KR102300234B1 - A method for providing disease information and device performing the same - Google Patents

Abstract

The present invention relates to a method for providing diagnostic assistance information by analyzing a medical image which comprises the following steps of: acquiring a brain image; labeling a feature value reflecting a region of a brain; determining a reference boundary within the brain image; calculating a first disease index and a second disease index; and providing diagnostic assistance information based on the first and second disease indexes.

Description

A method for providing diagnostic auxiliary information and a device for performing the same

The present invention relates to a method for providing diagnostic assistance information by analyzing a medical image, and to a system for performing the same, and more particularly, to a medical image segmented into a plurality of regions, and diagnosis based on a correlation between the plurality of regions It relates to a method for providing auxiliary information and a system for performing the same.

In the field of modern medical technology, there is a growing demand for various techniques for precisely analyzing a medical image and providing information about a disease more accurately. In this trend, next-generation medical technology that analyzes medical images and provides unified disease indicators is in the spotlight.

A method of calculating an index for providing information about a disease by analyzing a medical image has been disclosed through various literatures. However, most of these papers involve the clinical judgment of the doctor who analyzes the medical image, so the information on the disease provided to the patient is too subjective and has severe deviations depending on the doctor.

Also, even when image segmentation is performed to provide disease-related information from a medical image, a segmented region within the medical image varies according to an acquisition condition of the medical image. For this reason, it is difficult to obtain sufficient information about a disease in one specific medical image, and in order to provide accurate diagnostic auxiliary information, it is inconvenient to segment the medical images obtained under various conditions.
The following conventional documents disclose a method for calculating an index for providing information about a disease by analyzing a medical image.
* Korean Intellectual Property Office (KR) Registered Patent Publication No. 10-1754291 (2017.07.06)
* Korean Intellectual Property Office (KR) Registered Patent Publication No. 10-1676789 (2016.11.10)
In addition, the following prior documents disclose a method of segmenting a medical image according to various formats.
* US Patent Application Publication US2019/0096059 A1
* US Patent Application Publication US2020/0167930 A1

One problem to be solved by the present invention is to provide a standardized standard in calculating an index related to a disease by analyzing a medical image to eliminate the influence of a physician's subjective judgment.

One problem to be solved by the present invention is to calculate an objective and uniform disease index by analyzing a medical image using a highly trained artificial neural network.

One problem to be solved by the present invention is to provide disease information in which the influence of the diversity of medical images is removed by calculating a disease index corresponding to the propensity of the medical image.

One problem to be solved by the present invention is to increase the diversity of providing information about diseases by obtaining segmentation results having at least two or more unique characteristics from one medical image.

According to an embodiment of the present invention, there is provided a method comprising: obtaining first image data obtained under a first condition and including information related to a first region related to the brain and second image data obtained under a second condition; determining second region information in second image data to obtain first input data, wherein the second region information is generated to correspond to the second image data based on the first region information; learning the artificial neural network by inputting the first input data and the second image data to the artificial neural network so that the artificial neural network outputs first output data including information about a first target region; transforming the first output data into second input data by morphologically modifying at least a portion of the first target region in the first output data; and inputting the second input data and the second image data so that the artificial neural network outputs a second target area reflecting a first characteristic different from a second characteristic observed in the second image data to generate the artificial neural network. A learning method of an artificial neural network for segmentation of a medical image including the step of learning may be provided.

According to another aspect of the present invention, there is provided an apparatus for providing diagnostic auxiliary information, comprising: a communication module for acquiring input image data; a controller for analyzing the input image data; Including, the controller, the input image data reflecting the first characteristic corresponding to the first condition, using a learning set based on the second image data reflecting the second characteristic corresponding to the second condition A diagnosis aid comprising input into an artificial neural network to obtain result image data, wherein the result image data includes at least a first region indicating lesion information related to a first characteristic and a second region indicating structural information related to a second characteristic An information providing device may be provided.

According to the present invention, it is possible to provide objective and clear disease information from a patient's condition by providing a standardized standard for calculating a disease index from a medical image.

According to the present invention, by providing a method for calculating a criterion in which characteristics included in a medical image are reflected, a method for providing disease information that can be applied to various medical images can be provided.

According to the present invention, by analyzing a medical image using a highly trained artificial neural network, objective and clear disease information from which a doctor's clinical judgment is removed can be provided.

According to the present invention, by using an artificial neural network learned through medical images having at least two different characteristics, it is possible to accurately provide disease information on a medical image corresponding to the characteristics of a patient's disease.

1 is an exemplary embodiment of a system for providing diagnostic auxiliary information according to an exemplary embodiment.
2 is a block diagram schematically illustrating the configuration of a system for providing diagnostic auxiliary information according to an exemplary embodiment.
3 is a block diagram illustrating the configuration of an image acquisition apparatus according to an embodiment.
4 is a block diagram illustrating the configuration of an image analysis apparatus according to an embodiment.
5 is a schematic flowchart of a method for providing auxiliary diagnostic information performed in a system for providing auxiliary diagnostic information according to an exemplary embodiment.
6 is an example of a medical image according to an embodiment.
7 illustrates an example of a medical image according to various acquisition conditions according to an exemplary embodiment.
8 is a flowchart illustrating detailed operations of a medical image analysis operation performed by an image analysis apparatus according to an exemplary embodiment.
9 illustrates an example of a segmented medical image according to an embodiment.
10 is a flowchart illustrating an example of a method of extracting a disease index from a medical image according to an embodiment.
11 is an example of a medical image showing a reference region and a reference boundary according to an exemplary embodiment.
12 is an example of a process of obtaining area information of a target area according to an embodiment.
13 is another example of a process of obtaining area information of a target area according to an exemplary embodiment.
14 is an embodiment in which diagnostic auxiliary information obtained based on a disease index calculated according to an embodiment is provided.
15 is a flowchart illustrating an example of a method of providing diagnostic assistance information using a modified reference boundary according to an exemplary embodiment.
16 illustrates a case in which a reference boundary and an uninterested region overlap according to an exemplary embodiment.
17A and 17B illustrate examples of a method of calculating a modified reference boundary according to an embodiment.
18 illustrates an example of a partially modified reference boundary and a fully modified reference boundary according to an embodiment.
19 illustrates an example of a specific modification process of a partially modified reference boundary according to an embodiment.
20 is a schematic flowchart of a method for an image analysis apparatus to provide diagnostic assistance information from a plurality of medical images, according to an exemplary embodiment.
21 illustrates an example of a plurality of medical images according to an embodiment.
22 is an example of slice images each including different information and an example of providing diagnostic auxiliary information through this example.
23 is a flowchart schematically illustrating a method of obtaining diagnostic assistance information from a candidate image according to an exemplary embodiment.
24 is an example of a candidate image according to an embodiment.
25 is a diagram illustrating an example of providing diagnostic auxiliary information from a candidate image according to an exemplary embodiment.
26 is another example of providing diagnostic auxiliary information according to an exemplary embodiment.
27 is a flowchart illustrating a method of obtaining a disease index based on a plurality of images according to an embodiment.
28 is an example illustrating a process of calculating a disease index of a second image in consideration of information on a first image according to an embodiment.
29 is another example illustrating a process of calculating a disease index of a second image in consideration of information on a first image according to an exemplary embodiment.
30 is a flowchart illustrating an example of a method of providing diagnostic assistance information based on a 3D medical model according to an embodiment.
31 illustrates an example of a process of calculating a disease index from a 3D medical model according to an embodiment.
32 illustrates another example of a process of calculating a disease index from a 3D medical model according to an embodiment.
33 is a diagram schematically illustrating a segmentation operation of a medical image performed by an image analysis apparatus according to an exemplary embodiment.
34 shows an embodiment of an artificial neural network model according to an embodiment.
35 shows another implementation of an artificial neural network model according to an embodiment.
36 exemplarily shows a segmentation result using an artificial neural network model according to an embodiment.
37 illustrates an example of a medical image segmented to include a plurality of characteristics according to an embodiment.
38 is a flowchart illustrating a learning process of an artificial neural network according to an embodiment.
39 is an example illustrating a process of matching a first image and a second image according to an exemplary embodiment.
40 illustrates an example of morphological modification according to an embodiment.
41 is a flowchart illustrating a deployment process using an artificial neural network of an image analysis apparatus according to an embodiment.
42 is an example of a final output result of an artificial neural network according to an embodiment.

The above-described objects, features and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. However, since the present invention may have various changes and may have various embodiments, specific embodiments will be exemplified in the drawings and described in detail below.

In the drawings, the thicknesses of layers and regions are exaggerated for clarity, and also, an element or layer may be referred to as “on” or “on” another component or layer. What is referred to includes all cases in which other layers or other components are interposed in the middle as well as directly on top of other components or layers. Throughout the specification, like reference numerals refer to like elements in principle. In addition, components having the same function within the scope of the same idea shown in the drawings of each embodiment will be described using the same reference numerals.

If it is determined that a detailed description of a known function or configuration related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. In addition, numbers (eg, first, second, etc.) used in the description process of the present specification are only identification symbols for distinguishing one component from other components.

In addition, the suffixes "module" and "part" for the components used in the following description are given or mixed in consideration of only the ease of writing the specification, and do not have a meaning or role distinct from each other by themselves.

The present specification relates to a method of analyzing a medical image and providing diagnostic auxiliary information related to the medical image, and a system for performing the same.

Here, 'medical image' in the present specification may be interpreted as encompassing various images that may be acquired in the medical industry. That is, the medical image in the present specification may collectively refer to images that can be acquired from various devices used in the medical industry. For example, the medical image may be an image obtained by a computer tomography (CT) apparatus such as magnetic resonance imaging (MRI). As another example, the medical image may be a medical image obtained by an X-ray imaging apparatus (X-ray). The present invention is not limited thereto, and may include all images that can be acquired in the medical industry, such as photos that can be acquired by a general camera device.

In addition, the target object of the medical image according to an embodiment may include all target objects that may be targets in the medical industry. For example, the medical image may be an image of a patient. Specifically, the medical image may be an image related to a specific part of the patient's body. As a specific example, the medical image may be a picture related to the brain of a patient suspected of having dementia, or may be an image related to the lungs of a patient suspected of having lung cancer. In addition, the target object of the medical image according to an embodiment may be a body tissue collected from a human.

That is, the medical image in the present specification may be an image of any object to be photographed in the medical industry for which disease information in the image is to be obtained by the image analysis method according to the spirit of the present invention, and various devices used in the medical industry It may be an image having a form obtained from . However, in the following description of the present specification, for convenience of explanation, magnetic resonance imaging (hereinafter, referred to as 'MRI') related to the brain of the human body will be mainly described, but the spirit of the present specification is not limited thereto.

Also, in the present specification, 'diagnostic auxiliary information' may mean comprehensive information that can be objectively determined in relation to a disease that can be determined from a medical image. For example, in the present specification, the diagnosis auxiliary information may indicate whether a specific disease is present or not. As another example, the diagnostic auxiliary information may be a degree of progression of a specific disease. As another example, the diagnostic auxiliary information may be the severity of a disease. As another example, the diagnostic auxiliary information may be a numerical representation of the degree of disease of a specific patient compared to the average group. In addition, as described above, all information related to a disease that can be determined in a medical image may be collectively expressed as 'diagnostic auxiliary information' in the present specification.

First, with reference to FIGS. 1 and 2 , a configuration and exemplary implementation example of a system for providing diagnostic assistance information according to an embodiment will be described.

1 is an embodiment of a system for providing diagnostic assistance information according to an embodiment, and FIG. 2 is a block diagram schematically illustrating the configuration of a system for providing diagnostic assistance information according to an embodiment.

1 and 2 , a system 10000 according to an exemplary embodiment may analyze a medical image to provide diagnostic assistance information.

The system 10000 according to an embodiment may include an image acquisition apparatus 1000 and an image analysis apparatus 2000 .

The image acquisition apparatus 1000 may acquire a medical image from the object to be photographed. Here, the image acquisition device 1000 may refer to various devices or systems capable of acquiring a medical image. For example, as shown in FIG. 1 , the image acquisition apparatus 1000 may be an MRI apparatus. However, as described above, it will be understood that the image acquisition apparatus 1000 as used herein is not limited to the MRI apparatus.

The image analysis apparatus 2000 may analyze the medical image obtained from the image acquisition apparatus 1000 to provide diagnosis auxiliary information. Specifically, the image analysis apparatus 2000 may extract various disease-related indices from the medical image, and calculate auxiliary diagnosis information based on the extracted indices. An operation of analyzing a medical image performed by the image analysis apparatus 2000 will be described later in detail.

In addition, in the above, the image acquisition apparatus 1000 and the image analysis apparatus 2000 have been described as separate devices, but this is only an example and may encompass various forms that can be implemented as the diagnostic auxiliary information providing system 10000 . have. That is, the image analysis apparatus 2000 according to an embodiment may be implemented as a separate server, or may be implemented as a program integrated into the image acquisition apparatus 1000 . However, in this specification, for convenience of description, the system 10000 implemented as a separate image analysis apparatus 2000 will be described.

Hereinafter, the configuration of the image acquisition apparatus and the image analysis apparatus according to an embodiment will be described with reference to the drawings.

3 is a block diagram illustrating the configuration of an image acquisition apparatus according to an embodiment.

Referring to FIG. 3 , the image acquisition apparatus 1000 according to an embodiment may include a first controller 1002 , an image acquisition module 1200 , a first memory 1400 , and a first communication module 1800 . have.

According to an embodiment, the first controller 1002 may transmit the medical image acquired by the image acquisition module 1200 to the image analysis apparatus 2000 through the first communication module 1800 .

Hereinafter, each configuration of the image acquisition apparatus 1000 according to an embodiment will be described.

According to an embodiment, the image acquisition module 1200 may acquire a photographing result of the object to be photographed. Here, the image acquisition module 1200 may include a configuration for acquiring various medical images. As an example, the image acquisition module 1200 may be configured to acquire a magnetic resonance imaging (MRI) image. As another example, the image acquisition module 1200 may be configured to acquire an X-ray or CT image.

Here, the first controller 1002 may adjust a setting parameter of the image acquisition module. For example, when the image acquisition module 1200 is configured to acquire an MRI image, the first controller 1002 may adjust the repetition time TR and the echo time TE of the MRI imaging apparatus. Accordingly, the MRI imaging apparatus may acquire a T1-weighted image or a T2-weighted image. Also, the first controller 1002 may adjust a parameter related to an inversion pulse to obtain a FLAIR image from the MRI imaging apparatus.

The first communication module 1800 according to an embodiment may communicate with an external device or an external server. The image acquisition apparatus 1000 may perform data communication with the image analysis apparatus 2000 or an external device (or server) through the first communication module 1800 . For example, the image acquisition apparatus 1000 may transmit a medical image or data related to the medical image to the image analysis apparatus 2000 or an external device by using the first communication module 1800 .

The first communication module 1800 is largely divided into a wired type and a wireless type. Since the wired type and the wireless type have their respective advantages and disadvantages, in some cases, the image acquisition apparatus 1000 may be provided with both the wired type and the wireless type at the same time.

Here, in the case of the wired type, LAN (Local Area Network) or USB (Universal Serial Bus) communication is a representative example, and other methods are also possible.

In addition, in the case of the wireless type, a communication method of a Wireless Personal Area Network (WPAN) series such as Bluetooth (RPuetooth) or Zigbee may be mainly used. However, since the wireless communication protocol is not limited thereto, the wireless communication module may use a wireless local area network (WLAN)-based communication method such as Wi-Fi or other known communication methods.

The first memory 1400 may store various types of information. Various data may be temporarily or semi-permanently stored in the first memory 1400 . Examples of the first memory 1400 include a hard disk (HDD), a solid state drive (SSD), a flash memory, a read-only memory (ROM), and a random access memory (RAM). ), and so on.

The first memory 1400 includes an operating system (OS) for driving the image acquisition apparatus 1000 or a program for operating each component of the image acquisition apparatus 1000 , as well as an operation of the image acquisition apparatus 1000 . Various data necessary for the .

The first controller 1002 according to an embodiment may control the overall operation of the image acquisition apparatus 1000 . For example, the first controller 1002 may receive a medical image from the image acquisition module 1200 and generate a control signal to transmit it to the data analysis apparatus 2000 through the first communication module 1800 .

The first controller 1002 may be implemented as a CPU (Central Processing Unit) or a similar device according to hardware, software, or a combination thereof. In hardware, it may be provided in the form of an electronic circuit that performs a control function by processing an electrical signal, and in software, it may be provided in the form of a program or code for driving a hardware circuit.

The image acquisition apparatus 1000 may have a separate power supply unit or may receive power from the outside by wire or wirelessly, and may have a switch for controlling the power supply unit.

4 is a block diagram illustrating the configuration of an image analysis apparatus according to an embodiment.

Referring to FIG. 4 , the image analysis apparatus 2000 may include a second controller 2002 , a second memory 2400 , a display module 2600 , and a second communication module 2800 .

According to an embodiment, the second controller 2002 acquires a medical image from the image acquisition device 1000 through the second communication module 2800 and uses a program for analysis stored in the second memory 2400 . By analyzing the medical image, diagnosis auxiliary information may be calculated from the medical image.

Hereinafter, each configuration of the image analysis apparatus 2000 according to an embodiment will be described.

The second memory 2400 may store various types of information of the image analysis apparatus 2000 .

In the second memory 2200 , various data necessary for the operation of the image analysis apparatus 2000 are stored in the second memory 2200 , including an operation program for driving the image analysis apparatus 2000 or a program for operating each component of the image analysis apparatus 2000 . can be saved. For example, a program for processing a medical image and/or a program for analyzing the processed medical image may be stored in the second memory 2400 . The programs may be implemented as a machine learning algorithm, and a detailed description will be given later.

Examples of the second memory 2400 include a hard disk (HDD), a solid state drive (SSD), a flash memory, a read-only memory (ROM), and a random access memory (RAM). ), and so on.

In the second memory 2400 , an operating program (OS) for driving the image analysis apparatus 2000 or a program for operating each component of the image analysis apparatus 2000, as well as an operation of the image analysis apparatus 2000 Various data necessary for the .

The second communication module 2800 may communicate with an external device or an external server. The image analysis apparatus 2000 may perform data communication with the image acquisition apparatus 1000 or an external server using the second communication module 2800 . For example, the image analysis apparatus 2000 may acquire a medical image necessary for providing the diagnosis auxiliary information from the image acquisition apparatus 1000 using the second communication module 2800 .

The second communication module 2800 is largely divided into a wired type and a wireless type. Since the wired type and the wireless type have their respective strengths and weaknesses, in some cases, the image analysis apparatus 2000 may be provided with both the wired type and the wireless type at the same time.

Here, in the case of the wired type, LAN (Local Area Network) or USB (Universal Serial Bus) communication is a representative example, and other methods are also possible.

In addition, in the case of the wireless type, a communication method of a Wireless Personal Area Network (WPAN) series such as Bluetooth (RPuetooth) or Zigbee may be mainly used. However, since the wireless communication protocol is not limited thereto, the wireless communication module may use a wireless local area network (WLAN)-based communication method such as Wi-Fi or other known communication methods.

The second controller 2002 may control the overall operation of the image analysis apparatus 2000 . For example, the second controller 2002 loads a program for image data processing and analysis from the second memory 2400, processes and analyzes the medical image acquired from the image acquisition device 1000, and displays the result. 2 A control signal may be generated to be provided to an external device or an external server through the communication module 2800 . A detailed method for providing diagnostic auxiliary information performed by the image analysis apparatus 2000 will be described later in detail.

The second controller 2002 may be implemented as a central processing unit (CPU) or a similar device according to hardware, software, or a combination thereof. In hardware, it may be provided in the form of an electronic circuit that performs a control function by processing an electrical signal, and in software, it may be provided in the form of a program or code for driving a hardware circuit.

Also, the image analysis apparatus 2000 may further include a separate display module 2600 for outputting a medical image analysis result. The display module may be implemented in various ways to provide information to the user. Here, the second controller 2002 may process and analyze the medical image acquired from the image acquisition apparatus 1000 , and generate a control signal for providing the result through the display module 2600 .

Hereinafter, a method of providing diagnostic auxiliary information by analyzing a medical image performed by the diagnostic auxiliary information providing system according to an exemplary embodiment will be described in detail.

5 is a schematic flowchart of a method for providing auxiliary diagnostic information performed in a system for providing auxiliary diagnostic information according to an exemplary embodiment.

Referring to FIG. 5 , the operation of providing diagnostic auxiliary information according to an exemplary embodiment includes obtaining a medical image (S1000), analyzing the obtained medical image (S1200), and providing diagnostic auxiliary information based on the analysis result. It may include a step (S1400).

First, the diagnostic assistance information providing system 10000 according to an embodiment may acquire a medical image from an object to be photographed. Specifically, the image acquisition apparatus 1000 may acquire a medical image for acquiring diagnostic auxiliary information from an object to be photographed, such as a patient or a sample. More specifically, the image acquisition module 1200 in the image acquisition device 1000 captures the object to be photographed to acquire a medical image, and uses the acquired medical image to the image analysis device 2000 through the first communication module 1800 . ) can be transmitted. Here, there may be a plurality of medical images. That is, the image acquisition apparatus 1000 may acquire a plurality of tomographic images for a specific direction or a specific plane from the object to be photographed.

Here, the plurality of medical images may include three-dimensional information. That is, the medical image may be acquired as a 3D image.

Also, here, the specific direction and specific plane are information about all directions and planes that can be obtained from the object to be photographed, such as the transverse plane, sagittal plane, coronal plane, etc. of the object to be photographed. may include.

Thereafter, the diagnostic auxiliary information providing system 10000 may analyze the medical image. (S1200) The image analysis apparatus 2000 may analyze the medical image received from the image acquisition apparatus 1000 . Specifically, the second controller 2002 may analyze the medical image received through the second communication module 2800 using a program for image analysis stored in the second memory 2400 . For example, the second controller 2002 may segment the medical image using a program stored in the second memory 2400 .

Thereafter, the diagnostic auxiliary information providing system 10000 may acquire the diagnostic auxiliary information based on the analysis result of the medical image ( S1400 ). That is, the image analysis apparatus 2000 may acquire the diagnostic auxiliary information included in the medical image from the analyzed medical image by using the program for obtaining the diagnostic auxiliary information stored in the second memory 2400 . The second controller 2002 may transmit the diagnostic auxiliary information calculated through the second communication module 2800 to an external device or an external server, or output the diagnostic auxiliary information using a separately provided display module 2600, etc. have.

In the above, a method for providing diagnostic auxiliary information performed by the diagnostic auxiliary information providing system 10000 according to an exemplary embodiment has been overviewed. Hereinafter, detailed operations or contents of each step of the method for providing diagnostic auxiliary information performed by the diagnostic auxiliary information providing system 10000 will be described.

First, in relation to the medical image acquisition step ( S1000 ), examples of various medical images that can be acquired by the image acquisition apparatus 1000 according to an embodiment will be described.

6 is an example of a medical image according to an embodiment.

Referring to FIG. 6 , a medical image according to an embodiment may be acquired by the image acquisition apparatus 1000 implemented in various forms. The medical image may be an image captured by various computed tomography imaging devices. Also, as described above, the medical image may be a set of tomographic images composed of a plurality of slice images.

For example, the medical image may be an MRI image captured by an MRI apparatus. (a) and (b) show various body parts of a patient photographed by an MRI apparatus. As another example, the medical image may be an X-ray image captured by an X-ray apparatus. (c) shows the X-ray image of the patient taken by the X-ray device. As another example, the medical image may be a CT image taken by a CT apparatus. (d) shows the CT image of the patient taken by the CT device.

Also, the medical image according to an embodiment may be an image of various body parts. Referring back to FIG. 6 , the medical image may be an image of a body part that encompasses not only organs such as the brain or lungs of a patient, but also a skeleton such as a spine and a neural structure.

In the medical image according to an embodiment, information included in the image may vary according to a setting parameter of the image acquisition apparatus 1000 for acquiring the medical image.

7 illustrates an example of a medical image according to various acquisition conditions according to an exemplary embodiment.

Referring to FIG. 7 , the image acquisition apparatus 1000 according to an embodiment may acquire various images according to parameter settings of the apparatus. For example, when the image acquisition apparatus 1000 is implemented as an MRI apparatus, the image acquisition apparatus 1000 may acquire a medical image including various information by adjusting a parameter related to a magnetic condition. have.

As a more specific example, the image acquisition apparatus 1000 may obtain a T1-weighted image by setting a parameter to have a short TR/TE time. T1-weighted images are useful for clearly distinguishing anatomical structures due to their high signal intensity. That is, the T1-weighted image is mainly used to determine the anatomical characteristics of the human body.

In addition, the image acquisition device 1000 may acquire a T2-weighted image by setting parameters to have a long TR/TE time, and using an inversion pulse to acquire a FLAIR (Fluid attenuated inversion recovery) image. may be The T2-weighted image and the FLAIR image are characterized in that the water-containing area appears white, and the lesion area is mainly used for detecting the lesion area because the water content is high.

In addition, medical images using special imaging techniques such as diffusion-weighted imaging (DWI), PET, and fMRI may also be included in the scope of the present specification.

In the above, examples of various medical images acquired in the medical image acquisition step (S1000) according to an embodiment have been described. Hereinafter, details of the step ( S1200 ) of analyzing the acquired medical image will be described.

8 is a flowchart illustrating detailed operations of a medical image analysis operation performed by an image analysis apparatus according to an exemplary embodiment.

Referring to FIG. 8 , the medical image analysis operation according to an embodiment includes the steps of segmenting the medical image (S2002), extracting a disease index from the segmented medical image (S2004), and assisting diagnosis based on the extracted disease index It may include obtaining information (S2006).

First, the image analysis apparatus 2000 may segment the medical image (S2002). Specifically, the second controller 2002 may segment the medical image using a program for image segmentation stored in the second memory 2400 . For example, the second controller 2002 may segment a brain-related medical image for each brain region. Here, the brain region may refer to a lesion found in the medical image, or may also refer to a structural region of the brain found in the medical image. That is, the region of the brain found in the medical image may include all objects that can be distinguished in the image. In summary, the second controller 2002 may segment the brain-related medical image into at least two or more regions.

A program for image segmentation stored in the second memory 2002 may be implemented as a machine learning algorithm, which will be described later in detail.

After the medical image segmentation step (S2002), the image analysis apparatus 2000 may extract a disease index from the segmented medical image (S2004). Specifically, the second controller 2002 may extract a disease index based on each region included in the segmented medical image using an algorithm stored in the second memory 2400 .

In the present specification, the disease index may be defined as an index expressing the relationship between each region of the brain in a medical image in relation to a specific disease. Disease index is illustratively Fazekas scale (Fazekas scale), ARWMC (Age-related whitematter change), Posterior atrophy score of parietal atrophy, MTA (Medial temporal lobe atrophy score), Orbito-Frontal, Anterior Cingulate, Fronto-Insula, It will be understood that various indicators capable of extracting disease-related information from a medical image may be used, including the Anterior Temporal Scale and the like.

After the disease index extraction step ( S2004 ), the image analysis apparatus 2000 may provide diagnosis auxiliary information based on the extracted disease index. Specifically, the second controller 2002 may use an algorithm stored in the second memory 2400 to obtain information about a disease from a medical image based on the extracted disease index, and output the obtained result.

Hereinafter, an example of a medical image segmented for disease index extraction will be first described, and various embodiments of extracting a disease index from the segmented medical image will be described with reference to FIG. 9 .

9 illustrates an example of a segmented medical image according to an embodiment.

Referring to FIG. 9 , examples of medical images segmented by the image analysis apparatus 2000 are illustrated.

Referring to FIG. 9 , the image analysis apparatus 2000 may segment a medical image. Here, the segmentation may mean that the image analysis apparatus 2000 assigns a specific value to unit cells included in the medical image. Specifically, the segmentation may mean that the second controller 2002 labels a feature value on a pixel or a voxel included in a medical image. As an example, the second controller 2002 may label a plurality of pixels included in the medical image with feature values indicating brain regions.

The image analysis apparatus 2000 may segment the medical image to correspond to the form of the medical image. Here, the form of the medical image may correspond to the acquisition condition when the medical image is acquired. For example, when the medical image is taken under a specific magnetic condition, the medical image may have the form of a T1-weighted image. In the following description in this specification, terms such as the form of the medical image and the conditions for obtaining the medical image will be used interchangeably for convenience of explanation, but it is understood that the spirit of the present specification does not change due to the use of the term. will be In addition, since characteristics (eg, anatomical characteristics or lesion characteristics) included in the medical image may vary depending on the form of the medical image or the conditions for obtaining the medical image, the term characteristics of the medical image is also used interchangeably. and will be able to be used

Specifically, the second controller 2002 may segment the medical image so that characteristics of the medical image are reflected by using the image segmentation program stored in the second memory 2400 .

The second controller 2002 may label a pixel included in the medical image with a value indicating a brain region. That is, if it is within a medical image implemented as a FLAIR image, the second controller 2002 may label the pixels divided into white matter, gray matter, ventricle, and the like, with values indicating respective regions.

Exemplarily, (a) is a segmentation of the T1-weighted MRI image by the image analysis apparatus 2000, and (b) is a segmentation of the T2-FLAIR MRI image by the image analysis apparatus 2000.

As shown in the figure, (a) T1-weighted MRI image is easy to grasp the anatomical characteristics of the brain. That is, when the image analysis apparatus 2000 segments the T1-weighted MRI image, the image analysis apparatus 2000 may segment the brain so that each part constituting the brain is distinguished. For example, the image analysis apparatus 2000 according to an embodiment may segment the brain so that organs included in the brain, such as the cerebrum, cerebellum, diencephalon, and hippocampus, are distinguished. It is also possible to segment the brain so that it can be distinguished according to each other, and it is also possible to segment them by combining them. As such, when the diagnostic auxiliary information is obtained by analyzing the T1-weighted MRI image in which the anatomical characteristics of the brain are well observed, the image analysis apparatus 2000 collects disease information related to the anatomical characteristics of the brain, such as atrophic Alzheimer's. can be easily analyzed.

In addition, (b) T2 FLAIR MRI image is easy to understand the lesion characteristics of the brain. When the image analysis apparatus 2000 according to an embodiment segments the T2-FLAIR MRI image, the image analysis apparatus 2000 may segment the medical image so that lesion characteristics of the brain are distinguished. Here, the lesion characteristics of the brain can be detected through white matter hyperintensity (hereinafter referred to as 'WMH') observed in the medical image. That is, the image analysis apparatus 2000 may segment the medical image in the T2-FLAIR MRI image so that the WMH and the material or other regions constituting the brain are distinguished. Here, the materials constituting the brain (or the head of a person), white matter (white matter), gray matter (graymatter), the skull (skull), etc. will be there, in other areas there may be a ventricle (ventricle) and the like. The image analysis apparatus 2000 may extract a disease index based on the relationship between the location of the WMH and other brain components, which will be described later in detail.

In the drawings, the T1-weighted MRI image and the T2-FLAIR MRI image have been mainly described, but as described above, the spirit of the present specification is not limited thereto. All medical images taken with other imaging devices such as X-ray and CT may be used.

Also, the image analysis apparatus 2000 may segment a medical image having a specific characteristic (ie, acquired under a specific acquisition condition) to include information about other characteristics. Specifically, the second controller 2002 uses a machine learning algorithm stored in the second memory 2400 to photograph a medical image having the first characteristic taken under the first condition under the second condition as well as the first characteristic. Segmentation may be performed to include information of the second characteristic that may be obtained if necessary. For example, when the T2-FLAIR MRI image is segmented, the image analysis apparatus 2000 may segment the T2-FLAIR image to include anatomical information that can be observed in the T1-weighted MRI image. In this regard, it will be described later in detail with reference to the drawings.

Hereinafter, an embodiment of a specific method for extracting a disease index from a segmented medical image will be described with reference to the drawings.

10 is a flowchart illustrating an example of a method of extracting a disease index from a medical image according to an embodiment, FIG. 11 is an example of a medical image showing a reference region and a reference boundary, and FIG. 12 is an area of the target region This is an example of obtaining information, and FIG. 13 is another example of obtaining area information of the target area. 14 is an exemplary embodiment in which diagnostic auxiliary information obtained based on a calculated disease index is provided.

First, a general flow will be described with reference to FIG. 10 . According to an embodiment, the method of extracting a disease index from a segmented medical image performed by the image analysis apparatus 2000 includes detecting a reference region in the segmented medical image (S2102), and a reference boundary based on the detected reference region. It may include setting (S2104) and extracting a disease index based on the relationship between the reference boundary and the target region (S2106).

The image analysis apparatus 2000 may detect the reference region from the segmented medical image. (S2102) Specifically, the second controller 2002 detects pixels labeled to indicate the reference region in the segmented medical image, and , a set of these can be determined as a reference area. Here, the reference region may refer to a region that is included in the medical image and serves as a reference for calculating a specific disease index, as will be described later. For example, the reference region may be a ventricle region. As another example, the reference region may be a white matter region.

Here, the reference area may be an area to which a buffer area of a predetermined area or more is added from a set of labeled pixels to indicate the reference area. The reference region may be defined as a region in which certain pixels of an outer portion are removed from a set of pixels labeled to indicate the reference region.

Also, the image analysis apparatus 2000 may detect a region corresponding to the analysis target from the segmented medical image. Specifically, the second controller 2002 may detect pixels labeled to indicate an analysis target, and determine a set of them as an analysis target area (hereinafter referred to as a 'target area'). Here, the target area is, As will be described later, it may be included in a medical image, and may mean an area to be analyzed for calculating a specific disease index. For example, the target area may mean a WMH area. Since the method of detecting the target region is similar to the method of detecting the ventricular region described above, a detailed description thereof will be omitted.

When the reference region is detected, the image analysis apparatus 2000 may set a reference boundary based on the detected reference region (S2104). Specifically, the second controller 2002 may set a boundary separated by a predetermined distance from the ventricular region detected in the medical image as the reference boundary.

Here, the reference boundary may refer to a boundary serving as a reference for extracting a disease index by deriving a relationship between the reference region and the target region, as will be described later. That is, when a disease index such as the Fazekas scale is used in determining disease information related to Alzheimer's, the positional relationship between the ventricle and WMH can be used as an important indicator, and the position between the ventricle and WMH As a criterion for figuring out the adversarial relationship, the criterion boundary is used. At this time, in the case of the ventricle, it may be determined as a reference area of the Fazekas scale, and in the case of WMH, it may be determined as a target area of the Fazekas scale.

Also, here, in order to obtain the reference boundary, a predetermined distance from any point in the reference area may be calculated. That is, for example, the reference boundary may mean a set of pixels separated by a predetermined distance from a pixel outside the reference area. In this case, the direction serving as a reference for the predetermined distance may preferably be a direction normal to the boundary surface of the reference region. As another example, the reference boundary may mean a set of pixels separated by a predetermined distance from the center of the reference area.

It will be described in relation to the reference boundary with reference to FIG. 11 .

Referring to FIG. 11 , in the T2-FLAIR MRI image, the results of segmentation of the ventricle, WMH, white matter, gray matter, and skull are shown.

The image analysis apparatus 2000 may use the Fazekas scale as a disease index. Specifically, the second controller 2002 may calculate the Fazekas scale by determining the ventricular region as the reference region, and determining the WMH region as the target region. Here, as a reference boundary for calculating the Fazekas scale, the second controller 2002 may set a set of pixels separated by a predetermined distance from a pixel labeled with a ventricle as the reference boundary. In addition, the reference boundary may be calculated from pixels on the outskirts of the ventricular region, and the predetermined distance may be calculated as 10 mm. However, this is only an exemplary value, and an arbitrary distance may be determined within the range of 5 mm to 15 mm.

Referring to FIG. 11 , it is shown that WMH is present both inside and outside the reference boundary. WMH existing inside and outside the reference boundary acts as an important factor in determining the diagnostic auxiliary information. Specifically, the relationship between the information related to the area occupied by the WMH (hereinafter 'region information') and the reference boundary is closely related to the diagnostic auxiliary information. Accordingly, a method of calculating the WMH region information from the segmented medical image by the WMH will be described with reference to FIGS. 12 to 13 .

According to medical research, in general, the white matter region relatively close to the ventricle of the brain tends to be more susceptible to degeneration than the white matter region relatively far from the ventricle. Therefore, the WMH region formed around the ventricle and the WMH region relatively far from the ventricle included in the medical image may each contain different information related to disease. Accordingly, when analyzing the WMH region included in the medical image, it is necessary to analyze the distance from the ventricle into consideration.

12 and 13 illustrate examples of a method of acquiring region information on a target region in consideration of the distance between the reference region and the target region.

Referring to FIG. 12 , the image analysis apparatus 2000 according to an embodiment may acquire region information of a target region in a medical image. Specifically, the second controller 2002 may acquire region information on the target region in the medical image based on the segmentation result. Here, the area information may mean a thickness or an area of a specific area. Also, as will be described later, when the medical image is implemented in 3D, the area information may mean a volume.

First, in the case of calculating the disease index based on the Fazekas scale according to an embodiment, the region information of the target region that exists near the reference region included in the medical image, that is, within the aforementioned reference boundary. to explain about it.

The image analysis apparatus 2000 may acquire area information of the target area based on the distance from the reference area to the target area. Specifically, the second controller 2002 may calculate area information of the target area based on the distance from the area labeled as the reference area in the medical image to the area labeled as the target area.

For example, the second controller 2002 may calculate the thickness of the WMH region based on the distance from the region labeled with the ventricle in the medical image to the region labeled with the WMH. Here, the reference of the distance can be defined in the normal direction from the boundary of the area labeled with the ventricle. Also, here, the largest length among the calculated thicknesses of the WMH may be measured as the thickness of the WMH. More specifically, the second controller 2002 determines the closest WMH region in the normal direction from the boundary of the region labeled with the ventricle as the first point, and determines the furthest WMH region as the second point, The distance between the first point and the second point may be determined as the thickness (width) of the WMH region. Here, if the WMH region is formed by connecting (adhering to) the ventricular region, the first point may be determined as a point on the ventricle.

That is, to express this differently, the second controller 2002 generates a ray in the normal direction at the boundary of the ventricular region in the medical image, and considers the contact point between the generated rays and the pixels labeled with WMH. Thus, the thickness of WMH can be calculated. Here, the contact point of the pixels labeled with ray and WMH may include a first contact point in which the distance between the ventricular area and the WMH area is measured to be close and a second contact point to be measured farthest. That is, the second controller 2002 may calculate the thickness (width) of the WMH region in consideration of the distance between the first contact point and the second contact point. Here, the first contact point may be present on the ventricular region as described above.

The method using a ray is preferably applied to the WMH located within the aforementioned reference boundary, but is not necessarily limited thereto, and the same method may be applied to a WMH other than the reference boundary.

Also, FIG. 13 shows a method of calculating area information for WMH other than the reference boundary.

Referring to FIG. 13 , the image analysis apparatus 2000 may acquire region information of a target region by using a principal component analysis technique. Specifically, the second controller 2002 may extract at least two or more principal components of the target region using principal component analysis, and calculate region information of the target region based on the extracted principal components.

For example, the second controller 2002 may extract principal components for at least two or more axes from the WMH-labeled area in the medical image. Here, the main component may be extracted in various ways, but it is preferable to extract it in the short axis and the long axis within the WMH-labeled region. When the principal component is extracted, the second controller 2002 may calculate the thickness of the region labeled WMH based on the extracted principal component.

That is, to express this differently, the second controller 2002 performs the principal component analysis technique on the region labeled with WMH to obtain directions and sizes for the major and minor axes, and based on the length for the major axis, the WMH labeled The thickness of the area can be calculated. Alternatively, it may be expressed that the second controller 2002 can calculate the width of the WMH region.

In the above, a method of calculating area information of a target area used to obtain a disease index from a medical image has been described. In addition, it has been described that the positional relationship between the target area and the reference area may be considered when obtaining area information of the target area.

Hereinafter, a method of extracting a disease index based on the calculated WMH region information and a reference boundary will be described with reference to FIG. 10 again.

Referring back to FIG. 10 , the image analysis apparatus 2000 may calculate the disease index based on the reference boundary set in the medical image and the region information of the target region (S2106).

Specifically, the second controller 2002 may calculate the disease index by comprehensively considering the correlation between the reference boundary and the target region and region information of the target region.

Here, the correlation may be, for example, whether the WMH is located inside or outside the reference boundary.

As a specific example, it is generally judged that the greater the thickness of the WMH, the higher the severity of the disease. It can be considered comprehensively whether or not

ranking first disease index Second disease index 3rd disease index 0 No white matter degeneration No white matter degeneration Calculated by considering the first disease index and the second disease index One Periventricular 'cap' or linear degeneration Degeneration of speckle shape on deep white matter 2 Degeneration in the form of soft 'halos' around the ventricles Initiation of confluence with periventricular degenerative areas 3 Irregular periventricular degeneration extending to the deep white matter Formation of periventricular degeneration area and large confluence area

ranking first disease index Second disease index 3rd disease index 0 No WMH No WMH When the first disease index is 0 and the second disease index is 0 One The thickness of the white matter degeneration within the reference boundary is less than 5 mm Less than 10 mm in thickness or length of white matter degeneration outside the baseline boundary When the first disease index is grade 1 or 2, and the second disease index is grade 1 2 Within the reference boundary, the thickness of white matter degeneration is 5 mm or more but less than 10 mm Thickness or length of white matter degeneration outside the reference boundary of 10 mm or more but less than 25 mm When the first disease index is grade 1 to 3 and the second disease index is grade 2, or when the first disease index is grade 3 and the second disease index is grade 1 3 The thickness of white matter degeneration measured from within the reference boundary is 10 mm or more 25 mm or more in thickness or length of white matter degeneration outside the reference boundary When the second disease index is grade 3 and the first disease index is grades 1-3

Table 1 is an example showing the criteria for calculating the disease index according to the subjectivity of the existing doctor.

Referring to Table 1, an example of a criterion for calculating the Fazeka scale used in the existing medical industry field is shown. The Fazekas scale is a method of calculating a disease index published in several papers, and it is used as a method of calculating a reliable disease index in the medical industry. However, as shown in Table 1, there is a problem in that the existing medical industry cannot provide objective diagnostic information or diagnostic auxiliary information to the patient because the disease index is calculated according to the subjectivity of the doctor based on ambiguous criteria.

Table 2 shows the criteria for calculating the disease index performed by the image analysis apparatus according to an embodiment.

Accordingly, referring to Table 2, the image analysis apparatus 2000 according to an embodiment may calculate a disease index according to a clear criterion. Specifically, the second controller 2002 may calculate the disease index from the medical image according to the criterion for calculating the disease index stored in the second memory 2400 .

Here, the criterion for the image analysis apparatus 2000 to calculate the disease index may be determined based on various diagnosis results or diagnosis auxiliary results obtained in advance in the medical industry field. That is, according to an embodiment, the image analysis apparatus 2000 calculates the disease index from the medical image by using a criterion quantified as a result of analyzing a plurality of medical images for which the disease index is calculated and obtained in advance from the medical industry site. it can be

That is, the image analysis apparatus 2000 may analyze the medical image to determine a disease index to be calculated according to a predetermined criterion, and calculate a grade of the disease index corresponding to the calculated disease index. Specifically, the second controller 2002 may determine a reference boundary from the reference region, determine a disease index to be calculated in consideration of the relationship between the determined reference boundary and the target region, and calculate a grade value corresponding to the determined disease index.

Here, the predetermined criterion for determining the disease index to be calculated may be the above-described reference boundary. For example, the type of disease index to be calculated by the image analysis apparatus 2000 may vary depending on whether the target region exists within the reference boundary or whether the target region exists outside the reference boundary.

Here, in calculating the same disease index, there may be a criterion for distinguishing the grade of the disease index. For example, the predetermined criterion may be a criterion related to region information of the target region.

Referring back to Table 2, the disease index according to an embodiment may include first to third disease indexes. The disease index shown in Table 2 may mean a Fazekas scale. The first disease index is a disease index related to Peri-ventricular white matter, the second disease index represents a disease index related to Deep white matter, and the third disease index is the first and second A total index calculated in consideration of the disease index may be represented.

Here, the reference boundary may be used to distinguish the first disease index and the second disease index. That is, as described above, the second controller 2002 may calculate the first disease index from the target region within the reference boundary by using the reference boundary. The second controller 2002 may calculate a second disease index from a target region other than the reference boundary by using the reference boundary.

In addition, classes can be distinguished within the same disease index. That is, the second controller 2002 may calculate a grade value corresponding to a specific disease index from the medical image according to a predetermined criterion.

Here, the predetermined criterion may be related to area information of the target area. For example, if there is no cell indicating the target region or it is determined to be less than or equal to a predetermined number, the second controller 2002 may determine that the grade value of the specific disease index is 0. As another example, when the thickness of the target region is obtained to be equal to or greater than the first value, the second controller 2002 may determine that the grade value of the specific disease index is 1. As another example, when the thickness of the target region is obtained to be equal to or greater than the second value, the second controller 2002 may determine that the grade value of the specific disease index is 2 . As another example, when the thickness of the target region is measured to be greater than or equal to the third value or connected to another target region, the second controller 2002 may determine that the grade value of the specific disease index is 3 .

That is, taking an example referring to Table 2, the second controller 2002 may calculate the first disease index based on the target area included in the reference boundary as a result of analyzing the medical image, and the target included in the reference boundary A grade value related to the first disease index may be calculated in consideration of region information of the region.

Also, the image analysis apparatus 2000 may calculate both values related to the first disease index and the second disease index in the same medical image. When the target region exists both inside and outside the reference boundary as a result of the segmentation of the medical image, the second controller 2002 may calculate both the grades of the first disease index and the second disease index.

Also, the image analysis apparatus 2000 may calculate the third disease index in consideration of the first disease index and/or the second disease index.

14 is an exemplary embodiment of a display providing a disease index and diagnostic auxiliary information based on the disease index, according to an exemplary embodiment.

Referring to FIG. 14 , the image analysis apparatus 2000 may provide diagnosis auxiliary information based on a disease index. Specifically, the second controller 2002 may obtain a disease index as a result of analyzing the medical image, and may provide diagnostic assistance information based on the acquired disease index.

As a specific example with reference to the drawings, the second controller 2002 may provide a disease index related to the target region within the reference boundary based on the analysis result of the medical image. Also, based on the disease index, information on the meaning of the disease index may be provided.

Here, as described above, the disease index may refer to a measure (or reference) regarding the degree of progression or severity of a disease that can be determined in the medical image. For example, if the criterion of the disease index is the Fazekas scale, the disease index may be expressed as a grade of the Fazekas scale.

In addition, the diagnosis auxiliary information may refer to information related to a disease that may be determined based on the disease index. For example, when the above-mentioned disease index is calculated as a specific grade using the Fazekas scale, the diagnostic auxiliary information is related to the disease that can be inferred from the disease index, such as the current disease progression, the difference from the normal category, and the difference from the average by age. It can mean information about

Diagnostic auxiliary information obtainable from the medical image and diagnostic auxiliary information to be compared, for example, an average value by age, a disease index value of a normal person, etc. may be acquired in advance and stored in the second memory 2400 and continuously can be updated.

As described above, by setting clear criteria in the image analysis apparatus 2000, extracting a disease index from a medical image, and providing diagnostic auxiliary information based on the extracted disease index, in the meantime, the subjective judgment of doctors in the medical industry The effect of being able to obtain more objective and accurate diagnostic auxiliary information than the information about the disease on which it relied can be derived.

In the above, a method using a predetermined reference boundary has been described as an embodiment of a method for providing diagnostic auxiliary information. However, depending on the characteristics of the medical image, rather than using a fixed reference boundary, it may be necessary to modify the reference boundary in consideration of the characteristics of the medical image to provide more accurate diagnostic auxiliary information.

Hereinafter, a method of using a modified reference boundary as another embodiment of a method of providing diagnostic auxiliary information according to an embodiment will be described with reference to the drawings.

First, with reference to FIG. 15 , another example of extracting a disease index from a medical image and acquiring disease information according to an embodiment will be briefly described.

15 is a flowchart illustrating an example of a method of providing diagnostic assistance information using a modified reference boundary according to an embodiment.

15 , the method for providing diagnostic auxiliary information according to an exemplary embodiment includes detecting a reference region from a segmented medical image (S2202), setting a reference boundary based on the detected reference region (S2204), and a reference The method may include determining whether the boundary overlaps the non-interested region (S2206), correcting the reference boundary (S2208), and obtaining a disease index based on the corrected reference boundary (S2210).

First, in the image analysis apparatus 2000, similar to the operation performed in the embodiment described with reference to FIGS. 10 to 14 , the step of detecting a reference region ( S2202 ), setting a reference boundary based on the detected reference region ( S2202 ) S2204) can be performed, so a detailed description will be omitted.

When the reference boundary is set, the image analysis apparatus 2000 may determine whether the set reference boundary overlaps the non-interested region of the segmented medical image (S2206). Specifically, the second controller 2002 is the medical image. , it may be determined whether a pixel corresponding to a set reference boundary is included in a pixel labeled with a value indicating an uninterested region. Here, the non-interested region may be a gray matter region.

It will be briefly described with reference to FIG. 16 .

16 illustrates a case in which a reference boundary and an uninterested region overlap according to an exemplary embodiment.

Referring to FIG. 16 , it is shown that the reference boundary overlaps the gray matter-labeled region in the medical image.

In general, WMH mainly occurs in the white matter, and WMH is not found in the gray matter region. For this reason, when the disease index is extracted as before when the reference boundary invades the gray matter region, the disease index may be calculated inaccurately. rises Therefore, when the reference boundary overlaps the gray matter region, it is necessary to correct the reference boundary and calculate the disease index by the modified reference boundary.

In the following description of the present specification, a region in which a target region (eg, WMH) can be observed, such as a white matter region, will be referred to as a 'region of interest', and the target region (WMH) is not observed like a gray matter region. The term 'non-interested area' should be used for non-interested areas.

Accordingly, after segmenting the medical image and setting the reference boundary, the second controller 2002 may correct the reference boundary when the reference boundary overlaps with at least a portion of the gray matter-labeled area (S2208).

Referring back to FIG. 15 , when the reference boundary is corrected, the image analysis apparatus 2000 may acquire a disease index based on the corrected reference boundary ( S2210 ). A method of acquiring the disease index is described in FIGS. 10 to 10 . Since it is similar to the embodiment described with reference to 14, a detailed description will be omitted.

Hereinafter, an example of modifying the reference boundary will be described with reference to the drawings.

17A and 17B are various examples of a method of calculating a modified reference boundary according to an embodiment.

17A and 17B , the image analysis apparatus 2000 may correct a reference boundary. Specifically, the second controller 2002 may modify the reference boundary so that the modified reference boundary is located between the area labeled as the uninterested area and the area labeled as the reference area. Here, the reference boundary may be positioned to internally divide the interval between the reference region and the non-interested region at a predetermined ratio. Here, the predetermined ratio may be set in consideration of various factors. For example, the predetermined ratio may be determined in consideration of the distance between the non-interested region and the reference region. That is, the second controller 2002 controls various factors that can be considered in the medical image, including the distance between the reference region and the non-interest region, the position of the reference boundary overlapping the non-interest region, and/or the shape of the reference region. The reference boundary can be modified by reflecting it.

To give a specific example with reference to the drawings, in the case of FIG. 17A , the distance between the non-interest region (gray matter) and the reference region (ventricle) is relatively close, so that the reference boundary overlaps the non-interest region in a relatively large portion. is shown. In this case, in consideration of the distance between the non-interested region and the reference region, the second controller sets the reference boundary so that the modified reference boundary is located in the middle of the gray matter region and the ventricular region (to internalize the distance between the ventricle and the gray matter by 5:5). can be modified.

In addition, in the case of Figure 17b, the distance between the non-interest region (gray matter) and the reference region (ventricle) is relatively long, it is shown that the reference boundary overlaps only a relatively small portion of the non-interest region. In this case, the second controller 2002 may modify the reference boundary so that the modified reference boundary is closer to the non-interested region.

Also, according to an embodiment, a reference boundary determined in one medical image may overlap a plurality of non-interested regions. That is, for example, situations such as FIGS. 17(a) and 17(b) may occur within one medical image.

In this case, the second controller 2002 may correct the reference boundary so that the first interval and the second interval overlapping the non-interested area within the reference boundary have different ratios. Exemplarily, the second controller 2002 may modify the reference boundary as described above with reference to FIG. 17A in the first interval, and may modify the reference boundary as described above with reference to FIG. 17B in the second interval.

In other words, in the second controller 2002, at a position where the distance between the non-interest region and the reference region is the first distance, the ratio of the distance from the reference region to the reference boundary to the distance from the reference boundary to the non-interest region is set to be the first value, and at a position where the distance between the non-interest region and the reference region is the second distance, the ratio of the distance from the reference region to the reference boundary to the distance from the reference boundary to the non-interest region is the second value The reference boundary can be modified as much as possible. Here, when the second distance is greater than the first distance, the reference boundary may be modified such that the second value is smaller than the first value. For example, as the distance between the reference region and the non-interest region is relatively increased, the corrected reference boundary may be more biased toward the non-interest region.

As a specific example, the second controller 2002 corrects the reference boundary so that the distance between the gray matter and the ventricle is internally divided by 5:5 in the case of the first region where the distance between the gray matter and the ventricle is relatively close, and the gray matter and the In the case of the second region where the distance between the ventricles is relatively long, the reference boundary can be modified to internally divide the distance between the gray matter and the ventricles by 3:7. Here, the ratio of internalizing the distance between the gray matter and the ventricle may be preset.

Also, the second controller 2002 may correct the reference boundary only for an area where the reference boundary and the non-interested area overlap, or may correct the entire reference boundary. In addition, the second controller 2002 may correct the reference boundary for an additional buffer region in addition to the region where the reference boundary and the gray matter region overlap. This will be described in detail below with reference to FIG. 18 .

18 illustrates an example of a partially modified reference boundary and a fully modified reference boundary according to an embodiment.

Referring to FIG. 18 , the image analysis apparatus 2000 may correct all or part of a reference boundary. Specifically, the second controller 2002 may correct all or part of the reference boundary in consideration of the relationship between the reference region and the non-interested region.

For example, if the distance between the reference region and the non-interest region is close only in a specific portion, the reference boundary may be modified only in a partial section of the reference boundary corresponding to the specific portion, and the distance between the non-interest region and the reference region as a whole If is measured similarly, it would be possible to modify the reference boundary as a whole. However, this is only an example, and the situation in which the reference boundary is fully or partially modified may vary.

It will be described by way of example with reference to the drawings. In the drawing, the reference boundary modified entirely and the reference boundary partially modified are shown.

First, let's look at an example in which the reference boundary has been modified in general.

The image analysis apparatus 2000 may correct the reference boundary as a whole. Specifically, the second controller 2002 may entirely modify the reference boundary in consideration of the distance between the reference region and the non-interested region.

The second controller 2002 may modify all of the reference boundaries so as to correspond to the correction of the reference boundary in a section in which the reference boundary and the non-interest region overlap.

Alternatively, the second controller 2002 may correct the reference boundary in consideration of the distance between the reference region and the non-interest region in all sections of the reference boundary as well as the section in which the reference boundary and the non-interest region overlap.

For example, the second controller 2002 considers the distance between the non-interest region and the reference region in the first region, and sets the reference boundary to correspond to a position obtained by internally dividing the distance between the non-interest region and the reference region in the first region. It can also be expressed that it can be modified in general.

Also, for all sections of the reference boundary, the second controller 2002 may generally modify the reference boundary by considering the distance between the overall area of the reference area and the non-interested area.

Also, the image analysis apparatus 2000 may correct only a part of the reference boundary. Specifically, the second controller 2002 may correct the reference boundary only for a partial section corresponding to a section in which the reference boundary and the non-interested region overlap. Here, the second controller 2002 may correct the reference boundary even for a portion in which a buffer section is added to a section in which the reference boundary and the non-interested region overlap, or the reference boundary only for a section in which some sections are removed.

In other words, the second controller 2002 determines the reference boundary so that the reference boundary and the reference region have a first distance for a section in which the distance between the reference boundary and the non-interest area is closer than a predetermined distance, and the reference boundary For a section in which the and non-interested regions are farther than a predetermined distance, the reference boundary may be determined such that the reference boundary and the reference region have a second distance. Here, the predetermined distance may correspond to the distance between the reference boundary and the reference region described above with reference to FIGS. 10 to 14 .

Referring to the drawings, the partially modified reference boundary, the fully modified reference boundary 1, and the fully modified reference boundary 2 all have similar shapes in the section where the non-interested area (gray matter) and the reference area are located close to each other. However, it is well shown that the overall corrected reference boundary 1 is separated from the reference region by the same distance even in a section in which the distance between the non-interest region and the reference region is increased. On the other hand, in the reference boundary 2 that has been modified as a whole, it was determined that the reference boundary is located at a position where the non-interested area and the reference area are internally divided by a predetermined ratio in all sections of the reference boundary.

However, even in the case of a partially modified reference boundary, it is necessary to smoothly connect an area in which the reference boundary is modified and the existing reference boundary so that the modified reference boundary maintains a smooth shape.

Hereinafter, when a part of the reference boundary is corrected, a connection relationship between the existing reference boundary and the modified reference boundary will be described with reference to the drawings.

19 illustrates an example of a specific modification process of a partially modified reference boundary, according to an embodiment.

Referring to FIG. 19 , the image analysis apparatus 2000 may connect the modified reference boundary with the existing reference boundary. Specifically, the second controller 2002 may determine a range in which the reference boundary is to be corrected in consideration of the distance between the reference region and the non-interested region, and connect the existing reference boundary from some regions of the modified reference boundary.

First, the second controller 2002 controls the reference boundary from the first point at which the distance between the reference area and the non-interest area is equal to or less than the threshold distance to the second point at which the distance between the reference area and the non-interest area is equal to or less than the threshold distance. can be modified. However, this is an example, and the first point may be selected as an arbitrary point where the reference boundary and the non-interest region overlap, or an arbitrary point where the distance between the non-interest region and the reference region is less than or equal to a predetermined distance. In addition, as for the method of correcting the reference boundary, the above-described embodiments or other methods integrated with the spirit of the present invention may be used.

Thereafter, the second controller 2002 may connect the first point or the second point, which is the end point of the region in which the reference boundary is modified, and the third point, which is a section in which the existing reference boundary starts again. Here, the distance between the third point and the reference region according to the existing reference boundary may be predetermined as described above with reference to FIGS. 10 to 14 .

The second controller 2002 may connect the first point or the second point and the third point by interpolating between the position value of the first point or the second point and the position value of the third point. Here, the length of the interpolation section between the second point and the third point may be variously set. As an example, the interpolation period may be set in consideration of the distance between the non-ROI and the ROI.

With reference to the drawings, a specific example will be described.

The second controller 2002 may partially modify the reference boundary for a section in which the distance between the gray matter and the ventricle is close or the existing reference boundary overlaps the gray matter region. The second controller 2002 may interpolate between a first point located on one side of the modified reference boundary and a third point located on the existing reference boundary to connect them to each other. Here, the second point may be selected as a point at which the distance between the gray matter and the ventricle becomes more than a critical distance. Alternatively, the second point may be selected as a point at which the corrected reference boundary and the uninterested region are equal to or greater than a threshold distance.

As such, the image analysis apparatus 2000 according to an embodiment can more clearly set the reference boundary for calculating the disease index by smoothly connecting the existing reference boundary and the modified reference boundary, and use the clearly set reference boundary. Accordingly, the accuracy of the disease index obtained by the image analysis apparatus 2000 may also be improved.

When the disease index is calculated by the modified reference boundary, the image analysis apparatus 2000 may provide diagnosis auxiliary information based on the disease index obtained using the modified reference boundary. Specifically, the second controller 2002 may obtain a disease index as a result of analysis of the medical image using the modified reference boundary, and may provide diagnostic assistance information based on the acquired disease index. Since the provided disease index or disease information is similar to that described above with reference to FIG. 14 , a detailed description thereof will be omitted.

As described above, by using the modified reference boundary, it is possible to obtain an accurate disease index from more diverse images than in the case of using a uniform reference boundary.

In addition, in the above description, it has been described that the reference boundary is corrected only when the reference boundary overlaps the gray matter region, but it is also possible to set the reference boundary as described in FIGS. 15 to 19 from the beginning. Depending on the bar, an appropriate reference boundary setting method may be adopted and used.

In the above description of the present specification, a method of analyzing one medical image has been mainly described.

Hereinafter, when the image analysis apparatus 2000 analyzes a plurality of medical images, a method of providing diagnosis auxiliary information from the plurality of medical images will be described.

20 is a schematic flowchart of a method for the image analysis apparatus 2000 to provide diagnostic assistance information from a plurality of medical images, according to an exemplary embodiment.

Referring to FIG. 20 , a method of providing diagnostic auxiliary information from a plurality of medical images includes obtaining a plurality of medical images ( S3200 ), obtaining a disease index from the plurality of medical images ( S3400 ), and a plurality of disease indexes It may include a step (S3600) of providing diagnostic auxiliary information based on the .

First, the image analysis apparatus 2000 may acquire a plurality of medical images ( S3200 ). Specifically, the second controller 2002 may acquire a plurality of medical images from the image acquisition apparatus 1000 through the second communication module 2800 . Here, the plurality of medical images may be a set of tomographic images obtained by photographing the object to be photographed. For example, the plurality of medical images may be a set of MRI slice images of an object to be photographed.

Thereafter, the image analysis apparatus 2000 may obtain a disease index from a plurality of medical images (S3400). Specifically, the second controller 2002 may acquire disease indices for a plurality of medical images by using a program for acquiring disease indices stored in the second memory 2400 . Here, the second controller 2002 may obtain a disease index for each of a plurality of medical images, or may acquire a disease index for some of the plurality of medical images. Since the method of acquiring the disease index is similar to that described above, a detailed description thereof will be omitted.

When a plurality of disease indices are obtained, the image analysis apparatus 2000 may provide diagnosis auxiliary information based on the plurality of disease indices ( S3600 ). In detail, the second controller 2002 may comprehensively consider disease indexes related to a plurality of acquired medical images to acquire and provide diagnostic auxiliary information. Here, the second controller 2002 may provide diagnostic auxiliary information for all of the plurality of medical images for which the disease index is calculated, or may provide diagnostic auxiliary information for only some of the plurality of medical images for which the disease index is calculated. have. A method in which the image analysis apparatus 2000 provides diagnostic auxiliary information for a plurality of medical images will be described later in detail.

Hereinafter, an example of a plurality of medical images and a method of providing diagnostic assistance information based on the plurality of medical images according to an embodiment will be described with reference to the drawings.

21 illustrates an example of a plurality of medical images according to an embodiment.

Referring to FIG. 21 , the image analysis apparatus 2000 may acquire a plurality of tomographic images related to an object to be photographed. Specifically, the second controller 2002 may acquire a plurality of medical images including images of a plurality of cross-sections of the object to be photographed through the second communication module 2800 . Here, as described above, the plurality of tomographic images may be composed of a set of slice images.

21 shows an example of a plurality of medical images acquired by the MRI imaging apparatus.

Referring to FIG. 21 , the plurality of medical images includes slice images of a plurality of planes parallel to a first axis of an object to be photographed. In addition, each slice image may include information about the object to be photographed at positions spaced apart from each other by a predetermined distance. That is, each slice image may include different information about the object to be photographed. As such, since the plurality of medical images comprehensively include information on various cross-sections of the object to be photographed, the image analysis apparatus 2000 can obtain diagnostic auxiliary information about the object to be photographed overall by analyzing the plurality of medical images. have.

22 is an example of slice images each including different information and an example of providing diagnostic auxiliary information through this example.

Referring to FIG. 22 , an example of a plurality of medical images obtained by photographing different cross-sections for one object to be photographed is illustrated. Specifically, FIG. 22 shows a set of several slice images of an MRI image of a human brain.

In case of (a), an MRI slice image of a cross-section of the upper side of the brain is shown, and in case of (b), an MRI slice image of a cross-section of the middle side of the brain is shown.

(a) is an image of a cross-section of a region located in the upper part of the brain compared to (b), and the size of the ventricle was taken smaller than (b). In addition, the white matter region and the gray matter region included in (a) and (b) are also shown differently, and most importantly, it can be seen that the region information differentiated by WMH is included differently.

As described above, the disease index according to an embodiment may be obtained based on the relation between the reference region and the region information of the target region. Accordingly, different disease indices may be calculated in each of a plurality of medical images including cross-sections at various positions of the object to be photographed. As a specific example, in the case of (a), a cross-section of the upper side of the brain was taken, a disease index of grade 3 was calculated, and in the case of (b) of a cross-section of the middle side of the brain, a disease index of grade 2 was calculated.

In this case, both the auxiliary diagnosis information regarding (a) and the auxiliary diagnosis information regarding (b) can be obtained, and the image analysis apparatus 2000 provides diagnostic auxiliary information for all images or auxiliary diagnosis information for some images. can be obtained and provided.

That is, the image analysis apparatus 2000 according to an embodiment may calculate a disease index for a plurality of medical images obtained by photographing multiple cross-sections of an object to be photographed, and may provide diagnostic assistance information based on the calculated disease index. can

Also, the image analysis apparatus 2000 may provide the diagnosis auxiliary information with respect to all the medical images for which the disease index is calculated, or may provide the diagnosis auxiliary information with respect to only some of the medical images.

The image analysis apparatus 2000 may select some images from among the medical images for which the disease index is calculated, and may provide diagnostic assistance information for the selected medical images. Specifically, the second controller 2002 selects an image including predetermined information from among the medical images for which the diagnostic index is calculated, calculates a disease index for the image including the predetermined information, and provides diagnostic auxiliary information. can provide

Here, the criteria for selecting a medical image to which the diagnosis auxiliary information is to be provided may vary. That is, the image analysis apparatus 2000 may select a medical image to provide diagnostic assistance information based on the type or grade of the disease index.

For example, the image analysis apparatus 2000 may provide diagnostic assistance information on a medical image that is a basis for calculating a grade of a disease index indicating the most severe disease progression among medical images for which the disease index is calculated. . As another example, the image analysis apparatus 2000 may provide diagnostic assistance information on the medical image that is the basis for calculating the grade of the disease index indicating that the state of the disease is the most improved among the medical images for which the disease index is calculated. As another example, the image analysis apparatus 2000 is a medical image based on the calculation of the grade of the disease index representing the average value (or the median value) of the disease index calculated from a plurality of medical images among the medical images for which the disease index is calculated. It is also possible to provide diagnostic auxiliary information for

As another example, the image analysis apparatus 2000 may provide diagnostic auxiliary information for a medical image having a high first disease index in consideration of the type of the disease index, and a medical image having a high second disease index. It is also possible to provide diagnostic auxiliary information for the , or a combination thereof to provide diagnostic auxiliary information. Also, it goes without saying that the image analysis apparatus 2000 may select a medical image to provide diagnostic assistance information in consideration of both the type and grade of the disease index.

As such, since different diagnostic auxiliary information may be calculated according to the type or grade of the disease index, the image analysis apparatus 2000 considers the characteristics of the object to be photographed (for example, mainly in the case of a patient), and the type or grade of the disease index By selecting a medical image to provide diagnostic auxiliary information in consideration of

As another embodiment, the image analysis apparatus 2000 may select a candidate image for calculating a disease index from a plurality of medical images, and provide diagnosis auxiliary information based on the disease index obtained from the selected candidate images.

Hereinafter, a method in which the image analysis apparatus 2000 provides diagnostic auxiliary information based on a disease index obtained from candidate images will be described with reference to the drawings.

23 is a flowchart schematically illustrating a method of obtaining diagnostic assistance information from a candidate image according to an exemplary embodiment.

Referring to FIG. 23 , the method for the image analysis apparatus 2000 to obtain diagnostic auxiliary information from a candidate image includes the steps of acquiring a plurality of medical images (S1202), segmenting the plurality of medical images (S1212), and the segmented Determining a candidate image from a plurality of images (S1222), obtaining a disease index for the selected candidate image (S1402), and providing diagnostic auxiliary information based on the obtained disease index (S1602). have.

First, the image analysis apparatus 2000 may acquire a plurality of medical images from the image acquisition apparatus 1000 ( S1202 ). Since the step of acquiring a plurality of medical images ( S1202 ) is similar to that described with reference to FIG. 20 , a detailed description thereof will be omitted.

Thereafter, the image analysis apparatus 2000 may segment a plurality of medical images (S1212). Segmentation of the medical image may also be performed similarly to the above, so a detailed description thereof will be omitted.

When segmentation is performed, the image analysis apparatus 2000 may determine a candidate image from a plurality of segmented medical images (S1222). Specifically, the second controller 2002 may determine candidate images satisfying the predetermined criterion from among a plurality of segmented medical images according to a predetermined criterion. A detailed description of the candidate image will be described later with reference to the drawings.

When the candidate image is determined, the image analysis apparatus 2000 may obtain a disease index from the candidate image (S1402). Specifically, the second controller 2002 may obtain the disease index for all or some of the candidate images by using a program for calculating the disease index stored in the second memory 2002 . Since the method of obtaining the disease index may be performed similarly to the above, a detailed description thereof will be omitted.

When the disease index is obtained, the image analysis apparatus 2000 may obtain and provide diagnostic auxiliary information based on the acquired disease index (S1602). Specifically, the second controller 2002 may provide diagnostic assistance information based on a disease index obtained from one or more candidate images. Here, as for the method for the second controller 2002 to provide the auxiliary diagnosis information from the plurality of candidate images, as described with reference to FIGS. 20 to 22 , the method of providing the auxiliary diagnosis information from the plurality of medical images may be applied. have.

Hereinafter, an example in which a candidate image is extracted from a plurality of medical images, an example in which a disease index is calculated from the extracted candidate images, and an example in which the diagnosis auxiliary information is provided will be described with reference to the drawings.

24 is an example of a candidate image according to an embodiment.

Referring to FIG. 24 , the image analysis apparatus 2000 may select a candidate image. Specifically, the second controller 2002 may select one or more candidate images from among a plurality of medical images acquired from the image acquisition apparatus 1000 . Here, the candidate image may be an image satisfying a predetermined criterion among a plurality of medical images.

Here, the predetermined criterion may be determined to correspond to a disease index to be calculated. For example, the predetermined criterion may be whether the reference region is included. That is, the second controller 2002 may select an image including information about the reference region from among a plurality of medical images as a candidate image. As another example, the predetermined criterion may be whether the target area is included. That is, the second controller 2002 may select an image including information on the target region from among a plurality of medical images as a candidate image. Also, the second controller 2002 may select a medical image including both the target region and the reference region as a candidate image.

Also, here, the candidate image may be determined as one or more images included in the candidate image range as shown in the drawing. Here, the candidate image range includes a first medical image that satisfies a predetermined condition among tomographic medical images having a continuous sequence obtained by photographing the same object to be photographed in multiple planes about one axis, and a first medical image that satisfies the predetermined condition. It may mean a range between 2 medical. Here, the range between the first medical image and the second medical image may include a plurality of medical images sequentially positioned between the first medical image and the second medical image.

That is, to give a specific example with reference to the drawings, if the reference region is the ventricle region, among a plurality of medical images obtained by taking a transverse plane with respect to the saggital axis as a continuous cross-section, the lowermost medical image including the ventricle A range between the first medical image and the second medical image located at the top of the medical images including the ventricle may be set as the candidate image range.

In other words, the second controller 2002 may select the candidate image or the range of the candidate image based on whether the number of pixels corresponding to a preset condition included in the medical image is less than or equal to a threshold value.

Specifically, the second controller 2002 may analyze the segmentation result of the medical image to determine whether the number of pixels labeled with the ventricle is less than or equal to a threshold value. The second controller 2002 may determine the medical image as a candidate image when the number of pixels labeled with the ventricle included in the medical image is equal to or greater than a threshold value.

In addition, when the second controller 2002 sets the range of the candidate image, as a result of analyzing two adjacent medical images, the number of pixels labeled with the ventricle included in one image is less than or equal to the threshold value, and the number of pixels included in the other image is less than or equal to the threshold. If the number of pixels labeled with the ventricle is greater than or greater than the threshold, the other image may be set as one boundary of the range of candidate images.

25 is a diagram illustrating an example of providing diagnostic auxiliary information from a candidate image according to an exemplary embodiment.

Referring to FIG. 25 , the image analysis apparatus 2000 may provide diagnosis auxiliary information from a candidate image. Specifically, the second controller 2002 may obtain a disease index from the candidate image, and provide diagnostic auxiliary information based on the acquired disease index.

Since the method of calculating the disease index has been described above, a detailed description thereof will be omitted.

As shown in FIG. 25 , in case of (a), it is a medical image that is not a candidate image, and in case of (b), it is a medical image determined as a candidate image.

In calculating the disease index according to an embodiment, it may be important that a specific region of the object to be photographed is included in the medical image. For example, the disease index acquisition method according to an embodiment may be derived by the correlation between the ventricle and WMH. In this case, if the ventricle is not included in the medical image, it may be difficult to calculate the disease index, which Accurate diagnostic assistance information may not be provided. That is, in the case of (a) of FIG. 25 , it is a medical image obtained by photographing the upper part of the brain, and since the ventricular region is not included in the medical image, it may be difficult to use the disease index calculation method according to an embodiment.

In the case of (b), since the ventricular region is included in the medical image, the second controller 2002 can calculate the disease index by analyzing the medical image as in (b), and diagnosis auxiliary information based on the calculated disease index can be obtained.

In addition, when all medical images related to the object to be photographed are analyzed, the amount of computation of the image analysis apparatus 2000 may be excessive and the computation speed may be slowed, and thus an error may occur in the computation process.

However, according to the method for providing diagnostic assistance information using candidate images according to an embodiment, after segmenting a plurality of images, a candidate image for which a disease index is to be calculated is first determined, and a disease index is calculated only for the determined candidate images, so the amount of computation is By reducing

26 is another example of providing diagnostic auxiliary information according to an exemplary embodiment.

Referring to FIG. 26 , the image analysis apparatus 2000 according to an embodiment may provide diagnostic auxiliary information on at least one image among candidate images. Specifically, the second controller 2002 may provide diagnostic auxiliary information on at least one image satisfying a predetermined condition among medical images selected as the candidate image.

Here, the predetermined condition may be determined in consideration of the disease index. For example, the predetermined condition may be a disease index value indicating the most severe disease progression. That is, the second controller 2002 may select, as a medical image for providing diagnosis auxiliary information, a specific medical image indicating the progress of a disease having the most severe disease index among the disease indexes of the candidate images.

In addition, when a plurality of disease indices can be calculated from the medical image, the predetermined condition may be determined in consideration of all disease indices. That is, the second controller 2002 may select all of the specific candidate images for each of the plurality of disease indices as medical images for providing the diagnostic auxiliary information, or provide only the specific candidate images for some of the plurality of disease indices as the diagnostic auxiliary information. It can be selected as a medical image for

As a specific example with reference to the drawings, the second controller 2002 may provide, through the display module 2600, diagnostic auxiliary information on a medical image that satisfies a predetermined condition regarding the first disease index among a plurality of medical images. can Here, the predetermined condition may be whether the grade of the first disease index among the plurality of candidate images is calculated to be the highest. In addition, the second controller 2002 may provide, through the display module 2600 , diagnostic auxiliary information on a medical image that satisfies a predetermined condition regarding a second disease index among a plurality of medical images through the display module 2600 , It is also possible to provide diagnostic auxiliary information for both medical images that satisfy a predetermined condition regarding the index and the second disease index.

In the above, a method for determining a candidate image for calculating a disease index from a plurality of medical images and a method for calculating a disease index for one medical image have been mainly described.

However, according to an embodiment, the disease index may be calculated using a plurality of medical images in order to more accurately calculate the disease index, which will be described with reference to FIGS. 27 to 29 .

27 is a flowchart illustrating a method of obtaining a disease index based on a plurality of images according to an embodiment.

Referring to FIG. 27 , a method for obtaining a disease index based on a plurality of images according to an embodiment includes obtaining a plurality of medical images (S4200), and performing segmentation on the plurality of medical images (S4400). , obtaining the target area information of the first image (S4600) and obtaining the disease index of the second image based on the target area information of the first image (S4800).

First, acquiring a plurality of medical images (S4200) and performing segmentation on the plurality of medical images (S4400) may be performed similarly to the above-described operation, and thus a detailed description thereof will be omitted.

After the segmentation is performed on the medical image, the image analysis apparatus 2000 may acquire information on the target region of the first image ( S4600 ). Specifically, the second controller 2002 may acquire information on the target region corresponding to the disease index to be calculated. Here, the target region may mean a region essential for calculating a specific disease index. Specific details will be described later.

When information on the target region is obtained, the image analysis apparatus 2000 may calculate a disease index of the second image based on the target region information of the first image. Specifically, the second controller 2002 may calculate the disease index related to the second image based on the target area information obtained from the first image and the area information related to the object to be photographed included in the second image. Examples related thereto will be described later.

Hereinafter, a method of calculating the disease index of the second image in consideration of the information of the first image with reference to the drawings will be described with reference to a specific example.

28 is an example illustrating a process of calculating a disease index of a second image in consideration of information on a first image according to an embodiment.

According to an embodiment, the image analysis apparatus 2000 may calculate the disease index of the second image directly from the first image. Specifically, the second controller 2002 may calculate the disease index of the second image based on the relationship between the first region information included in the first image and the second region information included in the second image. Here, the first area information may be information about the reference area included in the first image. Also, the second area information may be information about the target area included in the second image.

In calculating the disease index, information on the reference region may not be included in the medical image, but information on the target region may be included. In this case, it may be difficult to calculate the disease index because the requirements for calculating the disease index are not met, but in some cases, it is necessary to calculate the disease index even for these medical images.

That is, illustratively referring to FIG. 28 , in the case of the first slice image, information about WMH is included, but information about the ventricle is not included. In this case, the disease index derived from the relationship between the ventricle and WMH may be difficult to derive from the first slice image. However, since WMH is likely to contain important information on disease, it is desirable to calculate a disease index for the image in which WMH was found in order to provide more accurate diagnostic auxiliary information.

In this case, the second controller 2002 may calculate the disease index of the second image in consideration of information on the first image based on the relationship between the first image and the second image. Here, the relationship between the first image and the second image may be obtained in advance. For example, the relationship between the first image and the second image may be a distance between slice images.

With reference to the drawings again, specific examples will be described.

The second controller 2002 may first determine, as a result of segmentation of a plurality of medical images, a first image in which a target region for calculating a disease index is included in the medical image, but a reference region is not included. For example, the second controller 2002 includes only information about WMH in the medical image, even though information about the ventricle and WMH is required to calculate the disease index according to an embodiment, and information about the ventricle is A first image that is not included may be selected.

Thereafter, the second controller 2002 may select a second image including a reference region not included in the first image from among other medical images near the first image. Here, the second image may be an image immediately adjacent to the first image, but is not limited thereto. For example, the second controller 2002 may be configured to calculate a disease index among medical images located around the first image (eg, a distance from the first image in the sagittal axis is less than or equal to a predetermined distance). A second image containing the ventricular region may be selected.

When the second image is selected, the second controller 2002 may calculate a disease index based on the relationship between the reference region included in the second image and the target region included in the first image. For example, the second controller 2002 determines the disease in the second image based on distance information from the ventricular region included in the second image to the WMH included in the first image and area information of the WMH included in the second image. index can be calculated. Here, the distance between the first image and the second image may be obtained and stored in advance.

Also, here, a criterion for deriving a relationship between the reference region included in the second image and the target region included in the first image may be variously determined. As an example, the second controller 2002 may calculate the disease index based on the distance from the center of the ventricular region included in the second image to the WMH region included in the first image. As another example, the second controller 2002 is a disease index based on the distance from the ventricular region closest to the WMH region included in the first image among the ventricular regions included in the second image to the WMH region included in the first image. can also be calculated. In addition to the methods mentioned by way of example, if it is a criterion for calculating the distance from the ventricular region included in the second image to the WMH region included in the first image, it is understood by those skilled in the art that various methods can be incorporated into the spirit of the present invention. it will be understood

Hereinafter, a method of calculating the disease index of the second image indirectly from the first image will be described with reference to the drawings.

29 is another example illustrating a process of calculating a disease index of a second image in consideration of information on a first image according to an embodiment.

According to an embodiment, the image analysis apparatus 2000 may indirectly calculate the disease index of the second image from the first image. Specifically, the second controller 2002 determines the second image based on the relationship between the reference cell in the second image and the target region included in the second image, which is indirectly obtained from information about the reference region included in the first image. can calculate the disease index of

With reference to the drawings, a specific example will be described.

As described with reference to FIG. 27 , first, as a result of segmentation on a plurality of medical images, the second controller 2002 includes a target region for calculating a disease index in the medical image, but does not include a reference region. A first image may be determined.

For example, the second controller 2002 includes only information about WMH in the medical image, even though information about the ventricle and WMH is required to calculate the disease index according to an embodiment, and information about the ventricle is A first image that is not included may be selected.

Thereafter, the second controller 2002 may select a second image including a reference region not included in the first image from among other medical images adjacent to the first image. For example, the second controller 2002 may select a second image including a ventricular region for calculating a disease index from among medical images located near the first image.

When the second image is selected, the second controller 2002 may determine a reference cell corresponding to the reference region included in the second image on the first image based on information on the reference region included in the second image. . For example, the second controller 2002 may determine a reference cell corresponding to the ventricular region included in the second image on the first image, based on information about the ventricular region included in the second image. have.

When the reference cell is determined, the second controller 2002 may calculate a disease index based on information about the reference cell and the target region included in the first image. That is, for example, the second controller 2002 considers the reference cell of the second region corresponding to the ventricular region included in the first image as the ventricle, and derives the relationship with the WMH included in the first image to determine the disease index. it can be calculated As for the method of calculating the disease index based on the relationship between the reference cell and the WMH, it goes without saying that a method of calculating the disease index similar to that described above may be used.

As described above, according to an exemplary embodiment, the image analysis apparatus 2000 may calculate a disease index by using a plurality of medical images to obtain an accurate disease index, and may provide diagnosis auxiliary information based on this.

In addition, the diagnostic assistance information providing system 10000 according to an exemplary embodiment may calculate a disease index in a 3D medical model obtained by 3D modeling a medical image, and provide diagnostic assistance information.

Hereinafter, a method of providing diagnostic assistance information based on a 3D medical model will be described with reference to FIGS. 30 to 32 .

30 is a flowchart illustrating an example of a method of providing diagnostic assistance information based on a 3D medical model according to an embodiment.

According to an embodiment, a method of providing diagnostic auxiliary information based on a 3D medical model performed by the image analysis apparatus 2000 includes acquiring a plurality of medical images (S5200), and performing segmentation on the plurality of medical images. It may include a step (S5400), a step of performing 3D modeling based on the segmented medical image (S5600), and a step of obtaining a disease index based on the 3D modeling (S5800).

The steps of obtaining a plurality of medical images ( S5200 ) and performing segmentation on the plurality of medical images ( S5400 ) may be performed similarly to those described above, and thus a detailed description thereof will be omitted.

When segmentation is performed on a plurality of medical images, the image analysis apparatus 2000 may acquire a 3D medical model based on the plurality of segmented medical images (S5600). Specifically, the second controller 2002 may generate a 3D medical model by processing a plurality of medical images obtained by photographing an object to be photographed as a continuous cross-section. Here, the 3D medical model may include 3D pixel information that is generated based on a plurality of segmented medical images and reflects information on a plurality of brain regions. That is, it may be segmented by a plurality of brain regions in 3D. In addition, the aforementioned reference boundary may be formed in three dimensions. (Explain appropriately: 3D pixel information corresponding to each of a plurality of brain regions, including boundaries, etc.) can be generated.

When the 3D medical model is obtained, the image analysis apparatus 2000 may obtain a disease index based on the obtained 3D medical model and provide diagnosis auxiliary information. Specifically, the second controller 2002 may calculate the disease index from the 3D medical model using the program for acquiring the disease index stored in the second memory 2400 , and obtain diagnostic auxiliary information from the calculated disease index. .

An example of calculating a disease index from a 3D medical model and providing diagnostic auxiliary information will be described below with reference to the drawings.

31 shows an example of a process of calculating a disease index from a 3D medical model according to an embodiment, and FIG. 32 shows another example of a process of calculating a disease index according to an embodiment.

Referring to FIG. 31 , the image analysis apparatus 2000 may acquire a disease index based on region information included in the 3D medical model. Specifically, the second controller 2002 may calculate the disease index based on information on one or more specific regions of the object to be photographed included in the 3D medical model. Here, the one or more specific regions may include the target region and the reference region as described above. Specifically, it may be a cell (eg, a voxel) labeled with a target region or a reference region expressed in 3D. Also, here, the second controller 2002 may set a criterion for deriving a relationship between regions included in one or more to-be-photographed objects.

As an example, the second controller 2002 may calculate the disease index based on the ventricular region and the WMH region included in the 3D medical model. Here, the second controller 2002 may set a reference boundary in order to derive a relationship between the ventricular region and the WMH region, similarly to the above-described example. That is, the second controller 2002 may set a set of cells (cells) separated by a preset distance from the ventricular region in three dimensions as a reference boundary. Also, similarly to the reference boundary set in the medical image described above, the second controller 2002 may set or modify the reference boundary in various ways.

The second controller 2002 may calculate the disease index in consideration of the relationship with the WMH region based on the reference boundary formed in 3D. Here, the relationship between the reference boundary and the WMH is similar to that described above. In addition, as three-dimensional 3D modeling is performed on a two-dimensional medical image, information on the area of the object to be photographed, for example, area information on the target area, may be a three-dimensional physical quantity. That is, the second controller 2002 may acquire even three-dimensional information such as the thickness of a specific region included in the object to be photographed, which could not be obtained in the two-dimensional medical image, through 3D modeling. As another example, information on WMH may be a volume (or number) corresponding to cells tagged with WMH.

Referring to FIG. 32 , the image analysis apparatus 2000 may calculate a disease index based on a plurality of planes included in the 3D medical model and provide diagnosis auxiliary information. Specifically, the second controller 2002 calculates a disease index based on one or more planes including information on at least one specific region related to an object to be photographed included in the 3D medical model, and provides diagnostic auxiliary information. can do.

As 3D modeling is performed, there is an advantage in that it becomes easy to derive a relationship between specific regions for calculating a disease index. That is, when calculating the disease index by analyzing the slice image as described above, if the slice image does not contain information on the essential region included in the object to be photographed for calculating the disease index, the disease index can be easily obtained. Not only that, but the accuracy of the acquired disease index may be lowered.

However, when calculating the disease index by performing 3D modeling, the accuracy of the disease index may be improved by determining a plane to include all regions essential for calculating the disease index and calculating the disease index with respect to the corresponding plane.

It will be described as a specific example with reference to the drawings.

Referring back to FIG. 32 , the second controller 2002 sets the first plane to correspond to a preset criterion to include both the ventricular region and the first WMH region, and calculates the disease index for the set first plane. have. Here, the criteria for setting the first plane may be various. For example, the first plane may be a plane set in consideration of the distance between the target region and the reference region. As a more specific example, the first plane may be a plane set to include a straight line in which the distance between the WMH region and the ventricular region is the minimum.

As another example, the second controller 2002 may set the second plane in consideration of the number of target regions included in the plane, and calculate a disease index for the set second plane. For example, the second controller 2002 may set the second plane to include the largest number of target areas in addition to including information about the reference area. Also, the second controller 2002 may set the second plane so that the number of pixels labeled as the target area is maximum.

That is, the second controller 2002 may set the first plane to include the first target region and the reference region, or set the second plane to include the second target region and the reference region, according to various criteria. In this case, the first plane and the second plane may form a predetermined angle. As such, the image analysis apparatus 2000 may derive a relationship between the reference region and the target region even for a plane that forms a predetermined angle with the plane including the medical image serving as the base of the 3D medical model, and calculates the disease index or provide diagnostic assistance information.

In the above, according to various embodiments, an example of a method of obtaining a disease index from one or a plurality of medical images performed by the diagnostic assistance information providing system 10000 and providing diagnostic assistance information based on the acquired disease index was explained about.

As described above, according to the diagnostic auxiliary information providing method performed by the diagnostic auxiliary information providing system 10000 according to an embodiment, a disease index based on an automated algorithm or program with a clear and uniform standard from a medical image By calculating , a more objective disease index can be calculated by excluding human subjective judgment, and thus more accurate diagnostic auxiliary information can be provided.

Hereinafter, a method in which the diagnostic assistance information providing system 10000 performs segmentation on specific regions included in the medical image from the medical image according to an exemplary embodiment will be described.

According to an embodiment, the diagnostic assistance information providing system 10000 may segment the medical image. Specifically, the image analysis apparatus 2000 may obtain a medical image from the image acquisition apparatus 1000 and segment the obtained medical image. More specifically, the second controller 2002 may segment the medical image using an image segmentation algorithm stored in the second memory 2400 .

Hereinafter, it will be described with reference to the drawings.

33 schematically illustrates a segmentation operation of a medical image performed by an image analysis apparatus according to an exemplary embodiment.

According to an embodiment, the image analysis apparatus 2000 may receive input data and output output data. Specifically, the second controller 2002 may input the medical image acquired from the image acquisition apparatus 1000 as input data to an algorithm for image segmentation, and may acquire output data in which the medical image is segmented.

The image analysis apparatus 2000 according to an embodiment may use various algorithms for image segmentation.

As an example, an algorithm for image segmentation may be provided as a machine learning model. A representative example of a machine learning model may include an artificial neural network. Specifically, a representative example of an artificial neural network is a deep learning-based artificial neural network that includes an input layer that receives data, an output layer that outputs a result, and a hidden layer that processes data between the input and output layers. .

Detailed examples of the artificial neural network include a regression analysis artificial neural network, a recurrent neural network, a deep neural network, and the like. It should be interpreted as a comprehensive meaning that includes both artificial neural networks in the form of artificial neural networks and artificial neural networks in the form of a combination thereof, and does not necessarily have to be a deep learning series.

In addition, the machine learning model does not necessarily have to be in the form of an artificial neural network model, and may include nearest neighbor algorithm (KNN), random forest, support vector machine (SVM), principal component analysis (PCA), etc. , may include all of the above-mentioned techniques in an ensemble form or in a form in which other methods are combined in various ways. On the other hand, it is stated in advance that the artificial neural network can be replaced with another machine learning model unless otherwise specified in the embodiments mentioned mainly with the artificial neural network.

Furthermore, in the present specification, an algorithm for image segmentation is not necessarily limited to a machine learning model. That is, the algorithm for image segmentation may include various judgment/decision algorithms rather than a machine learning model.

Therefore, it is clarified in advance that the algorithm for image segmentation in the present specification should be understood as a comprehensive meaning including all types of algorithms that analyze a medical image and distinguish a region included in the medical image.

An example of implementation of an artificial neural network model will be described with reference to FIGS. 34 and 35 .

34 shows an embodiment of an artificial neural network model according to an embodiment.

Referring to FIG. 34 , the image analysis apparatus 2000 according to an embodiment of the present application may utilize U-net as an artificial neural network for image segmentation.

U-net, which is widely used for image segmentation, may have an architecture including a contraction path and an expansion path.

Specifically, the contraction path of the U-net may be configured such that two convolutions and max pooling are successively performed. In this case, image-related characteristics may be extracted from the contraction path of the U-net.

However, since the size of the characteristic map is also reduced in the constricted path, the U-net may be configured to recover the size of the characteristic map by additionally including the expansion path.

The extension path of the U-net may be configured such that up-convolution and two convolutions are successively performed. In this case, the size of the image and the characteristic map may be extracted from the extension path of U-net.

Additionally, the U-net architecture is configured to concatenate the characteristic map of the same level, so that it is possible to provide positional information related to the characteristic from the contraction path to the expansion path.

At this time, based on the difference between the label of the input image and the label of the output segmentation map, the parameter of at least one node of the included layer of U-net so that the difference between the label of the input image and the label of the output segmentation map is minimized Or you can adjust the weight.

35 shows another implementation of an artificial neural network model according to an embodiment.

Referring to FIG. 35 , the image analysis apparatus 2000 according to an embodiment may utilize U-net++ as an artificial neural network for image segmentation. U-net++ is an artificial neural network model using the high-density block idea of DenseNet to improve the performance of U-net. A convolutional layer exists in the skip path to bridge the semantic gap between the encoder and decoder characteristic maps, and to the skip path. It differs from U-net in that dense skip connections exist to improve gradient flow.

Specifically, the image analysis apparatus 2000 may be implemented to input an input image to the input layer of the U-net++ neural network model and obtain label information output through the output layer. In this case, the image analysis apparatus 2000 may adjust the parameter or weight of at least one node of the hidden layer included in Unet++ based on the difference between the label information included in the input image and the label information output from the neural network model.

More specifically, the second controller 2002 is implemented to repeatedly perform the above-described operation of adjusting the parameters or weights of at least one node, so that the difference between the label information included in the input image and the label information output from the neural network model It is possible to obtain the parameter or weight of the node in which n is minimized.

36 exemplarily shows a segmentation result using an artificial neural network model according to an embodiment.

According to an embodiment, the image analysis apparatus 2000 may segment the medical image based on characteristics of the medical image. Specifically, the second controller 2002 may segment the medical image so as to correspond to a condition in which the medical image is obtained.

Here, the condition under which the medical image is obtained may vary according to a setting parameter of the image acquisition apparatus 1000 . For example, the condition under which the medical image is obtained may be a preset magnetic condition parameter value of the image acquisition apparatus 1000 . Specifically, when the image acquisition apparatus 1000 is implemented as an MRI apparatus, a T1-weighted image or a T2-weighted image may be acquired according to the setting of the TR/TE value, and the second controller 2002 may display the acquired image. The image can be segmented to correspond to . Here, the characteristics according to the acquisition conditions of the medical image are the same as described above.

With reference to the drawings, a specific example will be described.

Referring to FIG. 36 , a segmentation result of a T1-weighted image and a T2- FLAIR image obtained from the same object to be photographed is shown.

According to an embodiment, the second controller 2002 may segment the T1-weighted image. Here, the T1-weighted image may be segmented to determine anatomical characteristics. As shown in the figure, it can be seen that the segmentation result of the T1-weighted image shows that the wrinkle part of the outer gray matter region is segmented more clearly than the result of segmenting the T2-FLAIR image.

Also, according to an embodiment, the second controller 2002 may segment the T2-FLAIR image. Here, the T2-FLAIR image may be segmented to determine the lesion characteristics. As shown in the figure, it is shown that the segmentation result of the T2-FLAIR image further includes information on the WMH region compared to the segmentation result of the T1-weighted image.

As such, even in the case of a medical image acquired from the same object to be photographed, the characteristics (eg, anatomical characteristics, lesion characteristics) included in the medical image vary according to the acquisition conditions of the image acquisition apparatus 1000 , and the characteristics of the medical image Accordingly, information that can be observed in the medical image may also vary.

However, if the characteristics of several medical images are comprehensively considered as necessary, more accurate diagnostic auxiliary information can be provided. Accordingly, in the following description of the present specification, a process of segmenting a medical image to include a plurality of characteristics will be described.

37 illustrates an example of a medical image segmented to include a plurality of characteristics according to an embodiment.

Referring to FIG. 37 , the image analysis apparatus 2000 according to an embodiment may segment the medical image so that at least two or more characteristics that vary according to acquisition conditions are included in one medical image. Specifically, the second controller 2002 may segment the first image to include characteristics of the second image based on the first image.

With reference to the drawings, a specific example will be described.

The second controller 2002 may input the first medical image including the first characteristic as input data to the artificial neural network, and obtain output data in which the first characteristic and the second characteristic related to the second medical image are labeled. . Here, the artificial neural network may be trained based on a learning set related to the second image having the second characteristic. For example, here the first characteristic may be an anatomical characteristic or a structural characteristic, and thus the first image may be a T1-weighted image. In addition, the second characteristic may be a lesion characteristic, and accordingly, the second image may be a T2-FLAIR image.

As a more specific example, as shown in (a), the second controller 2002 inputs the first image (in the example, the T2-FLAIR image) including the lesion characteristic as input data to the artificial neural network according to an embodiment. can do.

Here, when using an artificial neural network trained only with a running set related to the first image, the second controller 2002 outputs only the segmentation result including only the first characteristic (ie, lesion characteristic) as shown in (b). do.

However, the artificial neural network according to an embodiment in the present specification may learn using a learning set based on both the first image regarding the first characteristic and the second image regarding the second characteristic. For this reason, the second controller 2002 can input the first image into the artificial neural network and obtain a segmentation result for the first image in which both the first characteristic and the second characteristic are reflected as shown in (c).

As such, even when a medical image including one characteristic is segmented, information on a plurality of characteristics is acquired together, so that information on various diseases can be acquired from one medical image, and image analysis according to an embodiment The apparatus 2000 can provide more accurate and various diagnostic auxiliary information from the medical image.

Hereinafter, a learning process of an artificial neural network according to an embodiment will be described with reference to the drawings.

38 is a flowchart illustrating a learning process of an artificial neural network according to an embodiment.

The artificial neural network according to an embodiment may learn by using learning data. Here, the artificial neural network according to an embodiment may be learned through various devices capable of driving an artificial neural network algorithm. As an example, the artificial neural network may be learned through the image analysis apparatus 2000 . In the present specification, for convenience of explanation, it will be described that the artificial neural network according to an embodiment is learned through the image analysis apparatus 2000, but is not limited thereto, and may be learned through an apparatus for driving other artificial neural networks. It will be understood that there is

Referring to FIG. 39 , in the learning process of the artificial neural network according to an embodiment, the step of segmenting the first image including the first characteristic ( S6000 ), and the second image including the second characteristic Matching the segmentation information (S6200), performing the primary learning using the matched image and the second image (S6400), morphologically modifying the second image including the output result of the primary learning It may include the step (S6600) and the step of performing secondary learning using the second image including the morphologically corrected segmentation result (S6800).

First, the image analysis apparatus 2000 may segment the first image including the first characteristic (S6000). Specifically, the second controller 2002 may segment the first image to include information about the first characteristic acquired under the first acquisition condition. In other words, the first image including the first characteristic obtained when the image acquisition apparatus 1000 is set as the first parameter may be segmented.

After segmenting the first image, the image analysis apparatus 2000 may match the segmented first image and the second image based on the second image ( S6200 ). Specifically, the second controller 2002 matches the first image segmented to include the first characteristic on the second image including the second characteristic, and the second controller 2002 includes the segmented information regarding the first characteristic. image can be obtained. In other words, the first image segmented to include the above-described first characteristic may be matched with the second image including the second characteristic obtained when the image acquisition apparatus 1000 is set as the second parameter. have. This will be described later in detail.

After the first image and the second image are matched, the image analysis apparatus 2000 may perform primary learning using the matched second image and the second image not matched ( S6400 ). Specifically, the second controller 2002 outputs a second image matched to reflect the second image and the first characteristic so that the artificial neural network according to an embodiment outputs a segmentation result including the first characteristic from the second image. It can be used to train artificial neural networks.

That is, the artificial neural network according to an embodiment may learn by using a learning set including the second image labeled with the first characteristic and the original second image. Accordingly, the artificial neural network may receive the second image as input data and output the second image labeled with the first characteristic. Here, although the running set has been described as including the original second image, a processed second image may be used in addition to the original second image. As an example, the second image may be a second image segmented to include the second characteristic. Also, it may be the second image on which other pre-processing processes have been performed. However, in the following description of the present specification, for convenience of explanation, the artificial neural network will be mainly described using a learning set based on the original second image.

After the artificial neural network is primarily learned, the image analysis apparatus 2000 may perform morphological correction on the output image of the primary learned artificial neural network ( S6600 ). Specifically, the second controller 2002 inputs the original second image data to the artificial neural network and performs morphological correction on the segmented second image to include the output second characteristic in order to increase the accuracy of the segmentation result. can do.

The artificial neural network learned primarily may not accurately include information about the first characteristic. Accordingly, the image analysis apparatus 2000 according to an embodiment morphologically modifies the primary output data of the artificial neural network based on the first image including the first characteristic, so that the artificial neural network collects information about the first characteristic. The computational accuracy of the artificial neural network can be improved so that it can be more accurately distinguished on the second image.

When the morphological correction of the first output data of the artificial neural network is completed, the image analysis apparatus 2000 may secondaryly train the artificial neural network based on the morphologically corrected second image ( S6800 ). Specifically, the second controller 2002 may train the artificial neural network according to an embodiment based on a running set including a second image that is morphologically modified and labeled with the first characteristic. That is, the artificial neural network according to an embodiment learns by using a learning set including a second image labeled to reflect the morphologically modified first characteristic, so that information about the first characteristic is more accurately obtained from the second image. Segmentation can be performed to be reflected.

Hereinafter, an example of a learning process of an artificial neural network according to an embodiment will be described with reference to the drawings.

39 is an example illustrating a process of matching a first image and a second image according to an exemplary embodiment.

Referring to FIG. 39 , the image analysis apparatus 2000 according to an embodiment may match a first image and a second image. Specifically, the second controller 2002 may obtain a registration matrix between the first image and the second image. Here, the matching matrix may mean a conversion function between the first image and the second image. That is, in the registration matrix according to an embodiment, the correlation between the first point included in the first image and the second point included in the second image is correlated between the first image and the second image of the same object to be photographed. It can also be expressed as an indicated function. Here, the first point and the second point may mean a pixel or a set of pixels.

The image analysis apparatus 2000 according to an embodiment may process the second image to correspond to the segmentation result of the first image based on the first image segmented with respect to the first characteristic using the matching matrix. That is, the second controller 2002 may segment the second image so that the segmentation result of the first image related to the first characteristic is reflected in the second image. Here, as described above, the second image may be segmented in advance to include the second characteristic. That is, the second controller 2002 may segment the second image so that both the first characteristic and the second characteristic are reflected.

As a more specific example, the first image may be a T1-weighted image, and the second image may be a T2-FLAIR image. That is, the image analysis apparatus 2000 may acquire a registration matrix by first matching the T1-weighted image and the T2-FLAIR image.

Also, the image analysis apparatus 2000 may segment the T1-weighted image to reflect the anatomical or structural characteristics of the T1-weighted image. For example, the second controller 2002 may process the T1-weighted image so that a specific region of the brain is distinguished. More specifically, the second controller 2002 may segment the T1-weighted image so that organs included in the brain, such as the cerebrum, cerebellum, diencephalon, and hippocampus, are distinguished, and differentiated according to the location of each part of the brain such as the temporal lobe, frontal lobe, occipital lobe, etc. It is possible to segment the T1-weighted image as much as possible, and it is also possible to segment them by combining them.

Thereafter, the image analysis apparatus 2000 may use the matching matrix to register the segmented T1-weighted image to the T2-FLAIR image so that the segmentation result of the T1-weighted image is reflected in the T2-FLAIR image.

Accordingly, the image analysis apparatus 2000 may acquire the segmented T2-FLAIR image to reflect the anatomical characteristics. Specifically, the second controller 2002 reflects the above-described anatomical characteristics in the T2-FLAIR image in which the WMH observed in the medical image or the substances constituting the brain, such as white matter and gray matter, are distinguished. The T2-FLAIR image can be segmented as much as possible. The segmented T2-FLAIR image to reflect the anatomical characteristics may be used for primary learning of the artificial neural network as described above.

40 illustrates an example of morphological modification according to an embodiment.

Referring to FIG. 40 , the image analysis apparatus 2000 according to an exemplary embodiment may morphologically correct a segmentation result of a medical image. Specifically, the second controller 2002 may morphologically modify the segmentation result reflecting the first characteristic included in the second image. The second image including the morphologically corrected segmentation result may more accurately reflect the first characteristic.

It will be described as a specific example with reference to the drawings.

As described above, the artificial neural network can perform primary learning using the segmented T2-FLAIR image and the original T2-FLAIR image to reflect anatomical characteristics. As a result of the primary learning, the artificial neural network may receive a T2-FLAIR image and output a segmented T2-FLAIR image to reflect the characteristics (anatomical or structural characteristics) of the T1-weighted image. The T2-FLAIR image, which is the output result of the artificial neural network, is segmented to reflect the characteristics of the T1-weighted image, but it is incomplete in some areas and may require correction.

In the case of (a), as the primary learning result of the artificial neural network according to an embodiment, it is shown that the outer part of the segmented region is segmented by incompletely reflecting anatomical characteristics. In contrast, (b) is the primary output of the morphologically modified artificial neural network, and it is shown that the segmented outer part is modified to reflect the anatomical characteristics well.

More specifically, in the case of (a), some gray matter regions are not distinguished, and the boundary line of each region is partially broken. Also included are noises with regions mislabeled due to incomplete learning.

However, in the morphologically corrected medical image, gray matter regions are well distinguished, and the boundaries of each region clearly distinguish specific regions.

Here, according to an embodiment, the second controller 2002 may perform morphological correction based on the characteristics of the brain structure. In general, regions of the brain, including the ventricles, must be connected in 26 ways in three dimensions. Accordingly, the second controller 2002 may reduce noise included in the segmented medical image based on the connection relationship between each region of the brain.

Also here, the ventricle may be mislabeled as non-tissue because it is misidentified from other non-tissues. In this case, in order to compensate for incorrect labeling of the ventricular region, the second controller 2002 may label the ventricular region using a fill-hole method.

As described above, the artificial neural network according to an embodiment may re-learn as an input the morphologically modified result of the segmentation result, and output the result so that the characteristics to be included are more accurately reflected.

41 is a flowchart illustrating a deployment process using an artificial neural network of an image analysis apparatus according to an embodiment.

Referring to FIG. 41 , the deploying process using an artificial neural network of the image analysis apparatus 2000 according to an embodiment includes the steps of inputting a first image including a first characteristic ( S7000 ) and a first The method may include outputting the processed first image to include the segmentation result including the characteristic and the second characteristic ( S7200 ).

First, the image analysis apparatus 2000 may input a first image including the first characteristic to the artificial neural network (S7000). Specifically, the second controller 2002 may input the first image including the first characteristic to the artificial neural network that has been sufficiently learned and stored in the second memory 2400 . Here, the artificial neural network will be sufficiently trained through the above-described learning process.

When the first image is input, the image analysis apparatus 2000 is processed to include a first characteristic related to the first image and a second characteristic related to a second image of a different format from the first image as an output result of the artificial neural network. 1 image may be acquired (S7200). Specifically, the second controller 2002 may acquire the segmented first image so that the first characteristic related to the first image and the second characteristic related to the second image different from the first image are comprehensively distinguished.

As such, by using the artificial neural network sufficiently learned according to the learning process of the artificial neural network according to the embodiment, the image analysis apparatus 2000 inputs the medical image including only the first characteristic to the artificial neural network, but the first characteristic and the first characteristic A medical image segmented to include both the characteristic and the other second characteristic may be acquired. Accordingly, various information for analyzing a disease for which the related information is to be acquired can be acquired, and more diverse information can be provided to the user even in relation to the disease by analyzing the various information.

42 is an example of a final output result of an artificial neural network according to an embodiment.

Referring to FIG. 42 , the image analysis apparatus 2000 may obtain a first image processed to include various characteristics by inputting the first image to the artificial neural network. Specifically, the second controller 2002 may input the first image to which the first characteristic is reflected to the artificial neural network, and obtain a segmented first image to further reflect the first characteristic and other characteristics.

To describe as a specific example with reference to the drawings, the image analysis apparatus 2000 may input a T2-FLAIR image to an artificial neural network, and obtain an image processed to reflect the characteristics of the T1-weighted image.

Specifically, the second controller 2002 inputs the T2-FLAIR image to an artificial neural network that has been sufficiently trained based on the running set in which the characteristics of the T1-weighted image and the characteristics of the T2-FlAIR image are reflected, and the characteristics of the T1-weighted image A segmented T2-FLAIR image can be acquired to reflect this. Here, of course, the artificial neural network can be learned to segment by reflecting the existing characteristics of the T2-FLAIR image. That is, the artificial neural network according to an embodiment uses the T2-FLAIR image as input data to reflect the anatomical or structural characteristics that are easy to observe in the T1-weighted image, as well as the lesion that is easily observed in the T2-FLAIR image. It is possible to obtain an image segmented to include all of the enemy characteristics.

To give a more specific example, the final output of the artificial neural network according to an embodiment is based on a T2-FLAIR image in which lesion characteristics such as WMH and materials constituting the brain such as white matter and gray matter are well distinguished. It can be segmented to include all characteristics of the T1-weighted image, such as the location of organs and brain regions included in the .

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with reference to the limited embodiments and drawings, various modifications and variations are possible from the above description by those skilled in the art. For example, the described techniques are performed in a different order than the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (18)

A learning method of an artificial neural network for segmentation of a medical image performed by a diagnostic auxiliary device, the method comprising:
acquiring first image data obtained under the first condition and including information related to a first region related to the brain and second image data obtained under the second condition;
determining second region information in second image data to obtain first input data, wherein the second region information is generated to correspond to the second image data based on the first region information;
learning the artificial neural network by inputting the first input data and the second image data to the artificial neural network so that the artificial neural network outputs first output data including information about a first target region;
transforming the first output data into second input data by morphologically modifying at least a portion of the first target region in the first output data; and
The artificial neural network is learned by inputting the second input data and the second image data so that the artificial neural network outputs a second target region reflecting a first characteristic different from the second characteristic observed in the second image data. making; including
Learning method of artificial neural network for segmentation of medical images.
According to claim 1,
The first image data is a T1-weighted MRI image,
Learning method of artificial neural network for segmentation of medical images.
According to claim 1,
The second image data is a T2-FLAIR MRI image,
Learning method of artificial neural network for segmentation of medical images.
According to claim 1,
The first characteristic is an anatomical characteristic,
The first region includes information on at least one region of an organ, location, or region of the brain to which anatomical characteristics are reflected.
Learning method of artificial neural network for segmentation of medical images.
According to claim 1,
The artificial neural network includes a U-net ++ structure
Learning method of artificial neural network for segmentation of medical images.
According to claim 1,
The artificial neural network is
learned to output information about a second region that reflects a second characteristic corresponding to a second condition;
Learning method of artificial neural network for segmentation of medical images.
7. The method of claim 6,
The second characteristic is a lesion characteristic,
The second region information includes information related to a white matter high intensity signal.
Learning method of artificial neural network for segmentation of medical images.
obtaining input image data obtained under a first condition, wherein the input image data reflects a first characteristic corresponding to the first condition; and
Obtaining the resulting image data as an output result of the artificial neural network by inputting the input image data - the artificial neural network is learned using a learning set based on the second image data reflecting the second characteristic corresponding to the second condition step; including,
The resultant image data includes at least a first region indicating lesion information related to the first characteristic and a second region indicating structural information related to the second characteristic.
A method of segmenting a medical image using an artificial neural network.
9. The method of claim 8,
The input image data is a T2-FLAIR MRI image.
A method of segmenting a medical image using an artificial neural network.
9. The method of claim 8,
The second image data is a T1-weighted MRI image,
A method of segmenting a medical image using an artificial neural network.
9. The method of claim 8,
The lesion information includes white matter high intensity signal (WMH)
A method of segmenting a medical image using an artificial neural network.
12. The method of claim 11,
The resulting image data is
Provides anatomical location information where white matter high-intensity signals are generated.
A method of segmenting a medical image using an artificial neural network.
A computer-readable recording medium in which a program for driving the method of any one of claims 1 to 12 is recorded.
In the diagnostic auxiliary information providing device,
a communication module for acquiring input image data;
a controller for analyzing the input image data; including,
The controller inputs the input image data reflecting the first characteristic corresponding to the first condition to the artificial neural network learned by using a learning set based on the second image data reflecting the second characteristic corresponding to the second condition. to obtain the resulting image data,
The resultant image data includes at least a first region indicating lesion information related to the first characteristic and a second region indicating structural information related to the second characteristic.
A device for providing diagnostic auxiliary information.
15. The method of claim 14,
The input image data is a T2-FLAIR MRI image.
A device for providing diagnostic auxiliary information.
15. The method of claim 14,
The second image data is a T1-weighted MRI image,
A device for providing diagnostic auxiliary information.
15. The method of claim 14,
The lesion information includes white matter high intensity signal (WMH)
A device for providing diagnostic auxiliary information.
18. The method of claim 17,
The resulting image data is
Provides anatomical location information where white matter high-intensity signals are generated.
A device for providing diagnostic auxiliary information.

Images