KR102351373B1 - Apparatus for measuring body size through reverse direction fitting and method thereof - Google Patents

Abstract

The present invention relates to an apparatus and method for measuring body size through reverse fitting, wherein at least one feature point designated by a medical professional in a plurality of medical images for a specific body part and a predetermined feature point in a standard medical image are matched. , extracting a predetermined feature point from the standard medical image that most closely matches the input medical image provided by the user, and performing reverse fitting in which the input medical image provided by the user is assigned to the input medical image in the reverse direction. The present invention relates to an apparatus and method for accurately and simply measuring the size of a body part from a feature point assigned to the input medical image.

Description

Apparatus for measuring body size through reverse fitting and method therefor

The present invention relates to an apparatus and method for measuring body size through reverse fitting, and more particularly, matching at least one feature point designated by a medical professional to a plurality of medical images for a specific body part with a predetermined feature point in a standard medical image In this state, by performing reverse fitting in which a predetermined feature point is extracted from the standard medical image that most closely matches the input medical image provided by the user and assigned to the input medical image provided by the user in the reverse direction, The present invention relates to an apparatus and method for enabling a user to accurately and simply measure the size of a specific body part desired by a user from a feature point assigned to the input medical image.

In general, 2D and 3D precision medical images such as X-Ray, Computed Tomography (CT), and Magnetic Resonance Imaging (MRI) are used to find specific lesions, and Read the size and location of the lesion.

As an example for finding a specific lesion through a precision medical image as described above, in the prior art, various medical images for learning having various shapes and sizes for each body part are matched with a single standard medical image determined in advance to learn, A learning model was created, and an input medical image provided by the user was input to the created learning model to generate numerical data capable of quantitative analysis, and the condition was confirmed through the generated numerical data.

However, if the learning model is not sufficiently learned from a large number of learning data when generating the learning model, the input medical image provided by the user and the standard medical image do not match each other, and the result may be inaccurate. Therefore, there is a need to devise a way to increase the accuracy when matching the input medical image provided by the user with the standard medical image for each body part.

Therefore, in the present invention, by performing reverse fitting in which predetermined feature points are extracted from the standard medical image that most closely matches the input medical image provided by the user and assigned to the input medical image in the reverse direction, the specific body part desired by the user is obtained. We would like to suggest a way to measure the size accurately and easily.

That is, the present invention matches at least one feature point designated by a medical professional to a plurality of medical images for a specific body part and a preset feature point to a standard medical image, and the standard medical care that most closely matches the input medical image provided by the user By matching the feature points of the image to the input medical image in the reverse direction, it is possible to measure the size of a specific body part.

Next, the prior art existing in the technical field of the present invention will be briefly described, and then the technical matters that the present invention intends to achieve differently from the prior art will be described.

First, Korean Patent No. 1825719 (2018.02.06.) relates to a brain image processing method and a method and apparatus for matching a clinical brain image and a standard brain image using the same, and includes a plurality of first images for predefined standard brains. A step of extracting key points or features from each, a vector quantization step of quantizing feature vectors including key points or features extracted differently according to the size and shape of the brain into global features to match the length, and a vector quantization step for each of the plurality of first images and storing the first vector quantization information corresponding to each other in length.

That is, the prior art describes the content of finding a standard brain image corresponding to a clinical brain image having a different brain shape and size for each person and generalizing the brain image by matching the clinical brain image to the standard brain image. However, the prior art is different from the present invention in configuration because it is not a technique for matching a predetermined feature point of a standard medical image to an input medical image provided by a user in the reverse direction.

In addition, Korea Patent Registration No. 1740464 (June 8, 2017) relates to a method and system for diagnosing and predicting stroke, an image acquisition unit configured to receive a plurality of images including at least a part of a human brain; an image alignment unit for aligning the plurality of images based on a standard brain image; a lesion region detection and mapping unit for detecting lesion regions from the plurality of images, respectively, and mapping the detected lesion regions to generate a single mapping image; a matching and correcting unit that matches a standard brain image by scaling the mapping image and performs image correction on the mapping image; a three-dimensional image generator configured to generate a three-dimensional lesion image by accommodating the mapping image in a three-dimensional data space; and a stroke diagnosis unit for diagnosing a stroke based on the three-dimensional lesion image.

That is, the prior art describes the contents of accurately diagnosing stroke and reliably predicting the patient's condition of stroke. However, since the prior art is a technology that provides quantitative and statistical results by visualizing the cause and severity of stroke with class-specific accuracy, the characteristic points of a predetermined standard medical image are matched in the reverse direction with the input medical image provided by the user, It is different from the configuration of the present invention in which the size of a specific body part desired by the user is measured using the feature points matched in the reverse direction.

As a result of reviewing the prior art above, the prior art suggests a configuration for generalizing a brain image by matching a clinical brain image to a standard brain image, and a configuration to visualize and provide the cause and severity of a stroke with accuracy. Unlike the method of extracting the feature points of the input medical image provided by the user and matching the extracted feature points to the standard medical image in the forward direction, the preset feature points in the standard medical image are applied in the reverse direction to the input medical image provided by the user. By matching, the technical characteristics of measuring the size of a specific body part desired by the user are presented, and since there is no description or any suggestion in the prior art for the related configuration, the prior art and the present invention have clear technical differences will be.

In addition, by customizing the product needed by the user through the size information of a specific body part confirmed through the medical image provided by the user, it is possible to increase user satisfaction, improve sales, and facilitate inventory management. The configuration is a feature of the present invention that is not presented at all in the prior art.

The present invention was created to solve the above problems, and by breaking away from a method of extracting feature points from an input medical image and matching them in a forward direction to a standard medical image, the user provides predetermined feature points to a standard medical image. An object of the present invention is to provide an apparatus and method for accurately identifying feature points in the input medical image by matching the input medical image in the reverse direction.

In addition, in a state in which at least one feature point designated by a medical professional in a plurality of medical images for a specific body part and a preset feature point in a standard medical image are matched, the method is the closest match to the input medical image provided by the user. An object of the present invention is to provide an apparatus and method capable of measuring the size of a specific body part by performing reverse fitting by extracting predetermined feature points from a standard medical image and assigning them in the reverse direction to the input medical image.

In addition, the present invention generates a learning model by learning a matching relationship between at least one feature point designated by a medical professional in a plurality of general medical images for a specific body part and a feature point predetermined in a standard medical image, and through the created learning model Another aspect of providing an apparatus and method capable of measuring the size of a specific body part by performing reverse fitting in which predetermined feature points are assigned in the reverse direction to the standard medical image that most closely matches the input medical image provided by the user The purpose.

Another object of the present invention is to provide an apparatus and method for measuring an actual size through a medical image without a separate reference to the size of the image.

Another object of the present invention is to provide an apparatus and method capable of manufacturing a product required by a user through size information on a specific body part confirmed through a medical image provided by the user and a method therefor.

According to an embodiment of the present invention, there is provided an apparatus for measuring body size through backward fitting, comprising: a reverse fitting unit for allocating feature points of a standard medical image for a specific body part to an input medical image in a reverse direction; and a body size measuring unit that measures the size of the specific body part by using the feature points assigned to the input medical image, wherein a predetermined feature point from the standard medical image is assigned to the input medical image in a reverse direction It is characterized in that the body size for the specific body part is measured through fitting.

In addition, the backward fitting unit may include: a standard medical image feature point extraction unit for extracting at least one feature point for one standard medical image that most closely matches the input medical image for the specific body part; a feature point coordinate converter converting coordinates of the extracted feature points of the standard medical image into corresponding coordinates of the input medical image; and a feature point allocator for allocating the feature point to the input medical image using the converted coordinates.

In addition, at least one standard medical image is provided for each body part, and a feature point of the standard medical image is matched in advance with a corresponding feature point of a medical image for a plurality of corresponding body parts, and the feature point is for the body part. It represents a specific point of a medical image, and it is characterized in that it includes the coordinates of the point, a label, and a data value of the coordinates.

In addition, the body size measuring device generates a learning model by learning a matching relationship between at least one feature point of a standard medical image provided for each body part and a corresponding feature point of a general medical image for a plurality of corresponding body parts to generate a learning model It is characterized in that it further comprises;

In addition, the learning model generation unit generates a learning model by learning a matching relationship between a specific feature point designated for each general medical image for the body part and a corresponding feature point of the standard medical image, or in a general medical image for the body part. It is characterized in that the learning model is generated by one-to-one learning of a matching relationship between a specified feature point and a corresponding feature point of the standard medical image.

In addition, the body size measuring unit performs two-dimensional or three-dimensional modeling by connecting and combining the feature points assigned to the input medical image, and the size of the specific body part through the performed two-dimensional or three-dimensional model. is characterized by measuring

In addition, according to an embodiment of the present invention, a method for measuring body size through backward fitting includes a reverse fitting step of, in a body size measuring apparatus, allocating feature points of a standard medical image for a specific body part to an input medical image in a reverse direction; and a body size measuring step of measuring the size of the specific body part by using the feature points assigned to the input medical image, wherein a predetermined feature point from the standard medical image is assigned to the input medical image in a reverse direction It is characterized in that the body size for the specific body part is measured through fitting.

In addition, the backward fitting step may include: a standard medical image feature point extraction step of extracting at least one feature point for one standard medical image that most closely matches the input medical image for the specific body part; a feature point coordinate transformation step of transforming the coordinates of the extracted feature points of the standard medical image into the corresponding coordinates of the input medical image; and a feature point assignment step of allocating the feature point to the input medical image using the transformed coordinates.

In addition, at least one standard medical image is provided for each body part, and a feature point of the standard medical image is matched in advance with a corresponding feature point of a medical image for a plurality of corresponding body parts, and the feature point is for the body part. It represents a specific point of a medical image, and it is characterized in that it includes the coordinates of the point, a label, and a data value of the coordinates.

In addition, the body size measurement method generates a learning model that generates a learning model by learning a matching relationship between at least one feature point of a standard medical image provided for each body part and a corresponding feature point of a general medical image for a plurality of body parts. step; characterized in that it further comprises.

In addition, the learning model generating step may include generating a learning model by learning a matching relationship between a specific feature point designated for each general medical image of the body part and a corresponding feature point of the standard medical image, or a general medical image for the body part It is characterized in that the learning model is generated by one-to-one learning of the matching relationship between the feature points specified in , and the corresponding feature points of the standard medical image.

In addition, the body size measurement step performs two-dimensional or three-dimensional modeling by connecting and combining the feature points assigned to the input medical image, and the measurement of the specific body part through the performed two-dimensional or three-dimensional model. It is characterized by measuring the size.

As described above, according to the apparatus and method for measuring body size through backward fitting of the present invention, in order to identify or find a specific lesion, feature points are extracted from a medical image provided by a user and matched to a standard medical image in a forward direction In contrast to , a standard that most closely matches the input medical image provided by the user in a state where at least one feature point specified by a medical professional in a plurality of medical images for a specific body part and a predetermined feature point are matched to a standard medical image Through reverse fitting, which assigns a predetermined feature point to the medical image to the input medical image in the reverse direction, the feature point can be quickly and accurately confirmed from the user's medical image, and the size of the desired body part from the user's medical image can be conveniently determined based on the feature point It has a measurable effect.

In addition, the present invention can increase the user's product satisfaction by manufacturing and providing customized products to each user by using the size information of the body part measured through the user's medical image by the manufacturer of hats, clothing, etc. , it has the effect of remarkably reducing the possibility of product return, improving sales and easily performing inventory management.

1 is an overall block diagram of an apparatus for measuring body size through backward fitting according to an embodiment of the present invention.
2 is a diagram for explaining the difference between body size measurement and forward processing for lesion confirmation according to backward fitting applied to the present invention.
3 is a view for explaining a body size measurement process through backward fitting according to an embodiment of the present invention.
4 is a diagram for explaining an example of a matching table or a learning model generated according to a matching relationship between each feature point of a standard medical image and a general medical image applied to the present invention.
5 is a diagram for explaining an example of reverse fitting and body size measurement applied to the present invention.
6 is a detailed block diagram illustrating the configuration of an apparatus for measuring body size according to an embodiment of the present invention.
7 is a flowchart illustrating in detail a process of measuring a body size through backward fitting according to an embodiment of the present invention.
8 is a flowchart illustrating in detail a process of constructing a matching table or a learning model based on a matching relationship between a standard medical image for each body part and each feature point of a plurality of general medical images according to an embodiment of the present invention.
9 is a diagram illustrating a hardware structure of an apparatus for measuring body size according to an embodiment of the present invention.

Hereinafter, a preferred embodiment of an apparatus and method for measuring body size through reverse fitting of the present invention will be described in detail with reference to the accompanying drawings. Like reference numerals in each figure indicate like elements. In addition, specific structural or functional descriptions of the embodiments of the present invention are only exemplified for the purpose of describing the embodiments according to the present invention, and unless otherwise defined, all terms used herein, including technical or scientific terms They have the same meaning as commonly understood by those of ordinary skill in the art to which the present invention pertains. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present specification. It is preferable not to

1 is an overall block diagram of an apparatus for measuring body size through backward fitting according to an embodiment of the present invention.

As shown in FIG. 1 , the present invention is configured to include a body size measuring apparatus 100 , a database 200 , a plurality of medical image providing terminals 300 , a plurality of user terminals 400 , and the like.

The body size measuring apparatus 100 collects standard medical images of each body part from the medical image providing terminal 300 through a network, stores them in the database 200 and manages them.

At this time, the standard medical image is generated at least one or more to be used as a standard for each body part such as the head, chest, arm, leg, etc. At least one feature point indicating a pixel) value, a voxel value, etc.) is pre-designated, and each feature point of the standard medical image is previously matched with a corresponding feature point of a medical image for a plurality of corresponding body parts.

In addition, the body size measuring apparatus 100 collects general medical images for each body part from the medical image providing terminal 300 through a network according to a predetermined period or collects in real time, and collects the general medical images for each body part. The matching relationship between the image and the feature points of the standard medical image of the corresponding body part is stored and managed in the database 200 .

For example, the body size measuring apparatus 100 matches a predetermined feature point in the standard medical image with a corresponding feature point (determined by a medical professional directly by a medical professional) of a general medical image for a plurality of corresponding body parts, and each feature point Creating a matching table indicating the matching relationship between A matching table or a learning model is stored and managed in the database 200 .

In this case, the general medical image is a medical image of users who have consented to the use of personal information, and includes all medical images such as a two-dimensional X-ray image, a three-dimensional CT or MRI image, and personal information is deleted so as not to be exposed.

In addition, when generating the matching table or learning model based on the matching relationship between each feature point of the standard medical image and the general medical image, in the present invention, the matching table or learning model is generated by subdividing it according to gender, age, race, etc. Reliability can be greatly improved, and body size measurement accuracy can be improved.

In addition, when generating a learning model by learning the matching relationship between each feature point of the standard medical image and the general medical image, various learning methods such as supervised learning, unsupervised learning, and reinforcement learning can be used in the present invention.

On the other hand, the body size measuring apparatus 100 stores and manages the matching table or the learning model in the database 200, and the size for a specific body part desired by the user from the user terminal 400 through the network. When an input medical image is provided to confirm Alternatively, the learning model is loaded from the database 200 . In this case, the input medical image includes arbitrary key point information (that is, a point temporarily designated by a user who does not know the exact location of the key point) for designating a portion for which the user wants to measure the size.

In addition, a method of using a user interface or a method of utilizing artificial intelligence may be applied as a method of confirming which body part the input medical image is in the body size measuring apparatus 100 .

More specifically, the method of utilizing the user interface is described in detail, in which the body size measuring apparatus 100 allows the user to drag and drop an input medical image through a user interface divided by body parts displayed on the user terminal 400 . (drag & drop) to the corresponding body part, or the user directly inputs the input medical image for which body part in text form, or the user directly enters the input medical image on the body part provided by the body size measuring device 100 . By selecting at least one or more lists from among the list for the medical information, it is possible to immediately check which body part the input medical image is of.

In more detail, the method of utilizing the artificial intelligence is described in detail. When the input medical image is provided from the user terminal 400, the body size measuring apparatus 100 uses the input medical image to check a preset body part. By applying to the learning model, it is possible to check which body part the input medical image is.

Thereafter, the body size measuring apparatus 100 applies the input medical image provided by the user to the loaded matching table or learning model to extract at least one feature point for the most closely matched standard medical image, and the extracted The coordinates of each feature point of the standard medical image are converted into coordinates of the corresponding feature point of the input medical image, and the feature points of the standard medical image are assigned to the input medical image with the converted coordinates.

That is, reverse fitting is performed in which the feature points of the standard medical image that most closely match the input medical image provided by the user through the matching table or the learning model are assigned to the input medical image in the reverse direction.

In addition, the body size measuring apparatus 100 measures the size of a specific body part desired by the user using the feature points assigned to the input medical image, and provides the measured size to the user terminal 400 .

For example, by connecting and combining each feature point assigned to the input medical image, two-dimensional modeling or three-dimensional modeling is performed, and the exact body size for a specific body part is determined through the two-dimensional model or the three-dimensional model. It can be measured and provided to the user.

Accordingly, the user can check the exact size of a specific body part that he or she wants based on the reverse fitting that assigns the feature points of the standard medical image for a specific body part to the input medical image in the reverse direction, and can customize the product to the exact size. be able to

At this time, the difference between body size measurement and forward processing of a medical image for lesion confirmation according to backward fitting applied to the present invention will be described in detail with reference to FIG. 2 .

2 is a diagram for explaining the difference between body size measurement and forward processing for lesion confirmation according to backward fitting applied to the present invention.

As shown in (a) of FIG. 2, a general method of performing lesion identification through medical images is to create a learning model by matching various medical images with a predetermined standard medical image and learning, and then use the medical images provided by the user. It was a forward processing method in which a specific pattern (eg, feature point) was extracted by applying to the learning model, the extracted specific pattern was matched with a standard medical image to derive a result that could be analyzed quantitatively, and through this, the disease was judged.

However, in the present invention, as shown in FIG. 2B , at least one feature point for a standard medical image that most closely matches the input medical image provided by the user is extracted and assigned to the input medical image in the reverse direction. Based on the backward fitting, the feature points are accurately identified in the input medical image, and the feature points are connected to each other to measure the size of a specific body part. In other words, the present invention does not stop at simply finding a condition while converting a medical image provided by a user into a standard medical image in a forward direction, but also applies the feature points set in the standard medical image to the medical image provided by the user through reverse processing to body size It has a compositional characteristic that measures .

On the other hand, the most common usage form of the body size measurement service through backward fitting of the present invention is that the body size measurement device 100 is configured as an independent platform as shown in FIG. 1 , and the user uses the user terminal 400 . When an input medical image is provided through an application program installed in The size of a specific body part desired by a user is measured by performing backward fitting in which one or more feature points are assigned to the input medical image in the reverse direction.

However, the present invention is not limited in this way, and when computing power and environment are improved, the function of the body size measuring apparatus 100 may be directly implemented in the user terminal 400 .

The database 200 includes standard medical images for each body part collected from the plurality of medical image providing terminals 300 by the body size measuring device 100 and general for each body part used for generating the matching table or learning model. Storing and managing medical images. Here, the database 200 is treated and described as a database management system (DBMS) that serves to store and manage data in individual DBs.

In addition, the database 200 stores and manages a matching table or a learning model generated based on a matching relationship between each feature point of the standard medical image and the general medical image in the body size measuring apparatus 100, and manages the measurement of the body size. The matching table or learning model for each body part stored and managed according to the request of the device 100 is provided to the body size measuring device 100 .

In addition, the database 200 is provided by the user based on the generation of a matching table or learning model for each body part using the standard medical image and the general medical image performed by the body size measuring device 100, and the matching table or the learning model Extraction of feature points of a standard medical image that most closely matches a specific body part, reverse assignment of at least one feature point for the standard medical image to the input medical image, and measurement of the size of a specific body part using the assigned feature point It performs storage and update management of various operation programs for

The medical image providing terminal 300 includes a wired/wireless communication terminal operated or possessed by a medical institution or medical institution worker, a communication terminal operated in a data center that comprehensively manages medical images, and a communication terminal owned by individual users, and each A standard medical image for each body part and a general medical image are provided to the body size measuring apparatus 100 .

At this time, when the standard medical image is newly created or changed, it is immediately provided to the body size measuring device 100, and the general medical images for each body part are collected for a certain period according to the request of the body size measuring device 100. It can be provided in a predetermined period or in real time.

The user terminal 400 is a wired/wireless communication terminal such as a smart phone, PC, tablet, etc. used by general users who have various medical images such as X-ray, CT, and MRI, according to the operation of the user using a pre-installed application program. An input medical image of a specific body part is output to the body size measuring apparatus 100 .

In addition, the user terminal 400 receives information on the size of a specific body part that the user wants through the input medical image from the body size measuring device 100 and displays it through the application program.

Through this, users can make various products or manufacture various products according to the size of a specific body part they want measured through medical images, breaking away from the general method using manual measurement using a measuring tool, optical image of themselves, and avatar. order can be executed.

Meanwhile, the process of measuring the body size through the reverse fitting will be described in detail with reference to FIG. 3 as follows.

3 is a view for explaining a body size measurement process through backward fitting according to an embodiment of the present invention.

First, if the medical image providing terminal 300 provides standard medical images and general medical images for various body parts, such as the head, chest, hands, and feet, to the body size measuring device 100 through the network (①), The body size measuring device 100 stores and manages standard medical images and general medical images for each body part in the database 200 (②).

In addition, the body size measuring apparatus 100 includes a matching table generated according to the matching relationship between the standard medical image for each body part and the feature points of the general medical image provided from the medical image providing terminal 300, or the standard for each body part. The learning model generated by learning the matching relationship between the feature points of the medical image and the general medical image is stored and managed in the database 200 (③).

In this case, the process of generating the matching table or the learning model will be described in detail with reference to FIG. 4 as follows.

4 is a diagram for explaining an example of a matching table or learning model generated according to a matching relationship between each feature point of a standard medical image and a general medical image applied to the present invention.

As shown in (a) of FIG. 4 , the body size measuring apparatus 100 corresponds to at least one predetermined feature point in the standard medical image for each body part and each feature point in the standard medical image designated by a medical professional. A learning model by generating a matching table representing the matching relationship between at least one feature point of #1 to #n of a plurality of general medical images for each body part, or by learning the matching relationship between each feature point of the standard medical image and the general medical image , and the matching table or learning model generated for each body part is stored and managed in the database 200 .

As such, the reason for creating a matching relationship between each feature point of the standard medical image and the general medical image as a matching table or learning model is that the general medical image has a difference in the size or shape of the body part. This is because the locations are different, and accordingly, it is necessary to establish a matching relationship in advance for a plurality of general medical images for each body part by checking the matching relationship between each feature point designated in the standard medical image.

In particular, since there is one standard medical image used for generating the matching table or learning model, and there are a plurality of general medical images, the matching relationship created through the matching of the general medical image and the standard medical image is for each general medical image and each feature point. There are many, and it is necessary to create numerous matching relationships as a matching table or learning model.

Meanwhile, in the present invention, a method for creating and managing a matching relationship between each feature point of a standard medical image and a general medical image for each body part as a matching table is described, but the number of general medical images for each body part has greatly increased, so the matching table It would be desirable to apply a learning model because there is a limit to managing it.

When building a matching relationship between each feature point of a standard medical image for each body part and a general medical image as a learning model, the network structure for learning is shown in FIG. CNN is composed of an input layer, a convolution layer, a pooling layer, and a fully connected layer to which general medical images (that is, medical images for learning) for each body part are input.

In this case, the general medical image for each body part input to the input layer preferably uses a full color-based medical image, but is not limited thereto, and the general medical image for each body part is converted to gray scale. It can be used by transforming it, quantizing the region of interest (that is, the region where the feature point is located) and using it, or it can be used after being binary.

In addition, the convolution layer convolves a specific part of the image data for the training data and the weight of the kernel according to a preset stride (that is, the movement unit of the kernel) in which a kernel having a specific weight is set. It generates and outputs a feature map for a specific part of image data.

In addition, the pooling layer subsamples the feature map into at least one or more sub-image data by pooling a specific map output from the convolutional layer according to the sizes of the kernel and the stride.

In addition, the convolution layer and the pooling layer may be configured as a pair and implemented as at least one.

In addition, the fully associative layer connects at least one or more of the generated feature maps to output feature point information for a general medical image for each body part. In this case, the outputted feature point information has a probability value between 0 and 1.

Referring back to FIG. 3 , after establishing a matching relationship between the standard medical images for each body part and each feature point of a plurality of general medical images in the body size measuring apparatus 100 as a matching table or learning model, the user terminal ( 400) provides an input medical image of a specific body part to the body size measuring apparatus 100 according to a user's operation (④), and the body size measuring device 100 provides a specific input medical image provided from the user terminal 400 It is checked which body part the input medical image for the body part is, and a matching table or a learning model corresponding to the identified specific body part is loaded from the database 200 (⑤).

In addition, the body size measuring apparatus 100 applies the input medical image to the matching table or learning model of the corresponding body part loaded from the database 200 to determine at least one feature point for the closest matching standard medical image. is extracted, and the extracted feature points are assigned to the input medical image in the reverse direction (⑥).

In addition, the body size measuring apparatus 100 measures the size of a specific body part desired by the user using the feature points assigned in the reverse direction to the input medical image (⑦), and determines the measured size of the specific body part from the user terminal It is output to (400) so that the user can check the body size measurement result (⑧).

Hereinafter, the reverse fitting and the body size measurement process will be described in detail with reference to FIG. 5 .

5 is a diagram for explaining an example of reverse fitting and body size measurement applied to the present invention.

The body size measuring apparatus 100 generates the input medical image provided by the user in advance and applies it to a matching table or learning model of the corresponding body part, which is a standard medical image that most closely matches the input medical image. Extract the feature points for That is, through the matching table or the learning model, a feature point for a standard medical image having the most similar matching relationship to the input medical image of a specific body part including information on a feature point for which a user wants to measure a size is identified. .

As such, when a feature point of the standard medical image that most closely matches the input medical image is extracted, the body size measuring apparatus 100 sets the coordinates of at least one feature point for the standard medical image to the corresponding feature point of the input medical image. is converted into coordinates of , and each feature point is assigned to the converted coordinates.

Thereafter, the body size measuring apparatus 100 connects each feature point assigned to the input medical image to perform two-dimensional or three-dimensional modeling, and then measures the size of the body part desired by the user to determine the size of the user terminal 400 ) is provided.

Next, a detailed configuration of an apparatus for measuring body size through reverse fitting according to an embodiment of the present invention will be described in detail with reference to FIG. 6 .

6 is a detailed block diagram illustrating the configuration of an apparatus for measuring body size according to an embodiment of the present invention.

As shown in FIG. 6 , the body size measuring apparatus 100 includes a matching table generating unit 110 , a learning model generating unit 120 , a matching table or learning model selecting unit 130 , and a reverse fitting unit 140 . , a body size measurement unit 150 , and a measurement result output unit 160 .

In addition, although not shown in the drawing, the body size measuring apparatus 100 includes a power supply unit for supplying operating power to each component, an input unit for data input for various functions, a memory for storing various operation programs, and updating of various operation programs. An update management unit that manages the body size measurement device 100, a control unit that collectively controls each component, an interface unit for data transmission and reception with an external device such as a database, standard medical images and general medical images for each body part It may further include a collecting unit for collecting.

The matching table generating unit 110 generates a matching table indicating a matching relationship between the general medical image for each body part collected from the medical image providing terminal 300 and the feature points of the standard medical image of the corresponding body part, and the generation A matching table is output to the database 200 to be stored and managed. That is, a matching relationship including the distance between each feature point of the general medical image and the standard medical image, the difference between two-dimensional or three-dimensional coordinates, and data values, etc. is constructed and managed as a table. This can be equally applied to the following learning model.

The learning model generating unit 120 generates a learning model by learning the matching relationship between the general medical image for each body part collected from the medical image providing terminal 300 and the characteristic points of the standard medical image of the corresponding body part, and The generated learning model is output to the database 200 to be stored and managed. That is, the learning model is generated by learning the matching relationship between at least one feature point of the standard medical image provided at least one or more for each body part and the corresponding feature point of the general medical image for the plurality of body parts.

In this case, the learning model generator 120 may generate a learning model by learning a matching relationship between a specific feature point designated for each general medical image of the body part and a corresponding feature point of the standard medical image. For example, learning a matching relationship between a specific feature point (n pieces) for each of a plurality of general medical images for each body part and a specific feature point (one piece) of the standard medical image, n (specific for each general medical image) It is to create one learning model for each feature point) through 1 (specific feature point of standard medical image) learning.

In addition, the learning model generator 120 may generate a learning model by learning a matching relationship between the feature points specified in the general medical image of the body part and the corresponding feature points of the standard medical image on a one-to-one basis. For example, 1 (specific feature points for each general medical image) that learns the matching relationship between each feature point (one piece) of each body part and each feature point (one piece) of the standard medical image for each body part and one-to-one It is to create a learning model of 'general medical image X number of feature points for each body part' through one-to-one (specific feature points of standard medical images).

In addition, the learning model generator 120 verifies the error of the created learning model, and adjusts the learning parameters to minimize the error based on the verified result to optimize the learning model (that is, back propagation). (back-propagation) function) can be performed.

On the other hand, the matching table generating unit 110 and the learning model generating unit 120 select and use any one according to the usage environment (eg, the number of medical images for building the matching table or the learning model). it is preferable However, the present invention is not limited thereto, and the matching table and two learning models may be applied and used at the same time. If the matching table and the two learning models are simultaneously applied, the result derived from the matching table and the learning model are compared, and the one with higher accuracy as a result of the comparison is selected and applied.

The matching table or learning model selection unit 130 checks which body part the input medical image provided from the user terminal 400 corresponds to, and stores it in the database 200 according to the checked body part and manages it. A matching table or a learning model is selected, and information on the selected matching table or learning model is output to the reverse fitting unit 140 .

The reverse fitting unit 140 selects a matching table or learning model for the corresponding body part based on information on the matching table or learning model selected according to the specific body part provided from the matching table or the learning model selection unit 130 . Reverse direction of loading from the database 200 and assigning at least one feature point for a standard medical image that most closely matches the input medical image to the input medical image using the loaded matching table or learning model in the reverse direction perform fitting.

In this case, the backward fitting unit 140 includes a standard medical image feature point extractor 141 , a feature point coordinate transform unit 142 , and a feature point allocator 143 .

The standard medical image feature point extraction unit 141 applies the input medical image to the matching table or the learning model, and at least one feature point for the standard medical image that most closely matches the input medical image for the specific body part. to extract

The key point coordinate conversion unit 142 converts the coordinates of each key point of the standard medical image extracted by the standard medical image key point extraction unit 141 into corresponding coordinates of the input medical image.

The feature point allocator 143 allocates the feature point of the standard medical image to the corresponding position of the input medical image as coordinates for each feature point transformed by the feature point coordinate transformation unit 142 .

The body size measuring unit 150 measures the size of a specific body part desired by the user by using the feature points allocated to the input medical image through the backward fitting unit 140 .

That is, the body size measuring unit 150 performs two-dimensional or three-dimensional modeling by connecting and combining the feature points assigned to the input medical image, and through the performed two-dimensional or three-dimensional model, the user wants It measures the size of a specific body part.

The measurement result output unit 160 transmits the size of a specific body part measured by the body size measurement unit 150 to the user terminal 400, and the user through the application program installed in the user terminal 400 It allows you to immediately determine the size of a specific body part.

In this case, the size of the specific body part may be provided to the user terminal 400 in the form of text, graphics, or a combination thereof.

Next, an embodiment of a method for measuring body size through reverse fitting according to the present invention configured as described above will be described in detail with reference to FIGS. 7 and 8 . At this time, the order of each step according to the method of the present invention may be changed by the environment of use or by those skilled in the art.

7 is a flowchart illustrating in detail a process of measuring a body size through backward fitting according to an embodiment of the present invention.

As shown in FIG. 7 , the body size measuring apparatus 100 determines whether an input medical image is input from the user terminal 400 through the network (S110), and the input medical image from the user terminal 400 is Upon input, it is checked which body part the input medical image is, and a matching table or a learning model according to the body part of the input medical image is selected (S120).

In this case, the confirmation of the body part in step S120 is confirmed by using a user interface or artificial intelligence. For example, by dragging and inputting an input medical image to the corresponding body part through the user interface divided by body part, or by the user Information on body parts may be directly input in text form, a user may select at least one list from a list of body parts, or a method of automatically checking through a learning model for body part confirmation may be applied.

After selecting a matching table or a learning model according to the body part of the input medical image through step S120, the body size measuring apparatus 100 assigns feature points of the standard medical image for a specific body part to the input medical image in the reverse direction Perform the reverse fitting step.

To explain the backward fitting step in more detail, the body size measuring apparatus 100 extracts at least one feature point for a standard medical image that most closely matches the input medical image for a specific body part provided by the user. The standard medical image feature point extraction step is performed (S130), and the feature point coordinate transformation step of converting the coordinates of the extracted feature points of the standard medical image to the corresponding coordinates of the input medical image is performed (S140), and the converted coordinates A keypoint assignment step of allocating the keypoint to the input medical image is performed (S150).

At this time, in step S130, extracting the feature points for the standard medical image that most closely matches the input medical image provided by the user is the matching of a predetermined feature point in one standard medical image and a feature point specified in a plurality of general medical images. This means finding a feature point of a standard medical image that most closely matches the input medical image by using a matching table or a learning model generated based on the matching table.

After the backward fitting step is performed through steps S130 to S150, the body size measuring apparatus 100 measures the size of the specific body part using the feature points assigned to the input medical image. to perform (S160). That is, by connecting and combining the feature points assigned to the input medical image, two-dimensional or three-dimensional modeling is performed, and the size of the specific body part is measured through the performed two-dimensional or three-dimensional model.

In addition, the body size measuring apparatus 100 provides the size of the specific body part measured in step S160 to the user terminal 400 through the network, so that the user can check it immediately (S170).

On the other hand, the body size measuring apparatus 100 determines the matching relationship between the standard medical images for each body part used when measuring the size of a specific body part from the input medical image provided by the user and each feature point of a plurality of general medical images. A matching table or a learning model should be created based on it, stored in the database 200, and managed, which will be described in detail with reference to FIG. 8 as follows.

8 is a flowchart illustrating in detail a process of constructing a matching table or learning model based on a matching relationship between a standard medical image for each body part and each feature point of a plurality of general medical images according to an embodiment of the present invention.

First, the body size measuring apparatus 100 collects a standard medical image for each body part and a plurality of general medical images from the medical image providing terminal 300 (S210).

At this time, at least one standard medical image is provided for each body part, and the feature points of the standard medical image are matched in advance with corresponding feature points of the general medical image for a plurality of corresponding body parts, and the feature points are for the body parts. It indicates a specific point on the medical image, and includes the coordinates of the point, a label, and a data value of the coordinates.

After collecting a plurality of general medical images for each body part through step S210, the body size measuring apparatus 100 determines at least one feature point for each general medical image collected through a medical professional, and Labeling is performed (S220).

Thereafter, the body size measuring apparatus 100 matches the characteristic points predetermined in the standard medical image with the corresponding characteristic points in the general medical image for a plurality of corresponding body parts performed in step S220 to indicate a matching relationship between the characteristic points. A matching table is created (S230).

In addition to generating the matching table of step S230, the body size measuring apparatus 100 provides at least one feature point of the standard medical image provided for each body part and general information for a plurality of corresponding body parts performed through the step S220. A learning model generation step of generating a learning model by learning the matching relationship between corresponding feature points of a medical image is performed (S240), the generated learning model is verified, and weights are weighted in a direction to minimize an error based on the verified result to optimize the learning model (S250).

At this time, when generating the learning model through the step S240, the body size measuring apparatus 100 learns and learns a matching relationship between a specific feature point designated for each general medical image of the body part and a corresponding feature point of the standard medical image A model can be created (n-to-one), or a learning model can be created (one-to-one) by learning the matching relationship between the feature points specified in the general medical image for the body part and the corresponding feature points in the standard medical image on a one-to-one basis. have.

Meanwhile, the body size measuring apparatus 100 may select and use either one of the matching table generation in step S230 and the learning model generation in steps S240 and S250, or both may be selected and used.

In addition, the body size measuring apparatus 100 outputs the matching table generated in step S230 or the learning model created and optimized in steps S240 and S250 to the database 200 to store and manage it (S260).

9 is a diagram illustrating a hardware structure of an apparatus 100 for measuring body size according to an embodiment of the present invention.

As shown in FIG. 9 , the hardware structure of the body size measuring apparatus 100 includes a central processing unit 1000 , a memory 2000 , a user interface 3000 , a database interface 4000 , and a network interface 5000 . , and a web server 6000 and the like.

The user interface 3000 provides an input and output interface to a user by using a graphical user interface (GUI).

The database interface 4000 provides an interface between a database and a hardware structure.

The network interface 5000 provides a network connection between devices owned by a user.

The web server 6000 provides a means for a user to access the hardware structure via a network. Most users can remotely access the web server and use the body size quantification providing apparatus 100 .

Each step of the above-described configuration or method may be implemented as computer-readable code on a computer-readable recording medium or transmitted through a transmission medium. The computer-readable recording medium is a data storage device capable of storing data that can be read by a computer system.

Examples of computer-readable recording media include, but are not limited to, databases, ROMs, RAMs, CD-ROMs, DVDs, magnetic tapes, floppy disks, and optical data storage devices. The transmission medium may include a carrier wave transmitted through the Internet or various types of communication channels.

In addition, the computer readable recording medium may be distributed through a network coupled computer system so that the computer readable code is stored and executed in a distributed manner.

In addition, at least one or more components applied to the present invention may include or be implemented by a processor such as a central processing unit (CPU), a microprocessor, etc. that perform each function, and two or more of the components are one single It may be combined as a component to perform any operation or function for two or more combined components.

In addition, a part of at least one or more components applied to the present invention may be performed by other components among these components. In addition, communication between the components may be performed through a bus (not shown).

In this way, unlike the method of extracting feature points from the medical image provided by the user in order to identify or find a specific lesion and matching it in the forward direction with the standard medical image, the present invention provides a method that most closely matches the input medical image provided by the user. Through reverse fitting in which a predetermined feature point is assigned to the standard medical image in the reverse direction to the input medical image, the feature point can be quickly and accurately identified from the user's medical image, and the size of a desired body part from the user's medical image is determined based on the feature point It can be measured easily.

In addition, since the present invention can produce and provide customized products to each user, the user's product satisfaction can be increased, and sales improvement and inventory management can be easily performed.

As described above, the present invention has been described with reference to the embodiments shown in the drawings, but these are merely exemplary, and those of ordinary skill in the art can make various modifications and equivalent other embodiments therefrom. You will understand that it is possible. Therefore, the technical protection scope of the present invention should be determined by the following claims.

100: body size measuring device 110: matching table generating unit
120: learning model generation unit 130: matching table or learning model selection unit
140: reverse fitting unit 141: standard medical image feature point extraction unit
142: feature point coordinate conversion unit 143: feature point allocator
150: body size measurement unit 160: measurement result output unit
200: database 300: medical image providing terminal
400: user terminal

Claims (12)

configured to extract a feature point indicating a predetermined specific point of the standard medical image closest to the input medical image for a specific body part, and backward fitting by allocating the feature point indicating a predetermined specific point of the standard medical image to the input medical image a reverse fitting unit configured to perform; and
a body size measuring unit configured to measure the size of the specific body part by using the coordinates of the specific point of the feature point assigned to the input medical image by the backward fitting; body size measurement device. The method according to claim 1,
The reverse fitting part,
a standard medical image feature point extraction unit configured to extract at least one feature point for one standard medical image that most closely matches the input medical image;
a feature point coordinate converter configured to convert coordinates of at least one feature point of the extracted standard medical image into corresponding coordinates of the input medical image; and
and a feature point allocator configured to allocate the feature point of the standard medical image to the input medical image using the converted coordinates. The method according to claim 1,
The standard medical image is
At least one is provided for each body part,
The feature point indicating a predetermined specific point of the standard medical image is,
The apparatus for measuring body size through backward fitting, characterized in that it further includes a label of the specific point and data values of the coordinates, and is matched in advance with corresponding feature points of a medical image for a plurality of corresponding body parts. The method according to claim 1,
The body size measuring device,
A learning model is generated by learning the matching relationship between at least one feature point of a standard medical image provided for each body part and a corresponding feature point of a general medical image for a plurality of corresponding body parts, and the generated learning model is stored in a database. Body size measuring device through reverse fitting, characterized in that it further comprises; learning model generation unit configured to manage. 5. The method according to claim 4,
The learning model generation unit,
A learning model is generated by learning a matching relationship between a specific feature point designated for each general medical image of the body part and a corresponding feature point of the standard medical image, or
The apparatus for measuring body size through reverse fitting, characterized in that it is configured to generate a learning model by one-to-one learning a matching relationship between a feature point specified in a general medical image for the body part and a corresponding feature point of the standard medical image. The method according to claim 1,
The body size measurement unit,
2D or 3D modeling is performed by connecting and combining the coordinates of specific points of the feature points assigned by backward fitting to the input medical image, and the specific through the 2D or 3D model on which the modeling is performed A device for measuring body size through backward fitting, characterized in that it is configured to measure the size of a body part. In the body size measuring apparatus, a feature point indicating a predetermined specific point of the standard medical image closest to the input medical image for a specific body part is extracted, and the feature point indicating a predetermined specific point of the standard medical image is added to the input medical image. a backward fitting step of performing backward fitting by assigning; and
body size measurement step of measuring the size of the specific body part by using the coordinates of the specific point of the feature point assigned to the input medical image by the backward fitting; How to measure your size. 8. The method of claim 7,
The reverse fitting step is
a standard medical image feature point extraction step of extracting at least one feature point for one standard medical image that most closely matches the input medical image;
a feature point coordinate transformation step of converting coordinates of at least one feature point of the extracted standard medical image into corresponding coordinates of the input medical image; and
and a feature point assignment step of allocating feature points of the standard medical image to the input medical image using the converted coordinates. 8. The method of claim 7,
The standard medical image is
At least one is provided for each body part,
The feature point indicating a predetermined specific point of the standard medical image is,
The method for measuring body size through backward fitting, characterized in that it further includes a label of the specific point and a data value of the coordinates, and is matched in advance with corresponding feature points of a medical image for a plurality of corresponding body parts. 8. The method of claim 7,
The method of measuring the body size,
A learning model is generated by learning the matching relationship between at least one feature point of a standard medical image provided for each body part and a corresponding feature point of a general medical image for a plurality of corresponding body parts, and the generated learning model is stored in a database. A method of measuring body size through backward fitting, characterized in that it further comprises; generating a learning model to manage it. 11. The method of claim 10,
The learning model creation step is,
A learning model is generated by learning a matching relationship between a specific feature point designated for each general medical image of the body part and a corresponding feature point of the standard medical image, or
A method for measuring body size through reverse fitting, characterized in that a learning model is generated by one-to-one learning of a matching relationship between a feature point specified in a general medical image for the body part and a corresponding feature point of the standard medical image. 8. The method of claim 7,
The body size measurement step is,
2D or 3D modeling is performed by connecting and combining the coordinates of specific points of the feature points assigned by backward fitting to the input medical image, and the specific through the 2D or 3D model on which the modeling is performed A method of measuring body size through reverse fitting, characterized in that the size of the body part is measured.

Images