 are supported on the knee portion by the belt 5. Each belt 5 has a length of approximately 2 / 3 the circumference of the knee. It is used by wrapping it around approximately half the circumference of the knee. A hook-and-loop fastener or the like is provided at the tip of each belt 5, allowing paired belts 5 to be fixed to each other at any position. The paired belts 5 are wrapped around the knee about once and fixed. Each belt 5 is made of a material such as resin or chloroprene rubber, and is preferably elastic.

[0182] The belt 5 includes a first belt 51 and a second belt 52 . The first belt 51 supports the first pad portion 41 on the knee portion. The second belt 52 supports the second pad portion 42 on the knee portion from above the first pad portion 41.

[0183] For example, the first belt 51 includes belts 511, 512, 513, 514, 515, and 516. The base ends of the belts 511, 513, and 515 are inserted into belt fixing means in a first outer edge portion provided on the first base portion 412 of the first pad portion 41, and are joined and fixed by sewing, adhesive, welding, etc. The base ends of the belts 512, 514, and 516 are inserted into belt fixing means in a second outer edge portion provided on the first base portion 412 of the first pad portion 41, and are joined and fixed by sewing, adhesive, welding, etc. The belts 511 and 512, the belts 513 and 514, and the belts 515 and 516 are each an example of a pair of belts.

[0184] Similarly, the second belt 52 includes belts 521, 522, 523, 524, 525, and 526. The base ends of the belts 521, 523, and 525 are inserted into belt fixing means at a first outer edge portion provided on the second base portion 422 of the second pad portion 42, and are joined and fixed by sewing, adhesive, welding, etc. The base ends of the belts 522, 524, and 526 are inserted into belt fixing means at a second outer edge portion provided on the second base portion 422 of the second pad portion 42, and are joined and fixed by sewing, adhesive, welding, etc. The belts 521 and 522, the belts 523 and 524, and the belts 525 and 526 are each an example of a pair of belts.

[0185] By tightening the first belt 51, the stress exerted by the first core part 411 on the skin of the knee increases, thereby increasing the mutual frictional force. This makes it possible to prevent the first core part 411 from slipping off the knee and to prevent the wearing position from shifting. Furthermore, by loosening the first belt 51, it becomes easier to adjust the position of the first core part 411 relative to the knee. Tightening the second belt 52 increases the frictional force between the second core portion 421 and the first core portion 411. This makes it possible to prevent the second core portion 421 from slipping off and shifting of the mounting position.

[0186] In this way, the auxiliary appliance 40 holds the probe 30 at the position of the opening of the contact surface 421F in a state in which the second pad portion 42 is pressed against the first pad portion 41 by the second belt 52.

[0187] (connected structure) The first pad portion 41 and the second pad portion 42 are preferably connected to each other at one end of their vertical ends and configured to be able to lock together at the other end. The first pad portion 41 and the second pad portion 42 are connected, for example, at their lower end (one end) by a hinge 44. The upper end (other end) of the first pad portion 41 and the second pad portion 42 are provided with latches 43 (fasteners) that can lock together, and by closing the latches 43 (fasteners) while the first pad portion 41 and the second pad portion 42 are overlapped, it is possible to suppress movement of the first pad portion 41 and the second pad portion 42 along the contact surface 421F. The round ring 43A of the latch 43 is provided on the first pad portion 41, and the receiving seat 43B is provided on the second pad portion .

[0188] The first belt 51 attaches the first core portion 411 of the first pad portion 41 to the knee of the subject. The second belt 52 is attached to the knee of the subject by placing the second core portion 421 of the second pad portion 42, which is placed over the first pad portion 41, over the first core portion 411. The position at which the first pad portion 41 is fixed determines the position of the opening of the contact surface 411F relative to the knee of the subject. The second pad portion 42 is disposed over the first pad portion 41 and supports the probe 30 at the position of the opening in the contact surface 411F. In this way, the auxiliary appliance 40 holds the probe 30 at the position of the opening of the contact surface 411F in a state in which the second pad portion 42 is pressed against the first pad portion 41 by the second belt 52.

[0189] (How to fasten Belt 5) First belt 51 of first pad portion 41: When the assistive orthosis 40 is placed alone and each belt of the first belt 51 is stretched away from the first pad portion 41, the belts 513 and 514 stretch along a direction (Ys direction) perpendicular to the stretching direction of the first core portion 411. In contrast, the tips of the belts 511 and 512 stretch in a direction approaching the belts 513 and 514. Similarly, the tips of the belts 515 and 516 stretch in a direction approaching the belts 513 and 514.

[0190] Second belt 52 of second pad portion 42: When the assistive orthosis 40 is placed alone and each belt of the second belt 52 is stretched away from the second pad portion 42, the belts 523 and 524 stretch along a direction (Ys direction) perpendicular to the stretching direction of the second core portion 421. In contrast, the tips of the belts 521 and 522 stretch in a direction away from the belts 523 and 524. Similarly, the tips of the belts 525 and 526 stretch in a direction away from the belts 523 and 524.

[0191] As described above, the belt 5 is formed so that the first belt 51 and the second belt 52 are attached in different directions. The above is a description of the case where the auxiliary orthosis 40 is placed alone, but when wrapping the belt 5 around the knee joint, different paths can be used around the knee joint. This allows the belt 5 to surround the knee joint.

[0192] According to the above embodiment, the image analysis system 1 (1A) makes it possible to more easily obtain the state of the knee joint when walking by predicting the meniscus dynamics when the subject is walking from the features of an image of the inside of the knee obtained using ultrasound while the subject is standing or walking.

[0193] For example, the image analysis system 1 (1A) acquires an image of the inside of the knee of a subject standing still using an ultrasound diagnostic device 20. Such an image analysis system 1 (1A) may include an analysis processing unit that predicts the amount of meniscus dynamics of the subject walking from the feature amount of the image of the inside of the knee of the subject standing still, acquired using ultrasound.

[0194] In the image analysis system 1A, an auxiliary device 40 may be used to stably attach the probe 30 of the ultrasound diagnostic device 20 to the subject. The assistive orthosis 40 of the embodiment is used in an image analysis system 1A that predicts meniscus dynamics during walking based on features of internal knee images acquired using ultrasound while the subject is standing or walking. The assistive orthosis 40 includes a support section, which facilitates supporting a probe on the subject's knee, which emits ultrasound and detects reflected waves of the ultrasound. This allows for more stable acquisition of internal knee images while the subject is standing still, making it easier to obtain the state of the knee joint during walking. If necessary, video of the internal knee during walking can be used to compensate for meniscus movement. For example, in obese subjects with thick epidermal fat, still ultrasound images may not be clearly visualized. To address this issue, slight movement (walking) of the subject can be used to induce changes in tissue dynamics. Even when still ultrasound images are unclear, the target tissue (meniscus) can be more clearly identified by using the internal knee video acquired during this process. Such an image analysis system 1 (1A) and assistive device 40 can be applied to both when the subject is standing and when walking. Therefore, the image analysis system 1 (1A) and assistive device 40 can provide more stable detection and obtain better information, even when capturing not only still images but also moving images. The image analysis system 1 (1A) uses this information in various analytical processes, making it possible to more easily obtain the state of the knee joint while walking.

[0195] (Other embodiments) The above describes in detail an embodiment of the present invention with reference to the drawings, but the specific configuration is not limited to this embodiment, and design changes and the like are also included within the scope that does not deviate from the gist of the present invention.

[0196] For example, Fig. 36 is a configuration diagram of a second pad section 42B of an auxiliary orthosis 40B according to a modified example of the embodiment. The second pad section 42B shown here differs from the second pad section 42 of the auxiliary orthosis 40 in the following points. By using the second base portion 422B of the second pad portion 42B and the adapter 421A, it is possible to use probes 30 with different shapes. For example, the second base portion 422B is commonly used regardless of the shape of the probe 30. For example, the adapter 421A can be replaced according to the shape of the probe 30, allowing the adapter suitable for the shape to be used. The adapter 421A with the probe 30 attached thereto is fitted into an opening provided in the second base portion 422B.

[0197] FIG. 37A is an overhead view of a probe attached to the knee using a second assistive orthosis. FIG. 37B is a view of assistive orthosis 40a as viewed from the direction of the arrow in FIG. 37A. The second assistive orthosis 40a described below is characterized by being durable, easy to wear, and capable of capturing accurate ultrasound images of the inside of the knee, particularly while walking. The assistive orthosis 40a is used during measurements using the image analysis system 1A and is attached to the wearer's knee together with the probe 30. The assistive orthosis 40a is formed with a contact surface 40f that comes into contact with the knee of the wearer (subject). The assistive orthosis 40a functions as a flange for supporting the probe 30 of the ultrasound diagnostic device 20 on the subject's knee.

[0198] The support orthosis 40a includes a first pad portion 41a and a second pad portion 42a. The probe 30 is set on the fixing member 43a, and the fixing member 43a is sandwiched between the first pad portion 41a and the second pad portion 42a and pressed against the knee. The fixing member 43a is then secured to the knee with the belts 51a, 52a, and 53a. The first pad portion 41a has a joining portion 414a (first joining portion) and a joining portion 415a (first joining portion) made of Velcro (registered trademark) or the like on a joining surface 417a with the second pad portion 42a, and a recess (mounting portion) 416a for mounting the fixing member 43a. Similarly, the second pad portion 42a has a joining portion 424a (second joining portion) and a joining portion 425a (second joining portion) made of Velcro (registered trademark) or the like on a joining surface 427a with the first pad portion 41a, and a recess (mounting portion) 426a for mounting the fixing member 43a. The first pad portion 41a and the second pad portion 42a are integrated by sandwiching the fixing member 43a and joining the joints 414a and 424a, and by joining the joints 415a and 425a. The fixing member 43a is placed and fixed in the space formed by the recesses 416a and 426a formed on the inside (joint surfaces 417a and 427a) of the first pad portion 41a and the second pad portion 42a. The size and shape of the fixing member 43a are designed so that it is placed in this space and fixed without shaking even while walking. For example, the fixing member 43a has a hole that can hold the probe 30. The probe 30 is fitted into this hole to fix the probe 30 to the fixing member 43a, and the fixing member 43a is fixed in the space formed by the recesses 416a and 426a. The size and shape of the probe 30 vary depending on the manufacturer, etc., and the fixing member 43a may be prepared for each type of probe 30 so that the probe 30 can be fixed.

[0199] The first pad portion 41a has a first fixing portion 411a, a second fixing portion 412a, and a third fixing portion 413a. The second pad portion 42a has a first fixing portion 421a, a second fixing portion 422a, and a third fixing portion 423a. When viewed from the side ( FIG. 37A ), the first fixing portion 411a, the first fixing portion 421a, the third fixing portion 413a, and the third fixing portion 423a are made of thin members, and the second fixing portion 412a and the second fixing portion 422a are substantially trapezoidal. The inclined surface 412a-1 of the second fixing portion 412a and the inclined surface 422a-1 of the second fixing portion 422a are rounded. By wrapping the belt 52a tightly around the rounded arc surface, the assistive orthosis 40a can be worn stably. The inside of the belt 52a may be treated to prevent slipping.

[0200] FIG. 38 shows an image of the support brace 40a being attached to the knee of a subject. For example, the contact surface 40f is positioned closer to the inside of the patella around the circumference of the knee, and the circumference of the knee is supported by the belts 51a-53a. Specifically, with the support brace 40a pressed against the knee, the belt 51a is wrapped around the first fixing portion 411a, the first fixing portion 421a, and the upper part of the knee above the knee, the belt 52a is wrapped around the second fixing portion 412a, the second fixing portion 422a, and the center of the knee, and the belt 53a is wrapped around the third fixing portion 413a, the third fixing portion 423a, and the lower part of the knee below the knee, thereby securing the support brace 40a to the knee. FIG. 39(a) shows an ultrasound image taken while the subject was walking, using a probe from a dedicated device attached to the subject. 39(b) shows an ultrasound image taken while the subject was walking, using the general-purpose probe 30 attached to the subject using the assistive device 40a. As shown in the figure, it was confirmed that by using the general-purpose probe 30 and the assistive device 40a, it is possible to obtain images similar to those obtained using dedicated equipment.

[0201] Furthermore, Figure 40(a) shows an ultrasound image taken while a subject was walking, with a probe made of dedicated equipment attached to the subject. Figure 40(b) shows an ultrasound image taken while a subject was walking, with a wired, general-purpose probe 30 attached to the subject using an auxiliary orthosis 40a. Figure 40(c) shows an ultrasound image taken while a subject was walking, with a wireless, general-purpose probe 30 attached to the subject using an auxiliary orthosis 40a. As shown in the figure, it was confirmed that, whether the wired, general-purpose probe 30 or the wireless, general-purpose probe 30 was used, by using the auxiliary orthosis 40a, it was possible to obtain images similar to those obtained with dedicated equipment.

[0202] FIG. 41(a) shows a comparison of meniscus dynamics analyzed from ultrasound images acquired using dedicated equipment and meniscus dynamics analyzed from ultrasound images acquired using a general-purpose probe 30 and an assistive device 40a. FIG. 41(b) shows a comparison of actual measurements of meniscus dynamics. As shown in the figure, by using a general-purpose probe 30 and an assistive device 40a, meniscus dynamics can be estimated with the same accuracy as when dedicated equipment is used. Furthermore, the estimated meniscus dynamics can be applied to S15 above and used as training data for a trained model. As described with reference to FIGS. 39 to 41, by using a general-purpose probe 30 and an assistive device 40a, ultrasound images of the inside of the knee during walking can be acquired without using expensive dedicated equipment, and training data on meniscus dynamics can be expanded.

[0203] Also, for example, FIG. 42 is a configuration diagram of an image analysis system according to a modified example of the embodiment. In the above embodiment, an example was given in which the image analysis device 10 and the ultrasound diagnostic device 20 are configured as separate devices in the image analysis system 1 (1A). However, instead of this, the image analysis devices 10C and 10D in the image analysis systems 1C and 1D may be configured as dedicated devices having the functions of the ultrasound diagnostic device 20. The image analysis device 10D in the image analysis system 1D is an example of a configuration that transfers image data and the like to the probe 30D via wireless communication.

[0204] The image analysis systems 1 to 1D shown in the above embodiments predict the meniscus dynamics of a subject based on feature values ​​of internal knee images taken while the subject is standing still, or internal knee images taken while the subject is standing still and while walking. In this case, in addition to predicting the meniscus dynamics during walking primarily using internal knee images taken while the subject is standing still, the amount of meniscus deviation (meniscus movement amount) used in the analysis may be detected and supplemented, if necessary, based on instantaneous values ​​of the amount of meniscus deviation at each timing obtained from internal knee images (moving images) taken while walking. To supplement the "meniscus deviation amount (meniscus movement amount)," the amount of meniscus deviation based on the internal knee images taken while standing still may be correlated with the amount of meniscus deviation based on internal knee images (moving images) taken while walking, and a value with a relatively small amount of meniscus deviation may be selected.

[0205] Furthermore, when predicting the meniscus dynamics of a subject from the feature values ​​of internal knee images during walking, the image analysis systems 1 to 1D shown in the above embodiments may sequentially detect the amount of meniscus prolapse at each timing obtained from internal knee images (moving images) during walking, and detect the amount of meniscus prolapse (amount of meniscus movement) based on the instantaneous value of the amount of meniscus prolapse at each timing. In this case, the amount of meniscus prolapse (amount of meniscus movement) can be identified from the detection results from internal knee images (moving images) during walking.

[0206] The image analysis systems 1 to 1D described above predict the meniscus dynamics of a subject from the feature amounts of an image of the inside of the knee of the subject when the subject is standing or walking, which image is acquired using ultrasound. The image analysis method in the image analysis system and a computer program that executes the corresponding processing can also realize the image analysis system described above. The image information used for analysis by the image analysis systems 1 to 1D may be digitized image data generated by the ultrasound diagnostic device 20, or may be a video signal output by the ultrasound diagnostic device 20. In the former case, the generation of the image by the ultrasound diagnostic device 20 (capturing) and the analysis by the image analysis device 10 may be performed at different times. In the latter case, the acquisition unit 110 of the image analysis device 10 may receive the video signal output from the ultrasound diagnostic device 20 and digitize it to generate image data. The image data used in the above analysis is preferably uncompressed image data (RAW image).

[0207] A part or all of the above-described embodiments can be described as, but not limited to, the following supplementary notes. (Addendum) (1) An image analysis system according to one embodiment includes: An analysis processing unit that predicts the dynamic amount of the meniscus and the mechanical index of the knee joint of a subject from the feature amount of an internal image of the knee of the subject when standing or walking, obtained using ultrasound. The image analysis system is provided with: (2) In the image analysis system described in (1) above, The analysis processing unit The meniscus dynamic amount while the subject is walking is predicted from the feature amount of the knee internal image when the subject is standing still or walking. (3) In the image analysis system described in (1) above, The analysis processing unit A mechanical index of the knee joint of the subject is predicted from the features of an image of the inside of the knee when the subject is standing still, and the mechanical index of the knee joint of the subject is used to predict the amount of exercise and / or the range of exercise that is appropriate for the dynamic amount of the meniscus of the subject. (4) In the image analysis system described in (2) above, The feature quantities of the internal knee image include any of the following information: the amount of meniscus deviation, the tilt (direction) of the meniscus, the degree (angle) of slack in the medial collateral ligament, and the presence or absence of osteophytes. (5) In the image analysis system described in (3) above, The analysis processing unit The tibia and meniscus related to the knee joint are detected from the image of the inside of the knee, and the amount of deviation of the meniscus or the inclination of the meniscus is predicted based on the detection result. (6) In the image analysis system described in (4), The analysis processing unit The degree of slack in the medial collateral ligament is detected from the internal knee image. (7) In the image analysis system described in (3) above, The analysis processing unit Osteophytes of the tibia and femur related to the knee joint are detected from the image of the inside of the knee. (8) In the image analysis system described in (1) above, The analysis processing unit a first analysis processing unit that divides the image of the inside of the knee of the subject when the subject is standing still into regions corresponding to tissues inside the knee; a second analysis processing unit that extracts feature amounts of the image of the inside of the knee when the subject is standing still using the result of dividing the image into regions corresponding to the tissues inside the knee; a third analysis processing unit that predicts a meniscus dynamic amount when the subject is walking based on a feature amount of an internal knee image of the subject when the subject is standing still; Equipped with. (9) In the image analysis system described in (1) above, The third analysis processing unit Using a trained model, the dynamics of the meniscus while the subject is walking and the mechanical index of the subject's knee joint are predicted from the features of an image of the subject's internal knee when the subject is standing still. (10) In the image analysis system described in (1) to (9), The analysis processing unit The subject's meniscus dynamics and the mechanical index may be predicted from the features of a specific knee internal image identified from multiple knee internal images acquired while the subject is walking. (11) In the image analysis system described in (3) to (10) above, The analysis processing unit Diseases that can be prevented by the amount of exercise may be predicted from the predicted value of the amount of exercise. (12) The image analysis system according to (1) above, an assistive device that supports a probe on the knee of the subject, the probe forming a surface that contacts the subject using an elastic material and that irradiates ultrasound waves in the direction of the subject and detects reflected waves of the ultrasound waves; Equipped with The probe includes a specific surface that receives the reflected ultrasonic waves, The auxiliary device is When attached to the knee of the subject, the particular surface of the probe is supported facing the knee of the subject. (13) An assistive orthosis according to one aspect of the present invention includes: An assistive device used in an image analysis system that predicts meniscus dynamics and mechanical indices when a subject is walking from features of an internal image of the knee of the subject when the subject is standing or walking, obtained using ultrasound, a support section that supports a probe on the knee of the subject, the probe forming a contact surface that is made of an elastic material and that irradiates ultrasonic waves in the direction of the subject and detects reflected waves of the ultrasonic waves, on the knee of the subject; It is an assistive device that includes: (14) In the assistive device described in (13) above, The support portion is a first pad portion having a first core portion made of an elastic material having a contact surface that contacts the knee portion of the subject and a first belt for attachment to the knee portion; a second pad portion having a second core portion made of an elastic material formed so as to be able to support the probe and a second belt for attachment to the knee portion; Equipped with The probe includes a specific surface that receives the reflected ultrasonic waves, an opening that transmits the ultrasonic waves of the probe is provided on the contact surface of the first core portion; The second core portion is disposed over the first pad portion and maintains a state in which the specific surface of the probe is aligned with the position of the opening in the contact surface. (15) In the assistive device described in (14), The first belt is configured to attach the first core portion of the first pad portion to the knee portion of the subject, The second belt is attached to the knee of the subject by attaching the second core portion of the second pad portion, which is arranged over the first pad portion, to the knee of the subject, The position at which the first pad portion is fixed determines the position of the opening of the contact surface relative to the knee of the subject, The second pad portion is The probe is supported at the position of the opening in the contact surface in a state where it is placed over the first pad portion. (16) In the assistive device described in (14), The first pad portion and the second pad portion are connected to each other at one end of their vertical ends and are configured to be able to be locked to each other at the other end. (17) In the assistive device described in (15) above, The probe is held at the position of the opening in the contact surface with the second pad portion pressed against the first pad portion by the second belt. (18) In the assistive device described in (13) above, The support portion is a first pad portion having a contact surface to be placed on the knee of the subject; A second pad portion having a contact surface that contacts the knee of the subject; a fixing member formed to be able to support the probe; Belt and Equipped with the first pad portion has a first joining portion that can be joined to a second pad portion; the second pad portion has a second joining portion that can be joined to the first pad portion, With the fixing member sandwiched between the first pad portion and the second pad portion, the first pad portion and the second pad portion are joined together using the first joining portion and the second joining portion; The belt attaches the joined first and second pad portions to the knees of the subject. (19) An image analysis method according to one embodiment includes: Predicting meniscus dynamics during walking of a subject from features of images of the inside of the knee obtained using ultrasound while the subject is standing or walking. The image analysis method includes: (20) In the image analysis method described in (19), An image of the inside of the knee of the subject while standing still, obtained by irradiating the ultrasound from a position closer to the inner thigh than the front of the knee joint, is used to predict the state of the joint while walking. (21) In the image analysis method described in (19), a first index indicating a meniscus dynamic amount during walking of the subject, the first index being obtained from a first internal knee image of the subject while the subject is standing still, obtained using ultrasound in a first period; a second index indicating a meniscus dynamic amount during walking of the subject, the second index being obtained from a second internal image of the knee of the subject while the subject is standing still, the second image being obtained using ultrasound during a second period; Using This includes showing changes in meniscus dynamics and mechanical indicators of the knee joint of the subject between the first period and the second period. (22) An image analysis method according to one embodiment includes: Analyzing the dynamic amount of the meniscus while the subject is walking from an image of the inside of the knee of the subject while walking, which image is acquired using ultrasound, and using the analyzed dynamic amount as training data for a trained model that predicts the dynamic amount of the meniscus while walking and mechanical indicators of the knee joint. The image analysis method includes: (23) In one aspect, the program Predicting meniscus dynamics and knee joint mechanical indices when a subject is walking from features of images of the inside of the knee obtained using ultrasound while the subject is standing or walking. It is a program for causing a computer to carry out the above. [Explanation of symbols]

[0208] 1, 1A, 1B, 1C, 1D...Image analysis system 10, 10C, 10D...Image analysis device 20...Ultrasound diagnostic equipment 30, 30D... probe 40, 40D... Assistive devices 5...Belt 51...1st Belt 52...Second Belt

Claims

1. an analysis processing unit that predicts the meniscus dynamics and mechanical indices of the knee joint of the subject from feature amounts of images of the inside of the knee of the subject when standing or walking, obtained using ultrasound; An image analysis system comprising:

2. The analysis processing unit predicting the meniscus dynamics amount while the subject is walking from the feature amount of the knee internal image when the subject is standing still; The image analysis system according to claim 1 .

3. The analysis processing unit predicting a mechanical index of the knee joint of the subject from a feature amount of an internal image of the knee of the subject when the subject is standing still or walking, and predicting an amount of exercise and / or a range of amount of exercise suitable for the meniscus dynamics of the subject using the mechanical index of the knee joint of the subject; The image analysis system according to claim 1 .

4. The feature amount of the knee internal image includes any of the following information: the amount of meniscus deviation, the tilt of the meniscus, the degree of slack in the medial collateral ligament, and the presence or absence of osteophytes. The image analysis system according to claim 2 .

5. The analysis processing unit detecting the tibia and meniscus related to the knee joint from the knee internal image, and predicting the amount of deviation or the inclination of the meniscus based on the detection result; The image analysis system according to claim 3 .

6. The analysis processing unit detecting a degree of slack in the medial collateral ligament from the internal knee image; The image analysis system according to claim 4 .

7. The analysis processing unit Detecting osteophytes of the tibia and femur related to the knee joint from the knee internal image. The image analysis system according to claim 3 .

8. The analysis processing unit a first analysis processing unit that divides the image of the inside of the knee of the subject when the subject is standing still into regions corresponding to tissues inside the knee; a second analysis processing unit that extracts feature amounts of the image of the inside of the knee when the subject is standing still, using the result of dividing the image into regions corresponding to tissues inside the knee; a third analysis processing unit that predicts a meniscus dynamic amount when the subject is walking based on a feature amount of an internal knee image of the subject when the subject is standing still; The image analysis system according to claim 1 , comprising:

9. The third analysis processing unit predicting, using a trained model, a meniscus dynamics amount during walking of the subject and a mechanical index of the knee joint of the subject from feature amounts of an internal image of the knee of the subject when the subject is standing still; The image analysis system according to claim 8 .

10. The analysis processing unit predicting the meniscus dynamics and the mechanical index of the subject from the feature amount of a specific knee internal image identified from a plurality of knee internal images acquired while the subject is walking; The image analysis system according to claim 1 .

11. The analysis processing unit predicting a disease that can be prevented by the amount of exercise based on the predicted value of the amount of exercise; The image analysis system according to claim 3 .

12. an assistive device that supports a probe on the knee of the subject, the probe forming a surface that contacts the subject using an elastic material and that irradiates ultrasound waves in the direction of the subject and detects reflected waves of the ultrasound waves; Equipped with The probe includes a specific surface that receives the reflected ultrasonic waves, The auxiliary device is supporting the specific surface of the probe facing the knee of the subject while the probe is attached to the knee of the subject; The image analysis system according to claim 1 .

13. An assistive device used in an image analysis system that predicts a meniscus dynamic amount and a knee joint mechanical index of a subject when walking from a feature amount of an internal image of the knee of the subject when standing or walking, obtained using ultrasound, a support part that supports, on the knee part of the subject, a probe that emits ultrasonic waves in the direction of the subject and detects reflected waves of the ultrasonic waves, and that has a contact surface that is made of an elastic material and that is in contact with the knee part of the subject; An assistive device comprising:

14. The support portion is a first pad portion having a first core portion made of an elastic material having a contact surface that contacts the knee portion of the subject and a first belt for attachment to the knee portion; a second pad portion having a second core portion made of an elastic material capable of supporting the probe and a second belt for attachment to the knee portion; Equipped with The probe includes a specific surface that receives the reflected ultrasonic waves, an opening that transmits the ultrasonic waves of the probe is provided on the contact surface of the first core portion; the second core portion is disposed over the first pad portion and maintains a state in which the specific surface of the probe is aligned with the position of the opening of the contact surface. The assistive device according to claim 13.

15. The first belt is configured to attach the first core portion of the first pad portion to the knee portion of the subject, The second belt is attached to the knee of the subject by attaching the second core portion of the second pad portion, which is arranged over the first pad portion, to the knee of the subject, The position at which the first pad portion is fixed determines the position of the opening of the contact surface relative to the knee of the subject, The second pad portion is the probe is supported at the position of the opening in the contact surface in a state where the probe is placed over the first pad portion; The assistive device according to claim 14.

16. The support portion is a first pad portion having a contact surface to be placed on the knee of the subject; a second pad portion having a contact surface that is to be placed on the knee of the subject; a fixing member formed to be able to support the probe; Belt and Equipped with the first pad portion has a first joining portion that can be joined to a second pad portion; the second pad portion has a second joining portion that can be joined to the first pad portion, With the fixing member sandwiched between the first pad portion and the second pad portion, the first pad portion and the second pad portion are joined together using the first joining portion and the second joining portion; The belt attaches the joined first pad portion and second pad portion to the knee portion of the subject. The assistive device according to claim 13.

17. Predicting the meniscus dynamics and knee joint mechanical indexes of a subject when walking from the feature quantities of internal knee images of the subject when standing or walking, obtained using ultrasound; An image analysis method comprising:

18. Analyzing a dynamic amount of the meniscus while the subject is walking from an image of the inside of the knee of the subject while walking, which image is acquired using ultrasound, and using the analyzed dynamic amount as training data for a trained model that predicts the meniscus dynamic amount while walking and the mechanical index. The image analysis method according to claim 17, comprising:

19. Predicting the meniscus dynamics and knee joint mechanical indexes of a subject when walking from the feature quantities of internal knee images of the subject when standing or walking, obtained using ultrasound; A program that causes a computer to perform the above.

Images