Method for constructing general-purpose modality-agnostic artificial intelligence model by using virtual data and method for segmenting brain image by using constructed artificial intelligence model

Abstract

The present invention relates to a method for segmenting a brain image by using virtual data and, more specifically, to a method for constructing a general-purpose modality-agnostic artificial intelligence model by using virtual data and a method for segmenting a brain image by using the constructed artificial intelligence model. To this end, provided is a method for constructing a general-purpose modality-agnostic artificial intelligence model by using virtual data comprising: a step (S100) for inputting a label map (100) of a medical image including one or more regions related to a disease of a human body; a step (S120) in which an image generation model (120) generates a plurality of deformed images by deforming the label map (100); a step (S140) for training the general-purpose modality-agnostic artificial intelligence model (200) on the basis of at least one of the medical image, the label map (100), or the plurality of deformed images; and a step (S160) for determining that the general-purpose modality-agnostic artificial intelligence model (200) has been constructed when an error between an expected label map (220) output by the general-purpose modality-agnostic artificial intelligence model (200) and the label map (100) is within a prescribed range.

Description

Method for constructing a general modality AI model using virtual data and a brain image segmentation method using the constructed AI model

[0001] The present invention relates to a method for segmenting brain images using virtual data, and more specifically, to a method for constructing a general-purpose modality artificial intelligence model using virtual data and a method for segmenting brain images using the constructed artificial intelligence model.

[0002] In general, various forms of imaging data are used in the field of brain imaging or neuroimaging. For example, Fig. 1a is an amyloid scan image used to diagnose neurodegenerative diseases such as Alzheimer's disease. Additionally, Figs. 1b to 1d are examples of various neuroimaging images used to study the brain. These brain images play an important role in volumetric measurements, morphological analysis, connectivity, physiology, and molecular research.

[0003] Figure 2a is an example of a high-resolution T1-weighted image for medical research purposes, and Figure 2b is an example of a low-resolution MRI scan image that can be obtained in clinical settings. As shown in Figures 2a and 2b, high-resolution T1-weighted images for medical research purposes allow for quantitative measurements of the brain using various analysis software (e.g., FreeSurfer, SPM, FSL, Heuron AD, etc.). However, there are disadvantages, such as the small number of images (only 4,000 to 5,000) and the limitation of subjects to Caucasian subjects.

[0004] On the other hand, as shown in Fig. 2b, low-resolution MRI scan images that can be acquired in clinical settings are characterized by low resolution and varying imaging settings from hospital to hospital, even though more than 10 million brain images are taken annually. Such clinical brain images differ in imaging settings, contrast, resolution, and image size depending on the hospital. There were even cases where the results differed when images were taken at different hospitals, even when using the same modality (e.g., X-ray, CT, MRI, etc.).

[0005] [Table 1] is a table comparing the characteristics of such conventional image data processing and analysis.

[0006] Methodology Speed ​​Image Data Processing and Analysis (modality agnostic) Manual---+++Multi-Atlas Segmentation-+Bayesian Segmentation+++Supervised CNN++---

[0007] Therefore, when attempting to build an artificial intelligence model for these clinical brain images using a supervised learning method, there was a disadvantage in that data collection was costly because each brain image had to be manually segmented. Furthermore, conventional label map segmentation mostly relied on Convolutional Neural Network (CNN) models, which had limitations such as the domain gap problem and the inability to generalize to data with different resolutions or MR contrasts. Generalization was also poor even for data acquired with different parameters or hardware within the same modality. Additionally, although attempts were made to solve the problem using techniques such as data argumentation, there was a disadvantage in that additional costs were incurred because the model had to be retrained for new MRs, requiring data with labels.

[0008] In the case of registration, conventional methods estimated the deformation field between two images, but there were optimization issues and the inconvenience of having to estimate it again for every new image. Furthermore, not only did training take a long time, but deep learning-based registration also had the problem of being affected by learned MR contrast and modality.

[0009] Accordingly, the present invention has been devised to solve the aforementioned problems, and the objective of the present invention is to construct an artificial intelligence model with excellent reasoning capabilities regarding segmentation and analysis by modifying virtual data regarding brain imaging to increase the amount of data and training the artificial intelligence with the increased brain imaging data.

[0010] In other words, the objective of the present invention is to provide a method for constructing a general-purpose modality artificial intelligence model using virtual data and a method for segmenting brain images using the constructed artificial intelligence model.

[0011] However, the technical problems to be solved by the present invention are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below.

[0012] To achieve the above technical objective, a method for constructing a universal modality agnostic model using virtual data is provided, comprising: a step (S100) of inputting a label map (100) of a medical image including one or more regions related to human body diseases; a step (S120) in which an image generation model (120) transforms the label map (100) to generate a plurality of transformed images; a step (S140) of training a universal modality agnostic artificial intelligence model (200) based on at least one of the medical image, the label map (100), and the plurality of transformed images; and a step (S160) of determining that the universal modality agnostic artificial intelligence model (200) has been constructed when the error between the expected label map (220) output by the universal modality agnostic model (200) and the label map (100) is within a predetermined range.

[0013] In addition, the areas related to human diseases include at least one of the following: the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and ventricles of the brain; areas related to Alzheimer's disease; ARIA-E (Amyloid-Related Imaging Abnormalities with Effusion or Edema) areas; ARIA-H (Amyloid-Related Imaging Abnormalities with Hemosiderin) areas; WMH (White Matter Hyperintensities) areas; Lacunes (empty spaces occurring in deep brain structures such as white matter or basal ganglia); cerebral microbleeds; grey matter areas; thalamus areas; caudate areas; putamen areas; skull areas; and cerebrospinal fluid (CSF) areas.

[0014] In addition, the disease of the human body may be one of the following: dementia, Alzheimer's disease, vascular dementia, brain atrophy, Parkinson's disease, stroke, brain tumor, epilepsy, multiple sclerosis, amyloid vaccine side effect disease, or brain cancer.

[0015] Additionally, the image generation model (120) can generate multiple modified images using a Generative Adversarial Network Model (GANM).

[0016] Additionally, the image generation model (120) can generate a plurality of transformed images by applying a spatial transformation method that applies at least one of a rigid transformation method, an affine transformation method, and a non-linear transformation method.

[0017] Additionally, the image generation model (120) can generate multiple modified images using a Gaussian Mixture Model (GMM) or a specific rule-based method in which data is generated by a predetermined defined rule.

[0018] Additionally, the image generation model (120) can generate multiple modified images using an artifact application method or a noise application method.

[0019] Additionally, the image generation model (120) can generate multiple modified images by changing the resolution of the label map (100).

[0020] Additionally, a plurality of modified images include a generated image (140) and a modified label map (160), and the learning step (S140) may include: a step (S142) in which a general modality artificial intelligence model (200) outputs a predicted label map (220) based on the label map (100) and the generated image (140); a step (S144) in which an average die loss unit (170) calculates an average from the similarity between segmented regions and the die loss based on the predicted label map (220) and the modified label map (160); and a step (S146) in which a backpropagation unit (180) updates weights and biases to minimize the error between the predicted label map (220) and the input label map (100).

[0021] Additionally, the general-purpose modality artificial intelligence model (200) may be a Convolutional Neural Network (CNN) model, a Vision Transformer (ViT), or a mixed model thereof.

[0022] Additionally, during the input step (S100), the label map (100) is a plurality of label maps (100), and some of the label maps (100) among the plurality of label maps (100) may be label maps (100) in which the skull region is processed as the background of the region of interest (ROI) or label maps (100) in which the outer cerebrospinal fluid (Outer CSF) region is processed as the background of the region of interest (ROI).

[0023] Additionally, the label map (100) is a label map based on at least one of the original medical images, such as X-ray, MRI, and CT.

[0024] Additionally, the label map (100) is a label map of virtual data generated based on at least one of original medical images such as X-ray, MRI, and CT.

[0025] The objective of the present invention as described above can also be achieved, as another embodiment, by a brain image segmentation method using a universal modality artificial intelligence model with virtual data, characterized by comprising: a step (S200) of inputting a new medical image of the brain into a universal modality artificial intelligence model (200) constructed by the above-described method; and a step (S220) in which the universal modality artificial intelligence model (200) infers and outputs a label map based on the new medical image of the brain.

[0026] Additionally, in the input step (S200), the new medical image is an image that differs from the label map (100) input in the artificial intelligence model construction step in at least one of resolution, size, shooting position, contrast, and format.

[0027] According to one embodiment of the present invention, a highly accurate artificial intelligence model can be constructed by training the artificial intelligence using a large amount of virtual data.

[0028] Furthermore, regarding the input label images of the brain, even if the hospital, shooting settings, resolution, size, format, or contrast differ, inference results with high segmentation robustness for the label map can be obtained. Moreover, the artificial intelligence model constructed according to the present invention has the advantage of exhibiting accurate segmentation performance without error even when the modality changes.

[0029] However, the effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below.

[0030] The following drawings attached to this specification illustrate preferred embodiments of the present invention and serve to further enhance understanding of the technical concept of the present invention together with the detailed description of the invention provided below; therefore, the present invention should not be interpreted as being limited only to the matters described in such drawings.

[0031] FIGS. 1a to 1d are examples of various neuroimaging photographs used to study the brain.

[0032] Fig. 2a is an example of a high-resolution T1-weighted image for medical research purposes,

[0033] FIG. 2b is an example of a low-resolution MRI scan image that can be obtained clinically.

[0034] FIGS. 3a to 3d are examples of virtual synthetic data of a brain used in a method for constructing a general modality artificial intelligence model using virtual data according to the present invention.

[0035] FIG. 4 is a block diagram schematically illustrating a method for constructing a general modality artificial intelligence model using virtual data according to the present invention.

[0036] FIG. 5 is a block diagram illustrating the process of outputting virtual synthetic data of the brain from an arbitrary labeled image in FIG. 4.

[0037] FIG. 6 is a block diagram of another representation of the block diagram shown in FIG. 4.

[0038] Figure 7a is an example of a label map regarding the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and ventricular regions of the brain.

[0039] Fig. 7b is an example of a label map regarding the gray matter, white matter, thalamus, caudate nucleus, putamen, skull, and cerebrospinal fluid regions of the brain.

[0040] FIG. 8a is an example of a label image input without modification in the present invention.

[0041] FIG. 8b is an example of an incompletely labeled image in which the skull region is processed as the background of the region of interest (ROI) in the present invention.

[0042] FIG. 8c is an example of a fully labeled image in which the outer cerebrospinal fluid (Outer CSF) region in the image shown in FIG. 8b is processed as the background of the region of interest (ROI).

[0043] FIG. 9 is a flowchart schematically illustrating a method for constructing a general-purpose modality artificial intelligence model using virtual data according to the present invention and a method for segmenting brain images using the constructed artificial intelligence model.

[0044] Below, with reference to the attached drawings, embodiments of the present invention are described in detail so that those skilled in the art can easily implement the invention. However, since the description of the present invention is merely an example for structural or functional explanation, the scope of the present invention should not be interpreted as being limited by the embodiments described in the text. That is, since the embodiments are subject to various modifications and may take various forms, the scope of the present invention should be understood to include equivalents capable of realizing the technical concept. Furthermore, the objectives or effects presented in the present invention do not imply that a specific embodiment must include all of them or only such effects; therefore, the scope of the present invention should not be understood as being limited by them.

[0045] The meaning of the terms described in this invention should be understood as follows.

[0046] Terms such as "first" and "second" are intended to distinguish one component from another, and the scope of rights shall not be limited by these terms. For example, the first component may be named the second component, and similarly, the second component may be named the first component. When a component is referred to as being "connected" to another component, it should be understood that it may be directly connected to that other component, or that there may be other components in between. Conversely, when a component is referred to as being "directly connected" to another component, it should be understood that there are no other components in between. Meanwhile, other expressions describing the relationship between components, such as "between" and "exactly between," or "adjacent to" and "directly adjacent to," shall be interpreted in the same manner.

[0047] A singular expression should be understood to include a plural expression unless the context clearly indicates otherwise, and terms such as "include" or "have" are intended to specify the existence of the set-up features, numbers, steps, actions, components, parts, or combinations thereof, and should be understood not to preclude the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0048] Unless otherwise defined, all terms used herein have the same meaning as generally understood by those skilled in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having meanings consistent with the context of the relevant technology and should not be interpreted as having an ideal or overly formal meaning unless explicitly defined in this invention.

[0049] Method for constructing a general modality AI model

[0050] Hereinafter, the configuration of a preferred embodiment will be described in detail with reference to the attached drawings. FIGS. 3a to 3d are examples of virtual synthetic data of the brain used in the method for constructing a general-purpose modality artificial intelligence model using virtual data according to the present invention, FIG. 4 is a block diagram schematically showing a method for constructing a general-purpose modality artificial intelligence model using virtual data according to the present invention, and FIG. 9 is a flowchart schematically showing a method for constructing a general-purpose modality artificial intelligence model using virtual data according to the present invention and a method for segmenting brain images using the constructed artificial intelligence model.

[0051] As illustrated in FIGS. 4 and 9, first, a label map (100) containing one or more regions related to diseases of the human body is input (S100). An example of the input label map (100) is as shown in FIGS. 3a to 3d. The label map (100) is a label data structure primarily used in tasks such as image segmentation or object detection, and is a map that represents which class (or category) each pixel or object belongs to. That is, it is a map containing class information for each part of the brain image or 3D data. The label map (100) is a map that assigns class labels to each pixel, voxel, or object of the brain image or 3D data, and is primarily used in tasks such as image segmentation, object detection, and medical image analysis. Through this, it clearly defines which class each part belongs to and plays an important role in learning and performance evaluation. In particular, when pixels of a brain MRI image need to be divided into classes such as gray matter, white matter, and cerebrospinal fluid in a medical image, the label map (100) is an array that displays the label of each class (tissue) with numbers such as 0, 1, 2 for each pixel of the image.

[0052] Additionally, the label map (100) in the input step (S100) is a plurality of label maps (100). For example, FIG. 7a is an example of a label map regarding the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and ventricular regions of the brain, and FIG. 7b is an example of a label map regarding the gray matter, white matter, thalamus, caudate nucleus, putamen, skull, and cerebrospinal fluid regions of the brain.

[0053] And, FIG. 8a is an example of a labeled image input without modification in the present invention, FIG. 8b is an example of an incompletely labeled image in which the skull region is processed as the background of the region of interest (ROI) in the present invention, and FIG. 8c is an example of a fully labeled image in which the outer cerebrospinal fluid (Outer CSF) region is processed as the background of the region of interest (ROI) in the image shown in FIG. 8b.

[0054] As shown in FIGS. 8a to 8c, which are input in this manner, some of the label maps (100) among the plurality of label maps (100) are label maps (100) in which the skull region is processed as the background of the region of interest (ROI) or label maps (100) in which the outer cerebrospinal fluid (Outer CSF) region is processed as the background of the region of interest (ROI).

[0055] Additionally, the label map (100) is a label map of virtual data generated based on at least one of original medical images such as X-ray, MRI, and CT, and preferably based on a 3D T1 image among MRI images. Specifically, the label map (100) refers to a 2D, 3D, or 4D image and is a label map generated from MRI, X-ray, CT, T1-weighted images of MRI, T2-weighted images of MRI, etc.

[0056] In addition, the areas related to diseases of the human body include at least one of the following: the frontal lobe, parietal lobe, temporal lobe, occipital lobe, and ventricles of the brain; the area related to Alzheimer's disease; the ARIA-E (Amyloid-Related Imaging Abnormalities with Effusion or Edema) area; the ARIA-H (Amyloid-Related Imaging Abnormalities with Hemosiderin) area; the WMH (White Matter Hyperintensities) area; the Lacunes area, which is a void occurring in deep brain structures such as white matter or basal ganglia; the cerebral microbleeds area; the grey matter area; the thalamus area; the caudate area; the putamen area; the skull area; and the cerebrospinal fluid (CSF) area.

[0057] And, the disease of the human body may be one of the following: dementia, Alzheimer's disease, vascular dementia, brain atrophy, Parkinson's disease, stroke, brain tumor, epilepsy, multiple sclerosis, amyloid vaccine side effects, or brain cancer.

[0058] Next, the image generation model (120) transforms the label map (100) to generate a plurality of transformed images (S120). There are various ways to transform the label map (100) as follows.

[0059] For example, an image generation model (120) can generate multiple modified images using a Generative Adversarial Network Model (GANM). A Generative Adversarial Network Model (GANM) is a generative model designed with a structure in which two neural networks compete with each other to learn.

[0060] Additionally, the image generation model (120) can generate multiple transformed images by applying a spatial transformation method that applies at least one of a rigid transformation method, an affine transformation method, and a non-linear transformation method. The rigid transformation method is a method that simply transforms the position and direction without translation, rotation, or scale transformation while maintaining the shape of the object or image, and is used when rotating an image or moving it to a specific location.

[0061] Affine transformation is a method that can encompass translation, rotation, scaling, and distortion of images or objects. Affine transformations maintain linear properties but can alter angles or length ratios. In other words, parallel lines are preserved, but they can be deformed to be non-right angles. For example, an image of the brain can be enlarged or reduced by a fixed ratio and a gradient applied.

[0062] Non-linear transformation is a method of transforming coordinates in a non-linear manner rather than through linear transformation. It allows for the application of curvilinear distortions or complex non-linear deformations to brain images. For example, it is possible to perform transformations that distort parts of a brain image or stretch only specific areas.

[0063] Additionally, the image generation model (120) can generate multiple modified images using a Gaussian Mixture Model (GMM) or a specific rule-based method in which data is generated by a predetermined defined rule.

[0064] A Gaussian Mixture Model (GMM) is a method of modeling data by combining multiple Gaussian distributions. As a sampling technique, it is used to describe or generate data by representing complex distributions as a mixture of several simple Gaussian distributions.

[0065] Furthermore, specific rule-based sampling is a method of selecting or generating data according to specific rules or criteria. This specific rule-based approach is contrasted with random sampling and is used to select data that satisfies specific conditions. Rule-based sampling is generally a method in which data is selected or generated by defined rules or algorithms.

[0066] Additionally, the image generation model (120) can generate multiple deformed images using an artifact application method or a noise application method. Artifacts refer to abnormal or unintended errors, distortions, or anomalies that occur primarily in medical imaging, data processing, image processing, or signal processing. Artifacts can interfere with analysis and may occur mainly due to equipment, environmental factors, or the data processing process. An example of such artifacts is a phenomenon in which images are distorted or errors occur due to various factors, such as equipment limitations or patient movement, in medical imaging such as MRI, CT, and X-ray.

[0067] Additionally, the image generation model (120) can generate multiple modified images by changing the resolution of the label map (100).

[0068] The generated multiple modified images include a generated image (140) and a modified label map (160) as illustrated in FIG. 4. The generated image (140) is input into a general-purpose modality artificial intelligence model (200), and the modified label map (160) is input into an average die loss unit (170), respectively.

[0069] Next, a general modality agnostic artificial intelligence model (200) is trained using the generated image (S140). The specific detailed process is as follows. First, the general modality agnostic artificial intelligence model (200) outputs an expected label map (220) based on the label map (100) and the generated image (140) (S142). This general modality agnostic artificial intelligence model (200) can be a Convolutional Neural Network (CNN) model, a Vision Transformer (ViT), or a hybrid model thereof.

[0070] Next, the average dice loss (170) is calculated based on the similarity between the segmented regions and the dice loss, according to the expected label map (220) and the modified label map (160) (S144).

[0071] Next, the backpropagation unit (180) updates the weights and biases to minimize the error between the expected label map (220) and the input label map (100) (S146).

[0072] During the learning process, if the error between the expected label map (220) output by the general modality artificial intelligence model (200) and the input label map (100) is within a predetermined range, it is determined (S160) that the general modality artificial intelligence model (200) has been constructed, and the learning is terminated.

[0073] FIG. 5 is a block diagram illustrating the process of outputting virtual synthetic data of the brain from an arbitrary labeled image in FIG. 4, and FIG. 6 is a block diagram of another representation of the block diagram shown in FIG. 4. As shown in FIG. 5 and FIG. 6, virtual synthetic data is generated using arbitrary labels, and the generation module can generate synthetic brain image data using one or more of rule-based, machine learning, or artificial intelligence-based methods.

[0074] Brain image segmentation method using a constructed artificial intelligence model

[0075] Hereinafter, the operation of a preferred embodiment will be described in detail with reference to FIG. 9. First, a new medical image of the brain is input into a general-purpose modality artificial intelligence model (200) constructed by the method described above (S200). At this time, the new medical image in the input step (S200) is an image that differs from the label map (100) input in the artificial intelligence model construction step in at least one of resolution, size, shooting position, contrast, and format.

[0076] Next, the general modality artificial intelligence model (200) infers and outputs a label map based on a new medical image of the brain (S220). Through this process, a label map with robust segmentation can be obtained from the input label image.

[0077] As described above, the detailed description of the preferred embodiments of the present invention disclosed is provided to enable those skilled in the art to implement and practice the present invention. Although the present invention has been described with reference to preferred embodiments, those skilled in the art will understand that various modifications and changes can be made to the present invention without departing from the scope of the invention. For example, those skilled in the art may utilize each configuration described in the embodiments described above in combination with one another. Accordingly, the present invention is not intended to be limited to the embodiments shown herein, but to be given the broadest scope consistent with the principles and novel features disclosed herein.

[0078] The present invention may be embodied in other specific forms without departing from the spirit and essential features of the invention. Accordingly, the above detailed description should not be interpreted restrictively in all respects but should be considered exemplary. The scope of the invention shall be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the invention are included within the scope of the invention. The invention is not intended to be limited to the embodiments shown herein, but to be given the broadest possible scope consistent with the principles and novel features disclosed herein. Furthermore, embodiments may be constructed by combining claims that are not explicitly related in the claims, or by including them as new claims through amendments made after filing.

[0079] Explanation of the symbols

[0080] 100 : Label Map,

[0081] 120 : Image generation model,

[0082] 140 : Generated image,

[0083] 160 : Modified image,

[0084] 170 : Average Dice Loss,

[0085] 180 : Back Propagation

[0086] 200: General Modality Artificial Intelligence Model,

[0087] 220 : Expected label map.

[0088] 230 : Synthetic brain imaging data output unit.

Claims

1. A step (S100) of inputting a label map (100) of a medical image that includes one or more regions related to a disease of the human body; Step (S120) in which an image generation model (120) transforms the label map (100) to generate a plurality of transformed images; A step (S140) of training a general modality agnostic artificial intelligence model (200) based on at least one of the medical image, the label map (100), and the plurality of modified images; A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized by including the step (S160) of determining that the general-purpose modality artificial intelligence model (200) has been constructed when the error between the expected label map (220) output by the general-purpose modality artificial intelligence model (200) and the label map (100) is within a predetermined range.

2. In Paragraph 1, A method for constructing a general modality artificial intelligence model using virtual data, characterized in that the above-mentioned areas related to human diseases include at least one of the following: the frontal lobe, parietal lobe, temporal lobe, occipital lobe, ventricles of the brain, areas related to Alzheimer's disease, ARIA-E (Amyloid-Related Imaging Abnormalities with Effusion or Edema) area, ARIA-H (Amyloid-Related Imaging Abnormalities with Hemosiderin) area, WMH (White Matter Hyperintensities) area, Lacunes area occurring in deep brain structures such as white matter or basal ganglia, cerebral microbleeds area, grey matter area, thalamus area, caudate area, putamen area, skull area, and cerebrospinal fluid (CSF) area.

3. In Paragraph 1, A method for constructing a general modality artificial intelligence model using virtual data, characterized in that the above-mentioned human disease is one of dementia, Alzheimer's disease, vascular dementia, brain atrophy, Parkinson's disease, stroke, brain tumor, epilepsy, multiple sclerosis, amyloid vaccine side effect disease, or brain cancer.

4. In Paragraph 1, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the above image generation model (120) generates the plurality of modified images using a Generative Adversarial Network Model (GANM).

5. In Paragraph 1, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the above image generation model (120) generates the plurality of transformed images using a spatial transformation method that applies at least one of a rigid transformation method, an affine transformation method, and a non-linear transformation method.

6. In Paragraph 1, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the above image generation model (120) generates the plurality of modified images using a Gaussian Mixture Model (GMM) or a specific rule-based method in which data is generated by a predetermined defined rule.

7. In Paragraph 1, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the above image generation model (120) generates the plurality of modified images using an artifact application method or a noise application method.

8. In Paragraph 1, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the image generation model (120) changes the resolution of the label map (100) to generate the plurality of modified images.

9. In Paragraph 1, The above plurality of modified images includes a generated image (140) and a modified label map (160), and The above learning step (S140) is, Step (S142) in which the above general modality artificial intelligence model (200) outputs the expected label map (220) based on the label map (100) and the generated image (140); A step (S144) in which the average die loss section (170) calculates an average from the die loss and the similarity between the divided regions based on the predicted label map (220) and the modified label map (160); and A method for constructing a general modality artificial intelligence model using virtual data, characterized by including a step (S146) in which a backpropagation unit (180) updates weights and biases to minimize the error between the predicted label map (220) and the input label map (100).

10. In Paragraph 1, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the above general-purpose modality artificial intelligence model (200) is a Convolutional Neural Network (CNN) model, a Vision Transformer (ViT) model, or a mixed model thereof.

11. In Paragraph 1, In the above input step (S100), the label map (100) is a plurality of label maps (100), and A method for constructing a general modality artificial intelligence model using virtual data, characterized in that some of the above-mentioned multiple label maps (100) are label maps (100) in which the skull region is processed as the background of the region of interest (ROI) or label maps (100) in which the outer cerebrospinal fluid (Outer CSF) region is processed as the background of the region of interest (ROI).

12. In Paragraph 1, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the above label map (100) is a label map based on X-ray, MRI, and CT images.

13. In Paragraph 12, A method for constructing a general-purpose modality artificial intelligence model using virtual data, characterized in that the above label map (100) is a label map of virtual data generated based on X-ray, MRI, and CT images.

14. A step (S200) of inputting a new medical image of the brain into a general modality artificial intelligence model (200) constructed according to any one of claims 1 to 13; and A method for segmenting a brain image using a general-purpose modality artificial intelligence model with virtual data, characterized by including the step (S220) in which the general-purpose modality artificial intelligence model (200) infers and outputs a label map based on a new medical image of the brain.

15. In Paragraph 14, A brain image segmentation method using a universal modality artificial intelligence model using virtual data, characterized in that in the input step (S200), the new medical image is an image that differs from the label map (100) input in the artificial intelligence model construction step in at least one of resolution, size, shooting position, contrast, and format.

Images