KR102295929B1 - Apparatus and method for extracting interesting medical image based on spine position - Google Patents

Abstract

The present invention relates to an apparatus and method for selecting a medical image of interest based on a position of a spine, comprising: a first neural network learning unit that generates a plurality of first learning data in which image characteristics for each type of vertebra are reflected, and then repeatedly learns it in a first neural network; a spine image feature extractor configured to extract spine image features by detecting and analyzing a spinal object included in each medical image when a medical image series obtained by photographing the subject's abdomen is input; a spinal position identification unit that detects a type of vertebra corresponding to each feature of the vertebral image through the first neural network, and further analyzes a vertebral detection sequence to identify a vertebral position; and an image of interest selector for selecting and outputting medical images of interest after back-tracking a medical image in which the organ of interest is photographed from the result of identifying the position of the vertebrae when the organ of interest is determined.

Description

Apparatus and method for extracting interesting medical image based on spine position}

The present invention relates to an apparatus and method for selecting a medical image of interest based on a position of a spine, which automatically identifies a position of a spine through a medical image, and uses this to select and provide only a necessary image.

The spine is the pillar of the human body. At the center of the spine lies the spinal nerves that connect to the brain, controlling and regulating all functions of the body. The spine supports the body weight and protects the spinal nerves, the connectors that transmit commands from the brain to each part of the body.

31 pairs of spinal nerves extending from the spinal nerves exit between the vertebrae and are distributed in all organs, muscles, blood vessels, ligaments, and skin of the body. is formed

Therefore, when medical images (CT images, MRI images) of the patient's abdomen are acquired, many medical staff first check the position of the spine, then trace the position of the organ of interest from it, and select only the medical image in which the organ is photographed. to perform image reading.

However, in order to determine the exact position of the spine, it takes a considerable amount of time because the medical staff must check hundreds of medical images with the naked eye sequentially, descending from the first image.

In addition, in order to confirm the exact position of the spine while the medical staff is reading the image, it is necessary to go through the process of reconfirming the shape of the vertebrae and organs around the spine by reconfirming the images from the reading position image up and down again, which consumes additional time. .

Domestic Patent Publication 10-2015-0108701 (Published date: September 30, 2015)

Accordingly, in order to solve the above problems, the present invention provides an apparatus and method for selecting a medical image of interest based on the position of the spine, which automatically identifies the position of the spine and selects and provides a medical image of the organ of interest based on this. would like to provide

The object of the present invention is not limited to the object mentioned above, and other objects not mentioned will be clearly understood by those of ordinary skill in the art to which the present invention belongs from the following description.

As a means for solving the above problem, according to an embodiment of the present invention, there is provided a first neural network learning unit that generates a plurality of first learning data in which image characteristics for each type of spine are reflected, and then repeatedly learns it in the first neural network; a spine image feature extractor configured to extract spine image features by detecting and analyzing a spinal object included in each medical image when a medical image series obtained by photographing the subject's abdomen is input; a spinal position identification unit that detects a type of vertebra corresponding to each feature of the vertebral image through the first neural network, and further analyzes a vertebral detection sequence to identify a vertebral position; and an image of interest selector for selecting and outputting medical images of interest after backtracking a medical image captured by the organ of interest from a spinal position identification result when the organ of interest is determined. provides

The spinal position identification unit is characterized in that after detecting the type of the vertebra through the first neural network, the position of the vertebra is identified by sequentially checking which type of vertebra is repeatedly detected and how many times.

The first neural network divides the vertebrae into cervical vertebrae, thoracic vertebrae, and lumbar vertebrae, and the cervical vertebrae are characterized in that the cervical vertebrae are further divided into atlantic vertebrae, axial vertebrae, transverse projections, and fusion vertebrae.

The vertebra position identification unit detects whether the atlas vertebrae, whether the axial vertebrae are detected, how many transverse projections are detected based on the axial vertebrae, whether the vertebrae are detected, how many times the thoracic vertebrae are detected based on the vertebral vertebrae, or the number of times the lumbar vertebrae are detected based on the thoracic vertebrae By confirming whether it is, it is characterized in that the position identification operation of the currently detected vertebra is performed.

The apparatus for selecting a medical image of interest includes: a second neural network learning unit that generates a plurality of second learning data in which image characteristics for each organ type are reflected and then repeatedly learns the second neural network; an organ image feature extractor configured to extract organ image features by detecting and analyzing organ objects included in each of the medical images of interest; and an organ identification unit configured to detect an organ type corresponding to each of the organ image features through the second neural network.

In this case, the organ identification unit compares and analyzes the currently acquired organ image characteristics with the predefined image characteristics for each organ type to identify the organ type corresponding to the currently acquired organ image characteristics.

As a means for solving the above problem, according to another embodiment of the present invention, there is provided a method comprising: generating a plurality of first learning data in which image characteristics for each type of vertebra are reflected, and then repeatedly learning in a first neural network; extracting spine image features by detecting and analyzing a spinal object included in each medical image when a medical image series obtained by photographing the abdomen of a subject is input; detecting a spine type corresponding to each of the spine image features through the first neural network, and then further analyzing a spine detection sequence to identify a spine position; and when the organ of interest is determined, backtracking the medical image of the organ of interest from the spinal position identification result, and then selecting and outputting the medical image of interest as the medical image of interest. do.

The present invention automatically identifies the position of the spine through a neural network, and selects and provides only a necessary image from among large-capacity medical images using the result of identifying the position of the spine. In addition, by additionally checking and utilizing not only the spine position but also the organ position, more accurate image selection operation is possible.

That is, the present invention selects and provides only images of organs of interest from among large-capacity medical images acquired and provided by medical imaging devices such as CT and MRI devices, so that high-quality medical data sets can be more efficiently constructed. do.

1 is a diagram showing the positional relationship between the position of the spine and the long term.
2 to 4 are diagrams for explaining an apparatus for selecting a medical image of interest according to an embodiment of the present invention.
5 is a diagram for explaining an apparatus for selecting a medical image of interest according to another embodiment of the present invention.

The following is merely illustrative of the principles of the invention. Therefore, those skilled in the art will be able to devise various devices which, although not explicitly described or shown herein, embody the principles of the present invention and are included within the spirit and scope of the present invention. Moreover, it is to be understood that all conditional terms and examples listed herein are, in principle, expressly intended solely for the purpose of enabling the concept of the present invention to be understood, and not limited to the specifically enumerated embodiments and states as such. should be

Moreover, it is to be understood that all detailed description reciting the principles, aspects, and embodiments of the invention, as well as specific embodiments, are intended to cover structural and functional equivalents of such matters. It is also to be understood that such equivalents include not only currently known equivalents, but also equivalents developed in the future, i.e., all devices invented to perform the same function, regardless of structure.

Thus, for example, the block diagrams herein are to be understood as representing conceptual views of illustrative circuitry embodying the principles of the present invention. Similarly, all flowcharts, state transition diagrams, pseudo code, etc. may be tangibly embodied on computer-readable media and be understood to represent various processes performed by a computer or processor, whether or not a computer or processor is explicitly shown. should be

The functions of the various elements shown in the figures including a processor or functional blocks represented by similar concepts may be provided by the use of dedicated hardware as well as hardware having the ability to execute software in association with appropriate software. When provided by a processor, the functionality may be provided by a single dedicated processor, a single shared processor, or a plurality of separate processors, some of which may be shared.

In addition, clear use of terms presented as processor, control, or similar concepts should not be construed as exclusively referring to hardware having the ability to execute software, and without limitation, digital signal processor (DSP) hardware, ROM for storing software. It should be understood to implicitly include (ROM), RAM (RAM) and non-volatile memory. Other common hardware may also be included.

In the claims of the present specification, a component expressed as a means for performing the function described in the detailed description includes, for example, a combination of circuit elements that perform the function or software in any form including firmware/microcode, etc. It is intended to include all methods of performing the functions of the device, coupled with suitable circuitry for executing the software to perform the functions. Since the present invention defined by these claims is combined with the functions provided by the various enumerated means and combined in a manner required by the claims, any means capable of providing the functions are equivalent to those contemplated from the present specification. should be understood as

The above-described objects, features, and advantages will become more apparent through the following detailed description in relation to the accompanying drawings, and accordingly, those of ordinary skill in the art to which the present invention pertains can easily implement the technical idea of the present invention. There will be. In addition, in describing the present invention, if it is determined that a detailed description of a known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

2 is a diagram for explaining an apparatus for selecting a medical image of interest according to an embodiment of the present invention.

Referring to FIG. 2 , the apparatus for selecting a medical image of interest according to the present invention includes a first neural network learning unit 10 , a medical image input unit 20 , a spine image feature extraction unit 30 , a spinal position identification unit 40 , and an interest and an image selection unit 50 and the like.

For reference, as shown in FIG. 3 , the spine is largely divided into cervical (C), thoracic (T), and lumbar (L), and their anatomical structures are different from each other. In addition, the anatomical structure of the cervical vertebrae (C1), the axial vertebrae (C2), the transverse projections (C3~C6), and the vertebrae (C7) are different from each other.

Accordingly, the first neural network learning unit 10 connects to an external device such as a medical database, searches for and acquires large-capacity spinal imaging images, and has image features as an input condition and a spine type as an output condition based on these images. Generate a plurality of first training data. And by repeatedly learning the first neural network through a plurality of first learning data, the correlation between image features and the type of spine is learned in the first neural network.

The medical image input unit 20 receives a medical image series obtained by photographing the subject's abdomen from a medical imaging apparatus such as CT or MRI.

The spine image feature extraction unit 30 detects a spine object included in each of the medical image series, and extracts spine image features (anatomical structure, brightness distribution pattern, etc.) based thereon.

The spine position identification unit 40 analyzes the characteristics of the spine image through the first neural network to identify the type of spine corresponding to each medical image. In addition, the position of the vertebrae should be further distinguished in consideration of the type of vertebrae and the order of detection of the vertebrae.

That is, as in Figure 4, whether the circular vertebrae (C1) was detected, the axial vertebrae (C2) were detected, the transverse projections (C3) were detected based on the axial vertebrae (C2), how many times the vertebrae (C7) were detected, the vertebrae (C7) By sequentially checking how many times the thoracic vertebrae (T) are detected or how many times the lumbar vertebrae (L) are detected based on the thoracic vertebrae (T), the correct position of the currently detected vertebra can be automatically identified. do.

The image of interest selector 50 pre-stores and manages information in which the position of the spine and the long-term positional relationship are defined as shown in FIG. 1 . In addition, when an organ of interest to be imaged is determined by a medical team, etc., based on the information, a predetermined number of medical images in which the organ of interest is photographed are traced back from the spinal position identification result, and these are selected and output as medical images of interest.

As described above, according to the present invention, after automatically detecting the type of vertebrae based on the neural network, by additionally confirming which vertebrae are detected in what order, the correct position of each vertebra can be identified. Also, based on the result of identifying the position of the spine, it is possible to select and provide only the medical image of interest in which the organ of interest is captured from among the large-capacity medical images provided by various medical imaging apparatuses.

5 is a view for explaining an apparatus for selecting a medical image of interest according to another embodiment of the present invention, which relates to an apparatus for selecting a medical image of interest by additionally considering a position of an organ as well as a position of a spine.

Referring to FIG. 5 , the apparatus of the present invention includes a first neural network learning unit 10 , a medical image input unit 20 , a spine image feature extracting unit 30 , a spinal position identification unit 40 , and an image of interest selection unit ( 50), it further includes a second neural network learning unit 60 , an organ image feature extracting unit 70 , and an organ identification unit 80 .

The second neural network learning unit 60 connects to an external device such as a medical database to search and acquire large-capacity long-term images, and then based on these images, a plurality of images having image characteristics as input conditions and organ types as output conditions. Generates second training data of And by repeatedly learning the first neural network through a plurality of second learning data, the correlation between image features and organ types is learned in the second neural network.

The organ image feature extractor 30 detects organ objects included in each of the medical images of interest selected by the image of interest selector 50, and extracts organ image features (anatomical structure, brightness distribution pattern, etc.) based on this. extract

The organ identification unit 80 analyzes the organ image characteristics through the second neural network to identify the organ type corresponding to each medical image.

Then, the image of interest selector 50 selects and outputs only a predetermined number of medical images in which the organ of interest is photographed by considering the spinal position identification result and the organ position identification result together. That is, considering that the position of the organ may vary slightly depending on the individual's health and physical condition, a wide range of medical images of interest are first selected based on the position of the spine, and then the narrow By performing secondary selection of a medical image of interest within a range, a more accurate medical image of interest selection operation may be performed.

Also, the organ identification unit 80 pre-defines and stores information on image characteristics for each organ type, and compares and analyzes the currently acquired organ image characteristics with the information to identify the organ type corresponding to the currently acquired organ image characteristics. make it possible to do

As described above, the present invention automatically detects an organ type using a neural network-based or image processing method, and uses it together with a spinal position identification result to select and provide a medical image of interest.

The apparatus or method according to the present invention described above may be produced as a program to be executed on a computer and stored in a computer-readable recording medium, and examples of the computer-readable recording medium include ROM, RAM, CD-ROM, There are magnetic tapes, floppy disks, optical data storage devices, and the like, and also includes those implemented in the form of a carrier wave (eg, transmission through the Internet).

The computer-readable recording medium is distributed in a network-connected computer system, so that the computer-readable code can be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the method can be easily inferred by programmers in the art to which the present invention pertains.

In the above, preferred embodiments of the present invention have been illustrated and described, but the present invention is not limited to the specific embodiments described above, and it is common in the technical field to which the present invention pertains without departing from the gist of the present invention as claimed in the claims. Various modifications may be made by those having the knowledge of, of course, and these modifications should not be individually understood from the technical spirit or perspective of the present invention.

Claims (7)

a first neural network learning unit that generates a plurality of first learning data in which image characteristics for each vertebrae are reflected, and then repeatedly learns the first neural network;
a spine image feature extractor configured to extract spine image features by detecting and analyzing a spinal object included in each medical image when a medical image series obtained by photographing the subject's abdomen is input;
a spinal position identification unit that detects a type of vertebra corresponding to each feature of the vertebral image through the first neural network, and further analyzes a vertebral detection sequence to identify a vertebral position; and
When the organ of interest is determined, a medical image of the organ of interest photographed from the spinal position identification result is traced back, and an image of interest selector that selects and outputs the medical image of interest is included;
The first neural network divides the vertebrae into cervical, thoracic, and lumbar vertebrae, and the cervical vertebrae are further divided into atlas, axial, transverse projections, and vertebrae,
The vertebra position identification unit detects whether the atlas vertebrae, whether the axial vertebrae are detected, how many transverse projections are detected based on the axial vertebrae, whether the vertebrae are detected, how many times the thoracic vertebrae are detected based on the vertebral vertebrae, or the number of times the lumbar vertebrae are detected based on the thoracic vertebrae An apparatus for selecting a medical image of interest based on a spinal position, characterized in that by confirming whether the vertebra is present, a position identification operation of the currently detected vertebra is performed. According to claim 1, wherein the spinal position identification unit
The apparatus for selecting a medical image of interest based on a spinal position, characterized in that after detecting a vertebra type through the first neural network, the vertebral position is identified by sequentially checking which type of vertebra is repeatedly detected and how many times. delete delete According to claim 1,
a second neural network learning unit generating a plurality of second learning data in which image characteristics for each organ type are reflected, and then repeatedly learning the second neural network;
an organ image feature extractor configured to extract organ image features by detecting and analyzing organ objects included in each of the medical images of interest; and
The apparatus for selecting a medical image of interest based on the position of the spine further comprises; an organ identification unit configured to detect an organ type corresponding to each of the organ image features through the second neural network. The method of claim 5, wherein the organ identification unit
An apparatus for selecting a medical image of interest based on a spinal position, characterized in that the currently acquired organ image characteristics are compared and analyzed with the predefined image characteristics for each organ type to identify the organ type corresponding to the currently acquired organ image characteristics. A method for screening a medical image of interest based on a spine position of a medical image of interest screening device, the method comprising:
generating, by the apparatus for selecting a medical image of interest, a plurality of first training data in which image characteristics for each type of vertebra are reflected, and then repeatedly learning it in a first neural network;
extracting spine image features by detecting and analyzing a spine object included in each medical image when a medical image series obtained by photographing the abdomen of a subject is input by the medical image selection apparatus of interest;
detecting, by the apparatus for selecting a medical image of interest, a vertebral type corresponding to each of the vertebral image features through the first neural network, and then further analyzing a vertebral detection sequence to identify a vertebral position; and
When the organ of interest is determined by the apparatus for selecting the medical image of interest, back-tracking the medical image in which the organ of interest is photographed from the spinal position identification result, and then selecting and outputting the medical image of interest,
The first neural network divides the vertebrae into cervical, thoracic, and lumbar vertebrae, and the cervical vertebrae are further divided into atlas, axial, transverse projections, and vertebrae,
The step of identifying the position of the vertebrae includes whether the atlas vertebrae are detected, whether the axial vertebrae are detected, how many transverse projections are detected on the basis of the axial vertebrae, whether the vertebrae are detected, the number of times the thoracic vertebrae are detected on the basis of the vertebrae, or the lumbar vertebrae on the basis of the thoracic vertebrae. A method of selecting a medical image of interest based on a position of a spine, characterized in that the operation of identifying the position of the currently detected vertebra is performed by checking the number of times it is detected.

Images