JP6966131B1 - Orthotic device discrimination system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an orthotic device discrimination system for discriminating a plurality of types of orthotic devices provided with a band with high accuracy. SOLUTION: The orthotic device discrimination system 10 discriminates the type of the orthotic device 50 from the image data 11 of the orthotic device 50 provided with a band, and analyzes the teacher data creation unit 20, the target data input / result display unit 30, and the analysis. It is provided with a unit 40. The analysis unit 40 extracts the feature amount expressing the contour, the shape and the number of the band of the discrimination target device 50 from the image data 11 transmitted from the target data input / result display unit 30, and registers them in advance stored in the library 41. Calculates and outputs the degree of similarity with the attached equipment. [Selection diagram] Fig. 1

Description

The present invention relates to an orthotic device discrimination system that discriminates a type of orthotic device from image data of a photograph of an orthotic device provided with a band.

Conventionally, patients with walking disabilities wear orthotics on their feet to assist walking and rehabilitate. Treatment equipment is relatively expensive, and if the so-called health insurance insured person who purchased the equipment applies for medical treatment expenses, the insurance will be applied at the discretion of the insurer and will be refunded as medical treatment expenses at a later date. You can receive it. When applying for medical expenses to the insurer, make sure that the estimate of the medical expenses applied for is the same as the one given to the insured (patient, etc.). Is becoming obligatory to attach.

The insurer should have a photo of the orthotic device attached to the application for medical treatment expenses and a receipt (name and type of orthotic device, etc., and the cost amount by breakdown) submitted with the application. It is necessary to confirm whether the model number of the orthotic device described in (including the specification part) is correct. However, there are a plurality of types of orthotic devices, and it is difficult to distinguish them because the contour of the entire orthotic device, the number of bands of the orthotic device, the position, and the like are different. In addition, most of the insurers who conduct examinations are clerical workers who do not have knowledge of orthotic devices, so the photos of the orthotic device catalogs registered with the Ministry of Health, Labor and Welfare are the same as the photos of the application form. Determining whether or not it is an orthotic device requires a great deal of labor, and the cost burden such as labor costs increases.

Therefore, in recent years, an image discrimination system using artificial intelligence (AI) has been used in various fields in order to determine whether the object to be discriminated is correct or what it is. The technique shown in Patent Document 1 is known as an example of such an image discrimination system using artificial intelligence.

The discrimination system of Patent Document 1 is a discrimination system for a target work in which foods such as seasonings and cooking ingredients are enclosed and transferred, and is an imaging unit, a control unit, and an arm robot. The imaging unit captures a plurality of target works randomly stacked, and the control unit transfers the capture target work detection function that detects the capture target work from the information captured by the imaging unit and the capture target work. It has an arm robot control function that controls the arm robot so that it can be mounted. From a plurality of target works randomly piled up, the target work is discriminated based on the information obtained by artificial intelligence (AI), and the target work is detected as the target work to be captured in consideration of the distance from the arm robot for transfer. The arm robot is controlled and captured / transferred according to the orientation and inclination of the work to be captured.

However, in the discrimination system of Patent Document 1, the target work is one in which foods such as seasonings and cooking ingredients are enclosed, and the contours, columns, bands, the number and positions of the bands, etc. It is difficult to apply the discrimination system of Patent Document 1 to the configuration peculiar to the orthotic device. In particular, the orthotic device is composed of a support portion such as a main body portion and a support column, and a band for fixing to the body. Some bands are used not only for fixing to the body but also for correcting the body, and just by looking at images such as photographs, does the band influence the function and type of the orthotic device, or is it just? It is difficult to determine whether it is used for fixing.

For this reason, even if the image data of the orthotic device is input to the analysis method of the neural network that simply performs machine learning, it is not possible to obtain a highly accurate discrimination result that is practical.

Japanese Unexamined Patent Publication No. 2020-192677

In view of the above points, it is an object of the present invention to provide an orthotic device discrimination system that discriminates a plurality of types of orthotic devices provided with bands with high accuracy.

[1] An orthotic device discrimination system that discriminates the type of the orthotic device from the image data of the orthotic device provided with the band.
It has a teacher data creation unit, a target data input / result display unit, and an analysis unit.
The teacher data creation unit photographs the equipment having a predetermined model number and specifications (hereinafter referred to as a model number), creates teacher data for each model number from the image data of the model number and the photograph taken, and creates the teacher data for each model number. It is transmitted to the analysis unit.
The target data input / result display unit inputs image data of a photograph of the discrimination target device to be discriminated and transmits the image data to the analysis unit, and determines the discrimination result of the model number of the discrimination target device determined by the analysis unit. To display
The analysis unit extracts a feature amount expressing the contour of the device of a predetermined model number, the shape and the number of bands from the image data of the teacher data transmitted from the teacher data creation unit, and the model number from the feature amount. An estimation model for estimating the above is generated by machine learning, and the estimation model of the model number is stored in the library.
In addition, the feature amount expressing the contour of the discriminant target device, the shape and the number of bands is extracted from the image data transmitted from the target data input / result display unit, and the pre-registered device stored in the library is used. It is characterized in that the degree of similarity with is calculated and output, and is transmitted to the target data input / result display unit as the data of the discrimination result.

According to this configuration, the orthotic device discrimination system includes a teacher data creation unit, a target data input / result display unit, and an analysis unit. The analysis unit extracts a feature amount expressing the contour of the orthotic device of a predetermined model number, the shape and the number of bands from the image data of the teacher data, and generates an estimation model for estimating the model number from the feature amount by machine learning. In addition, from the image data transmitted from the target data input / result display unit, the feature amount expressing the contour of the discriminant target device and the shape and number of bands is extracted, and the similarity with the device registered in advance in the library is calculated. Output. Since the analysis unit extracts and discriminates the feature amount expressing the shape and number of the band in addition to the feature amount of the contour of the device of the predetermined model number and the device to be discriminated in the teacher data, a plurality of types of bands are provided. The orthotic device can be identified with high accuracy.

[2] Preferably, the analysis unit extracts a feature amount expressing the shape of the support column of the orthotic device of a predetermined model number and the position and angle of the band with respect to the support column from the image data of the teacher data.
In addition, the shape of the support column of the device to be discriminated and the feature amount expressing the position and angle of the band with respect to the support column are extracted.

According to such a configuration, the analysis unit extracts and discriminates the shape of the support column of the orthotic device of the predetermined model number and the device to be discriminated from the teacher data and the feature amount expressing the position and angle of the band with respect to the support column, so that the band can be determined. It is possible to discriminate a plurality of types of provided equipment with higher accuracy.

[3] Preferably, the analysis unit extracts the feature amount regardless of the color tone.

According to this configuration, the analysis unit extracts and discriminates the feature amount regardless of the color tone. Therefore, even if the orthotic device has a pattern, the feature amount is extracted only from the shape, and a plurality of types provided with bands. It is possible to discriminate the orthotic device with higher accuracy.

[4] Preferably, the analysis unit calculates the degree of matching of the feature amounts of the discrimination target device with respect to the pre-registered device and outputs a similar registered device.

According to such a configuration, the analysis unit calculates the degree of matching of the feature amounts of the discrimination target equipment with respect to the pre-registered equipment and outputs the similar registered equipment. It is possible to identify close equipment.

[5] Preferably, the teacher data creation unit, the target data input / result display unit, and the analysis unit are connected to each other via a network via the Internet.

According to this configuration, the teacher data creation unit, the target data input / result display unit, and the analysis unit are connected to each other via a network via the Internet, so that a band can be easily provided even at a remote location. It is possible to discriminate a plurality of types of equipment with high accuracy.

[6] Preferably, the target data input / result display unit is used by an insurer of health insurance.

According to this configuration, most of the persons in charge of examination within the insurer are clerical workers who do not have knowledge of orthotic devices, but even those who are not connoisseurs of orthotic devices have multiple types of orthotic devices with bands. Can be discriminated with higher accuracy.

[7] Preferably, the target data input / result display unit includes a mobile terminal with a camera.
The image data of the device to be discriminated is the photograph attached to the application form sent by the applicant on paper with the camera-equipped mobile terminal, or sent electronically by the applicant. It is attached to the application form.

According to such a configuration, the image data of the equipment to be discriminated is a photograph taken by a mobile terminal with a camera attached to the application form sent by the applicant on paper, or sent by the applicant as electronic data. Since it is attached to the application form, it can be handled whether the application form is paper or electronic data, and the equipment can be identified with higher accuracy.

It is possible to discriminate a plurality of types of orthotic devices provided with bands with high accuracy.

It is explanatory drawing of the equipment discriminating system which concerns on this invention. This is the flow from generating an estimation model using teacher data to saving it. This is the flow from determining the model number of the device to be discriminated using the image data of the photograph attached to the application form and displaying the result. It is a figure explaining various kinds of orthotics fixed to a leg. It is a figure explaining various kinds of orthotics fixed to a foot. It is a figure explaining various kinds of orthotics which are fixed from the thigh to the foot. This is a flow for classifying by the shape of the orthotic device and calculating the price of the orthotic device.

Embodiments of the present invention will be described below with reference to the accompanying drawings. It should be noted that the explanatory diagram of the orthotic device discrimination system conceptually (schematically) shows an example of discrimination of an arbitrary orthotic device.

As shown in FIGS. 1, 4, and 5, the orthotic device discrimination system 10 is a system that discriminates the type of the orthotic device 50 (50a, 50b, 50c, 50d, 50e, 50f) from image data 11 such as a photograph. .. The type of the orthotic device 50 is not limited to the orthotic devices 50a, 50b, 50c, 50d, 50e, and 50f shown in the examples, and there are many more types.

There are a plurality of types of the orthotic device 50, and as shown in FIG. 4A as an example, the orthotic device 50a for the knee is arranged along the vertical direction of the leg 70 with the left and right support columns 51 which are the main body portions. A cloth-like strap 52 provided on the support column 51 to cover the leg 70, a wide fixing band 53 provided on the support column 51 above the knee 71 to fix the orthotic device 50a to the leg 70, and a support column below the knee 71. A functional band 54 provided on the 51 and hung across the front side of the leg 70, and a functional band 55 provided on the support column 51 below the knee 71 and hung on the front side of the leg 70 in the left-right direction are provided. ing.

The fixing band 53 is simply for fixing the orthotic device 50a to the leg 70, and the leg 70 is provided from a predetermined position with respect to the support column 51 (for example, the center of the band width is 9 cm downward from the upper end of the support column 51). It extends in the left-right direction (for example, 90 degrees) to the middle of the front side of. On the other hand, the functional band 54 has a function of holding down the lower part of the knee 71 and is hung on the left and right columns 51. Further, the functional band 54 is arranged at a predetermined position (for example, a position where the center of the band width is 3 cm and a position 15 cm upward from the lower end of the support column 51) and a predetermined angle (for example, 60 degrees) with respect to the support column 51. Has been done. The functional band 55 is arranged at a predetermined position (for example, a position where the center of the band width is 9 cm upward from the lower end of the support column 51) and a predetermined angle (for example, 90 degrees) with respect to the support column 51.

As shown in FIG. 1, the orthotic device discrimination system 10 discriminates the type of the orthotic device 50 from the image data 11 of the orthotic device 50 provided with the band 54 (see FIG. 4) and the like. A teacher data creation unit 20, a target data input / result display unit 30, and an analysis unit 40 are provided.

The application form creation unit 12, the teacher data creation unit 20, the target data input / result display unit 30, and the analysis unit 40 are each connected via a network 15 via the Internet.

The application form preparation unit 12 will be described.
The application form preparation unit 12 prepares, for example, an application form 13 for a so-called health insurance insured person (applicant) who has purchased the orthotic device 50 to apply for payment of medical expenses. I have. At present, some insurers do not support formal electronic applications, and e-mail is the main method for submitting image data electronically. In the future, it is possible to upload the image data to the storage on the WEB and share the image data with the applicant and the insurer on the platform owned by the specified person. Here, the electronic submission method of such images is comprehensively expressed as the application preparation department. The applicant creates an application form 13 to which the image data 11 of the photograph of the device 50 is attached by the application terminal 14. This application form 13 is output to paper and sent to the target data input / result display unit 30 with the insurer (route 1), or as electronic data, the target data input / result display with the insurer is performed via the network 15. Send to section 30 (route 2). When the application form 13 is sent on paper (route 1), a photograph of the orthotic device 50 may be attached to the application form originally prepared on paper without preparing the application terminal 14.

Next, the teacher data creation unit 20 will be described.
The teacher data creation unit 20 includes a registration terminal 21 that can be connected to the network 15. As formal original data, for example, an orthotic device 50 having a predetermined model number approved by the Ministry of Health, Labor and Welfare is photographed by a registration terminal 21, and the model number (registered product name, registered orthotic device name) and image data 22 of the photographed photograph are taken. Create a plurality (large amount) of teacher data 23 for each model number. Further, the teacher data creation unit 20 transmits the created teacher data 23 to the analysis unit 40 via the network 15. It is preferable that the device 20 is photographed by taking a plurality of photographs from multiple directions by changing the conditions or the like, and preparing a plurality of image data 21 to be the teacher data 23.

Further, in the teacher data creation unit 20, a plurality of types of orthotic devices 50a, 50b, 50c, 50d, 50e, 50f, etc. of a plurality of types of predetermined model numbers (registered product names) as shown in FIGS. Create type teacher data 23. To create the teacher data 23, the image data 11 used in the application form may be newly added to the teacher data 23 to further improve the accuracy of the teacher data 23 toward the future. As a result, even if the accuracy of the estimation model, which will be described later, is insufficient at the beginning and the accuracy of the similarity is low, the image data 11 used in the application form can be newly added to the teacher data 23. It can be retrained to improve the accuracy of the estimation model and the accuracy of similarity.

Next, the target data input / result display unit 30 will be described.
The target data input / result display unit 30 is used, for example, by an insurer of health insurance. Further, the target data input / result display unit 30 inputs the image data 11 of the photograph of the discrimination target device 50 to be discriminated and transmits it to the analysis unit 40, and the model number of the discrimination target device 50 discriminated by the analysis unit 40. The determination result of (registered product name) is displayed.

The target data input / result display unit 30 receives and stores the electronic data of the application form 13 transmitted from the application form creation unit 12 via the network 15, and the server 31 and the server 31 by wire or wirelessly. It is equipped with a connectable examination terminal 32 (desktop PC or mobile terminal with a camera). The examination terminal 32 can be connected to the network 15 without going through the server 31.

The image data 11 of the discrimination target device 50 to be discriminated is a photograph taken by the examination terminal 32 attached to the application form 13 sent by the applicant on paper. The examination terminal 32 plays a role of digitizing an image sent on paper, and also plays a role of transmitting image data 11 and displaying a discrimination result. In the digitization of images, the examination terminal digitizes the images sent on paper by utilizing the camera function of the mobile terminal, a digital camera, or a device such as a scanner. The image data 11 of the discrimination target device 50 to be discriminated is attached to the application form 13 sent by the applicant as electronic data. In this case, it is not necessary to digitize the examination terminal 32. Further, the transmission of the image data 11 of the examination terminal 32 and the display of the determination result can also be performed by a mobile information terminal such as a smartphone, a desktop PC, or the like.

When the image data 11 of the identification target device 50 is attached to the application form 13 sent as electronic data, a camera-equipped mobile phone or desktop PC may be used among the examination terminals. A mobile terminal with a camera may be used, but instead, a terminal without a camera such as a desktop PC may be used.

In this way, when the application form is sent on paper, the insurer uses a mobile terminal with a camera such as a smartphone as the examination terminal 32 to take a picture of the device 50 to be discriminated in the application form and take an image. The data 11 is sent to the analysis unit 40, and the discrimination result and the degree of similarity returned from the analysis unit 40 can be viewed. In addition, when the application form is sent by electronic data, the insurer uses the image data 11 of the identification target device 50 of the application form uploaded to the server 31 on the Web via the network 15 to the desktop PC. The image data 11 is sent to the analysis unit by the so-called copy and paste method of pasting on the discrimination screen or the selection method, which is used as the examination terminal 32, and the discrimination result and the degree of similarity returned from the analysis unit 40 are viewed. be able to.

In order to discriminate the type or model number of the discriminating device 50, the insurer transmits the image data 11 of the discriminating device 50 from the examination terminal 32 to the analysis unit 40. Then, the determination result from the analysis unit 40 is displayed on the examination terminal 32.

As an example of displaying the discrimination result, in the display column 33, the image data 11 of the application photograph, the image data 22 of the registered orthotic device same as the applied model number (always fixed in the manufacturer's catalog image, etc.), and the model number (registered product name). , Similarity (discrimination accuracy) "%" etc. are displayed. The display field 33 is not limited to this, and in addition to the above, the applicant name and the necessary information a'determined amount of the orthotic device may be displayed.

Next, the analysis unit 40 will be described.
The analysis unit 40 has artificial intelligence (AI) and discriminates the image data 11 by using machine learning. In general, machine learning refers to a technique of learning regularity and judgment criteria from data and predicting and judging unknown things based on the learned regularity and judgment criteria, and an analysis technique related to artificial intelligence. Deep learning, which is included in machine learning, may be used. Deep learning is an extension of the neural network analysis method, which is a more basic and wide-ranging machine learning method, and is highly accurate analysis and utilization. It is a method that enables. In one embodiment of the present invention, machine learning is used in which teacher data corresponding to the correct answer, so-called supervised learning, is given. Regression analysis and decision trees are examples of analysis methods for supervised learning.

A neural network is also applied in the field of image recognition, and a neural network in which an intermediate layer (hidden layer) is layered into two or more layers is called deep learning. Deep learning learns the rules and regularities contained in the data by extracting the features that characterize the data by itself without humans instructing the specific features to focus on in the given data. It is a method to do.

The analysis unit 40 is a feature that expresses the contour of the equipment 50 of a predetermined model number and the shapes and numbers of the bands 53, 54, 55 (see FIG. 4) from the image data 22 of the teacher data 23 transmitted from the teacher data creation unit 20. The quantity is extracted, an estimation model for estimating the model number from this feature quantity is generated by machine learning, and the estimation model for the model number is stored in the library 41.

Further, the analysis unit 40 extracts in advance the contour of the discrimination target device 50 and the feature amount expressing the shape and number of the bands 53, 54, 55 from the image data 11 transmitted from the target data input / result display unit 30. The degree of similarity with the equipment registered in the library 41 is calculated and output, and is transmitted to the target data input / result display unit 30 as the data of the discrimination result. The image data 22 of the estimated model whose feature amount is closest to the feature amount of the discrimination target device 50 among the estimated models stored in the library 41 in advance may be output together with the model number and used as the data of the discrimination result.

Further, the analysis unit 40 is a feature that expresses the shape of the support column 51 (see FIG. 4) of the orthotic device 50 of a predetermined model number and the positions and angles of the bands 53, 54, 55 with respect to the support column 51 from the image data 11 of the teacher data 23. Extract the amount. Further, the analysis unit 40 extracts a feature amount expressing the shape of the support column 51 of the discrimination target orthotic device 50 and the positions and angles of the bands 53, 54, 55 with respect to the support column 51.

Further, the analysis unit 40 calculates and outputs the degree of matching of the feature amounts of the discrimination target device 50 with the feature amounts of the image data 11 of the device (product) registered in advance. Further, the analysis unit 40 may extract a feature amount that does not depend on the color tone. The analysis unit 40 may calculate and output the degree of coincidence of the feature amounts of the discriminant target device 50 with respect to the feature amounts of the closest estimated model.

Next, the flow from generating the estimation model using the teacher data to saving it will be described.

As shown in FIG. 2, in the orthotic device discrimination system 10, the registrant takes a plurality of multi-directional photographs for each model number of the orthotic device 50 in STEP 1 (STEP is referred to as S in the figure) in the teacher data creation unit 20. .. In STEP 2, the model number of the device 50 and the image data 22 are transmitted to the analysis unit 40 (AI terminal) as teacher data 23.

In STEP3, in the analysis unit 40, the feature amount expressing the contour of the orthotic device 50 and the shape of the support column 51 is extracted from the image data 22 which is the teacher data 23 by the AI terminal (artificial intelligence). In STEP4, the feature quantities expressing the shapes, numbers, positions, and angles of the bands 53, 54, and 55 are extracted. In STEP 5, an estimation model that estimates the model number from the features of the teacher data 23 is generated by machine learning. The estimated model generated in association with the teacher data 23 of the registered orthotic device name (registered orthotic device, model number) in STEP 6 is stored in the library 41.

As shown in FIG. 3, in STEP 7, the applicant submits (sends) the application form 13 with a photograph of the device to be discriminated 50 in paper or data in the application form preparation unit 12. In the target data input / result display unit 30, it is determined in STEP 8 whether or not the application form 13 is submitted by paper. If the discrimination is NO in STEP8 (when the application form is sent as electronic data), the registered equipment teacher will be based on the name of the registered equipment described in the application form in STEP9. Data 23 is automatically selected (prepared).

If the determination is YES in STEP 8 (when the application form is paper), the insurer takes a picture of the equipment 50 of the application form 13 and uses it as image data 11 in STEP 10, or the insurer takes a picture of the equipment 50 of the application form 13. The photograph is scanned into image data 11. In STEP 11, the insurer selects the name of the registered device entered in the application form from the registered device name (model number) on the examination terminal 32, and selects (prepares) the teacher data 23 of the registered device. The information of the registered equipment name selected (prepared) together with the image data 11 of the discrimination target equipment 50 applied in STEP 12 is transmitted to the analysis unit 40 (AI terminal).

The analysis unit 40 receives the information of the registered equipment name selected (prepared) together with the image data 11 of the equipment 50 to be discriminated in STEP 13. In STEP 14, the AI terminal (artificial intelligence) extracts the feature amount expressing the contour of the identification target device 50 and the shape of the support column 51 from the image data 11 of the applicant. In STEP 15, feature quantities expressing the shapes, numbers, positions, and angles of bands 53, 54, and 55 are extracted from the applicant's image data 11.

The comparison between the teacher data 23 (image data 22) of the registered device associated with the registered device name selected (prepared) in STEP 16 and the image data 11 of the applied discrimination target device 50 is started. A similar registered orthotic device may be calculated from the feature amount of the image data 11 of the orthotic device 50 to be discriminated by the applicant.

In STEP 17, the similarity (%) between the image data 22 of the registered device name (registered device) and the image data 11 of the device to be discriminated 50 is calculated. In STEP 18, the registered equipment name (registered equipment), its image data 22, and the similarity are output, and the data is returned to the target data input / result display unit 30 (insurer).

In the target data input / result display unit 30, the determination result is displayed on the insurer's examination terminal 32 in STEP 19.

Next, some types of the orthotic device 50 will be described.
The orthotic device 50a shown in FIG. 4A is as described above, and description thereof will be omitted here.

As shown in FIG. 4B, the knee orthotic device 50b is arranged along the vertical direction of the legs 70 and serves as a main body portion of the left and right columns 51, and a cloth provided on the columns 51 to cover the legs 70. A shaped strap 52, a wide fixing band 53 provided on the support column 51 below the knee 71 and fixing the orthotic device 50b to the leg 70, and a wide fixing band 53 provided on the support column 51 above the knee 71 and crossing the front side of the leg 70. It includes a functional band 54 that is hung, and a functional band 55 that is provided on the support column 51 above the knee 71 and is hung on the front side of the leg 70 in the left-right direction.

The fixing band 53 is simply for fixing the orthotic device 50b to the leg 70, and the leg 70 is provided from a predetermined position with respect to the support column 51 (for example, the center of the band width is 9 cm upward from the lower end of the support column 51). It extends in the left-right direction (for example, 90 degrees) to the middle of the front side of. On the other hand, the functional band 54 has a function of holding down the upper part of the knee 71 and is hung on the left and right columns 51. Further, the functional band 54 is arranged at a predetermined position (for example, a position where the center of the band width is 3 cm and a position 15 cm downward from the upper end of the support column 51) and a predetermined angle (for example, 45 degrees) with respect to the support column 51. Has been done. The functional band 55 is arranged at a predetermined position (for example, a position where the center of the band width is 9 cm downward from the upper end of the support column 51) and a predetermined angle (for example, 90 degrees) with respect to the support column 51.

As shown in FIG. 4C, the knee orthotic device 50c is arranged along the vertical direction of the legs 70 and serves as a main body portion of the left and right columns 51, and a cloth provided on the columns 51 to cover the legs 70. The shaped strap 52, the functional band 54 provided on the support column 51 above and below the knee 71 and diagonally hung on the front side of the leg 70, and the left and right sides of the front side of the leg 70 provided on the support column 51 below the knee 71. It includes a functional band 55 that is hung in the direction.

The functional band 54 has a function of holding down the upper part and the lower part of the knee 71, and is obliquely hung on the left and right columns 51. Further, the functional band 54 is arranged at a predetermined position (for example, a position where the center of the band width is intermediate in the vertical direction of the support column 51) and a predetermined angle (for example, 45 degrees) with respect to the support column 51. The functional band 55 is arranged at a predetermined position (for example, a position where the center of the band width is 9 cm upward from the lower end of the support column 51) and a predetermined angle (for example, 90 degrees) with respect to the support column 51.

As shown in FIG. 5A, the orthotic device 50d for the ankle foot 72 is used to fasten the shoe portion 56 into which the foot 72 is inserted, the strap 57 that covers the outside of the ankle 73 and extends diagonally inward, and the strap 57. The fixing band 57a of the above, the iron 58 provided on the shoe portion 56, the support 61 extending from the iron 58 along the leg 70 via the foot joint 59, and the leg provided on the upper end of these support 61. It is provided with a cuff band 62 for fixing the 70.

As shown in FIG. 5B, the orthotic device 50e for the ankle foot 72 is wrapped around the foot 72 and crosses the covering portion 63 that diagonally covers from the tip side of the foot 72 toward the ankle 73 and the front side of the ankle 73. It is provided with a fixing band 63a that fixes the covering portion 63 that is wound so as to be wound and extends diagonally.

As shown in FIG. 6, the equipment 50f for the long leg is attached to the entire leg 70, and the plastic shoe phone brace 64 located from the foot 72 to the leg 70 (short leg) and the shoe phone brace 64. A lower strut 65 extending from the upper end of the race 64 along the leg 70, an upper strut 67 extending upward through the lower strut 65 via a knee joint 66, and the upper strut 67 and the lower strut 65 are provided. It is provided with a knee pad 68 that holds down the knee 71, and a cuff band 69 that fixes the upper strut 67 device 50f to the leg 70.

Next, the flow of classification based on the orthotic device shape based on the neural network will be described.
As shown in (A) of FIG. 7, STEP 20 is used to classify by the shape of the orthotic device. When the band is fixed only (STEP21), the classification is adopted as it is (STEP22). If the band affects the function or product name (STEP23), the adopted name is output (STEP24).

As a background, as insurance coverage has begun to expand even for ready-made equipment (supporters) in recent years, many ready-made knee supporters are on the market, and ready-made knee supporters are devised for each manufacturer. The shape and design of the main body are designed for each to give elaborate originality. As the product lineup increases, there is concern that a photo of the orthotic device will be attached to the insurance application, and the burden on the insurer to examine the identity of the catalog photo and the actual product based on the photo will increase.

Therefore, in the present invention, the image of the equipment registered in advance by the Ministry of Health, Labor and Welfare or the like is used as the teacher data, and the ready-made product catalog image or the actual image is read. Based on that, we provide a mechanism to support insurers to perform more efficient examinations by creating a system that determines whether the image worn by the person applying for insurance is registered. The purpose is to do.

The orthotic device is composed of a support portion which is a main body portion and a band for fixing to the body. However, some bands are used for purposes other than fixing to the body, that is, for correcting the body, and just by looking at the pictures, does the band influence the function and type of the orthotic device, or (the purpose). I don't know if it's used just for fixing (because the function is obtained by the main body part). For this reason, simply inputting into a neural network that performs machine learning does not give practical results.

Therefore, in the orthotic device discrimination system according to the embodiment of the present invention, the information (shape of the support portion) and the band of the main body are focused on the relationship between the orthotic device main body and the band, instead of classifying from the shape of the conventional orthotic device. We decided to think separately and build a neural network that derives the type of orthotic device from the combination. As mentioned above, the constructed neural network will be able to identify the product from the image.

Furthermore, after that, custom-made products are also read as teacher data based on actual photographs. For ready-made products, the insurer may determine whether the product is registered with the Ministry of Health, Labor and Welfare, etc., and pay for the product specified for the registered product. However, on the other hand, for custom-made orthotic devices, the "shape" of the standard for insurance coverage is not defined, and the insurer mainly uses the actual orthotic devices used by patients who make insurance refund claims. The details of the image and the estimate of the amount of the insurance claim will be confirmed.

Since the price of the orthotic device is determined by the integration method, it is necessary to confirm how the orthotic device is configured for the support and parts, but it is very difficult for an insurer who has no knowledge of the orthotic device to perform an accurate examination. it is conceivable that. Further, as mentioned above, the problem of the band installed on the orthotic device makes it difficult to estimate. This is because the orthotic device is composed of a support portion which is a main body part and a band for fixing to the body, and the custom-made orthotic device has a band that can be integrated and a band that cannot be integrated as parts constituting the price. For this reason, even if input is performed into a neural network that simply performs machine learning from the entire image as in the conventional case, practical results cannot be obtained.

Therefore, in this system, instead of classifying based on the shape of the conventional orthotic device, we focus on the relationship between the orthotic device body and the band, consider the information on the main body (shape of the support part) and the band separately, and further consider the orthotic device price. We also detected the accessory parts added to the above, aiming to realize artificial intelligence that can confirm the consistency between the contents of the image and the submitted specification of the orthotic device.

Next, the flow for calculating the price of the orthotic device will be described.
As shown in (B) of FIG. 7, in the orthotic device image, the band to be priced is classified in STEP 25, and the band fee is added (STEP 26) and the band fee is not included (ATEP 27). .. In addition, other accessory parts are classified in STEP 28, and the case where the accessory cost is added (STEP 29) and the case where there is no accessory (STEP 30) are classified, and the final detected addition element is obtained. The description of the part and the equipment are collated (STEP31).

Next, the operation and effect of the orthotic device discrimination system 10 described above will be described.
In the embodiment of the present invention, the orthotic device discrimination system 10 includes a teacher data creation unit 20, a target data input / result display unit 30, and an analysis unit 40. The analysis unit 40 extracts a feature amount expressing the contour of the device 50 of a predetermined model number, the shape and the number of bands from the image data 22 of the teacher data 23, and generates an estimation model for estimating the model number from the feature amount by machine learning. do. Further, from the image data 11 transmitted from the target data input / result display unit 30, the feature amount expressing the contour of the discriminant target device 50 and the shape and number of bands is extracted, and the degree of similarity with the pre-registered device is calculated. ,Output. Since the analysis unit 40 extracts and discriminates the feature amount expressing the shape and number of the band in addition to the feature amount of the contour of the device of the predetermined model number of the teacher data 23 and the device to be discriminated 50, the band is provided. It is possible to discriminate a plurality of types of orthotic devices 50 with high accuracy. Further, since the analysis unit 40 calculates and outputs the degree of matching of the feature amounts of the discrimination target device 50 with respect to the feature amount of the closest estimated model, for example, by displaying the degree of matching in%, a guideline for discrimination can be obtained. Therefore, it is possible to discriminate with higher accuracy.

Further, in the embodiment of the present invention, the analysis unit 40 extracts the feature amount expressing the shape of the support of the orthotic device 50 of the predetermined model number of the teacher data 23 and the orthotic device 50 to be discriminated, and the position and angle of the band with respect to the support. Therefore, it is possible to discriminate a plurality of types of orthotic devices 50 provided with a band with higher accuracy.

Further, in the embodiment of the present invention, the analysis unit 40 extracts and discriminates the feature amount regardless of the color tone. Therefore, even if the orthotic device has a pattern, the feature amount is extracted only from the shape and a band is provided. It is possible to discriminate the plurality of types of orthotic devices 50 with higher accuracy.

Further, in the embodiment of the present invention, the analysis unit 40 calculates the degree of matching of the feature amounts of the discrimination target equipment 50 with respect to the pre-registered equipment 50 and outputs a similar registered orthotic device 50, so that the device is registered. The closest device 50 can be identified from the device 50.

Further, in the embodiment of the present invention, the teacher data creation unit 20, the target data input / result display unit 30, and the analysis unit 40 are connected to each other via the network 15 via the Internet, so that they are located at remote locations. However, it is possible to easily discriminate a plurality of types of orthotic devices 50 provided with bands with high accuracy.

Further, in the embodiment of the present invention, most of the persons in charge of examination within the insurer are clerical workers who do not have knowledge of the orthotic device, and even a person who is not a connoisseur of the orthotic device 50 is provided with a band. It is possible to discriminate a plurality of types of orthotic devices 50 with higher accuracy.

Further, in the embodiment of the present invention, the image data 11 of the discrimination target device 50 is a photograph of a photograph attached to an application form sent by the applicant on paper, taken by a camera-equipped mobile terminal or the like among the examination terminals 32. Or, since it is attached to the application form sent by the applicant as electronic data, it can be handled whether the application form is paper or electronic data, and the equipment 50 can be discriminated with higher accuracy. Can be done.

In the embodiment, the leg orthotic device is determined, but the present invention is not limited to this, and any orthotic device for the upper limbs or lower limbs may be used as long as a band is provided, and even a prosthetic limb may be used. There is no problem. Further, not only the band but also an accessory part such as a knee pad may be included in the discriminating element.

That is, the present invention is not limited to the examples as long as the actions and effects of the present invention are exhibited.

The technique of the present invention is suitable for an orthotic device discrimination system that discriminates a type of orthotic device from image data of an orthotic device provided with a band.

10 ... Equipment discrimination system, 11 ... Image data, 20 ... Teacher data creation unit, 21 ... Registration terminal, 22 ... Image data, 23 ... Teacher data, 30 ... Target data input / result display unit, 31 ... Server, 32 ... Examination terminal (desktop PC, portable terminal with camera), 40 ... Analysis unit, 41 ... Library, 50, 50a, 50b, 50c, 50d, 50e, 50f ... Equipment (discrimination target equipment, registered equipment, artificial limbs), 51, 61 ... struts, 53, 57a, 63a ... fixed bands, 54, 55 ... functional bands, 62, 69 ... cuff bands, 65 ... lower struts, 67 ... upper struts, 68 ... knee pads, 70 ... legs, 71 ... knees, 72 ... feet, 73 ... ankles.

Claims (7)

It is an orthotic device discrimination system that discriminates the type of the orthotic device from the image data of the orthotic device provided with the band.
It has a teacher data creation unit, a target data input / result display unit, and an analysis unit.
The teacher data creation unit creates teacher data for each model number from the image data of the model number and the photograph taken by photographing the orthotic device of a predetermined model number, and transmits the teacher data to the analysis unit. can be,
The target data input / result display unit inputs image data of a photograph of the discrimination target device to be discriminated and transmits the image data to the analysis unit, and determines the discrimination result of the model number of the discrimination target device determined by the analysis unit. To display
The analysis unit extracts a feature amount expressing the contour of the device of a predetermined model number, the shape and the number of bands from the image data of the teacher data transmitted from the teacher data creation unit, and the model number from the feature amount. An estimation model for estimating the above is generated by machine learning, and the estimation model of the model number is stored in the library.
In addition, the feature amount expressing the contour of the discriminant target device, the shape and the number of bands is extracted from the image data transmitted from the target data input / result display unit, and the pre-registered device stored in the library is stored. A device discrimination system characterized in that the degree of similarity with and is calculated and output, and the data of the discrimination result is transmitted to the target data input / result display unit. The orthotic device discrimination system according to claim 1.
The analysis unit extracts from the image data of the teacher data a feature amount expressing the shape of the support column of the orthotic device of a predetermined model number and the position and angle of the band with respect to the support column.
Further, the orthotic device discrimination system is characterized in that the shape of the support column of the device to be discriminated and the feature amount expressing the position and angle of the band with respect to the support column are extracted. The orthotic device discrimination system according to claim 1 or 2.
The analysis unit is an orthotic device discrimination system characterized in that it extracts the feature amount regardless of the color tone. The orthotic device discrimination system according to any one of claims 1 to 3.
The analysis unit is an orthotic device discrimination system characterized in that the analysis unit calculates the degree of matching of the feature amounts of the orthotic device to be discriminated with respect to the orthotic device registered in advance and outputs a similar registered orthotic device. The orthotic device discrimination system according to any one of claims 1 to 4.
The equipment discrimination system characterized in that the teacher data creation unit, the target data input / result display unit, and the analysis unit are connected to each other via a network via the Internet. The orthotic device discrimination system according to any one of claims 1 to 5.
The target data input / result display unit is an orthotic device discrimination system characterized in that it is used by an insurer of health insurance. The orthotic device discrimination system according to claim 6
The target data input / result display unit is provided with a mobile terminal with a camera.
The image data of the orthotic device to be discriminated is a photograph taken by the camera-equipped mobile terminal attached to the application form sent by the applicant on paper, or sent electronically by the applicant. An orthotic device discrimination system characterized by being attached to an application form.

Images