Method and apparatus for analyzing brain MRI image on basis of deep learning model

Abstract

Provided are a method, apparatus, and computer program for analyzing a brain MRI image on the basis of a deep learning model. The method according to an embodiment disclosed herein comprises the steps of: training a neural network model on the basis of training data that includes a plurality of brain MRI images; when the training is complete, inputting brain MRI images obtained from a subject into the trained neural network model, and thereby identifying a plurality of regions included in the brain of the subject, and determining the volumes of the plurality of regions; and providing numerical information about the volume of each of the regions included in the brain of the subject, wherein the training step comprises the steps of: applying a CDF mapping function, set on the basis of scan conditions, to each of the plurality of brain MR images included in the training data, and thereby obtaining augmented MR images corresponding to each of the brain MR images; and training the neural network model on the basis of training data that includes the augmented MR images.

Description

Method and apparatus for analyzing brain MR images based on a deep learning model

[0001] The present disclosure relates to deep learning technology in the medical field, and more specifically, to a method and apparatus for analyzing brain MR images using a deep learning model.

[0002] X-ray machines, ultrasound diagnostic devices, computed tomography (CT) scanners, and magnetic resonance imaging (MRI) machines are utilized to acquire internal body information for observing and diagnosing the inside of a patient's body. Among these, MRI is receiving attention for its utility compared to other imaging techniques, particularly because it allows imaging without exposing the patient to radiation or administering contrast agents, and offers high resolution and excellent soft tissue contrast. Recently, with the advancement of artificial intelligence technology, deep learning models are being applied in the field of MR image analysis, further increasing the utility of MR images. For example, this could involve using a pre-trained deep learning model to segment a patient's brain within an MR brain image into detailed regions and quantitatively analyzing each region to determine whether brain atrophy has occurred.

[0003] Meanwhile, in the case of MR images, brightness and contrast vary depending on the scanning conditions, specifically the scan parameters set on the MR device. This has caused a problem where the results of neural network models differ depending on the scanning conditions, even for MR images acquired from the same object. To address this, conventional methods focused on reducing the differences between MR images generated under different scanning conditions and ensuring the consistency of training data. Normalization methods were used to reduce differences between MR images by converting the brightness and contrast of each image to specific standards. However, while this method contributed to mitigating the differences in MR images based on scanning conditions, it had the problem of excluding or omitting non-average MR images from the training data. Consequently, the unique characteristics of the training data were lost, and data diversity was reduced, leading to deep learning models overfitting only to specific scanning conditions. This lowered the reliability and accuracy of analysis results when processing MR images acquired under new scanning conditions.

[0004] Consequently, a new approach is required to improve generalization performance so that deep learning models can maintain high accuracy and reliability despite various scan conditions.

[0005] The present disclosure is devised in response to the aforementioned background technology and aims to provide a method and apparatus for analyzing brain MR images based on a deep learning model.

[0006] However, the problems to be solved in this disclosure are not limited to those mentioned above, and other unmentioned problems may be clearly understood based on the description below.

[0007] A method for analyzing brain MR images based on a deep learning model, performed by a computing device including at least one processor according to one embodiment of the present disclosure for realizing the aforementioned tasks, comprises the steps of: training a neural network model based on training data including a plurality of brain MR images; when the training is completed, inputting a brain MR image obtained from a subject into the trained neural network model to identify a plurality of regions included in the subject's brain and to identify the volume of the plurality of regions; and providing numerical information regarding the volume of each region included in the subject's brain. The training step comprises the steps of: applying a CDF mapping function set based on scan conditions to each of the plurality of brain MR images included in the training data to obtain an augmented MR image corresponding to each brain MR image; and training the neural network model based on training data including the augmented MR image.

[0008] Alternatively, the CDF mapping function includes a plurality of CDF mapping functions corresponding to different scan conditions, and the step of acquiring the augmented MR image includes identifying the scan conditions applied to each of the brain MR images, selecting a CDF mapping function corresponding to a scan condition different from the scan conditions applied to each of the brain MR images for each of the brain MR images, and applying the CDF mapping function selected for each of the brain MR images to each of the brain MR images.

[0009] Alternatively, the step of acquiring the augmented MR image includes acquiring the augmented MR image corresponding to each brain MR image, which is acquired under scan conditions different from those applied to each brain MR image.

[0010] Alternatively, the augmented MR image and the brain MR image corresponding to the augmented MR image are characterized by having different contrasts and brightness for the same brain.

[0011] Alternatively, scan conditions applied to brain MR images obtained from the subject can be identified, and if the identified scan conditions correspond to new scan conditions that do not correspond to the plurality of CDF mapping functions, the CDF mapping function can be newly generated based on the new scan conditions.

[0012] Alternatively, the scan condition may be determined according to a set value of a scan parameter including at least one of a pulse sequence, magnetic field strength, slice thickness, and flip angle.

[0013] Alternatively, the method may include the step of identifying a first histogram regarding the pixel value distribution of a first reference MR image corresponding to a first scan condition and a second histogram regarding the pixel value distribution of a second reference MR image corresponding to a second scan condition, and generating a CDF mapping function between the first scan condition and the second scan condition.

[0014] Alternatively, the step of generating the CDF mapping function may include identifying a first CDF function corresponding to the first reference MR image based on the first histogram, identifying a second CDF function corresponding to the second reference MR image based on the second histogram, and obtaining a first CDF mapping function that converts from the first scan condition to the second scan condition and a second CDF mapping function that converts from the second scan condition to the first scan condition based on the first CDF function and the second CDF function.

[0015] Alternatively, the step of providing the numerical information may include providing numerical information including the volume value of each of the regions, the ratio of the total volume of the brain to the volume of the skull region of the subject, and the ratio of the volume of each of the regions to the volume of the skull region of the subject.

[0016] A computing device for analyzing brain MR images based on a deep learning model according to one embodiment of the present disclosure for realizing the above-described problem comprises a memory for storing at least one instruction and at least one processor that, by executing the at least one instruction, trains a neural network model based on training data including a plurality of brain MR images, and when the training is completed, inputs a brain MR image obtained from a subject to the trained neural network model to identify a plurality of regions included in the subject's brain, identifies the volume of the plurality of regions, and provides numerical information regarding the volume of each region included in the subject's brain. The at least one processor can each apply a CDF mapping function set based on scan conditions to the plurality of brain MR images included in the training data to obtain an augmented MR image corresponding to each brain MR image, and train the neural network model based on the training data including the augmented MR image.

[0017] A computer program stored on a computer-readable storage medium according to one embodiment of the present disclosure for realizing the above-described problem, wherein when the computer program is executed on one or more processors, it performs operations for analyzing brain MR images based on a deep learning model, wherein the operations include an operation of training a neural network model based on training data including a plurality of brain MR images, an operation of inputting a brain MR image obtained from a subject into the trained neural network model when the training is completed to identify a plurality of regions included in the subject's brain and to identify the volume of the plurality of regions, and an operation of providing numerical information regarding the volume of each region included in the subject's brain, wherein the training operation includes an operation of obtaining an augmented MR image corresponding to each brain MR image by applying a CDF mapping function set based on scan conditions to each of the plurality of brain MR images included in the training data, and an operation of training the neural network model based on training data including the augmented MR image.

[0018] According to a method for analyzing brain MR images based on a deep learning model according to an embodiment of the present disclosure, MR images acquired without separate additional shooting are augmented according to the contrast of the MRI scan to generate various augmented MR images, and training data can be constructed without data omission. By training a deep learning model based on such training data, non-uniformity in image quality caused by differences in scan conditions can be effectively corrected, and diversity of training data can be secured to obtain a deep learning model that provides high reliability in the quantitative analysis process. Furthermore, by efficiently training data acquired under various scan conditions based on this, the accuracy of the analysis can be improved, and the progression of lesions such as neurodegenerative diseases can be evaluated more precisely.

[0019] FIG. 1 is an exemplary diagram of a computing device for acquiring magnetic resonance images based on a deep learning model according to one embodiment of the present disclosure.

[0020] FIG. 2 is a block diagram of a computing device for analyzing brain MR images based on a deep learning model according to one embodiment of the present disclosure.

[0021] FIG. 3 is a flowchart of a method for controlling a computing device that analyzes brain MR images based on a deep learning model according to one embodiment of the present disclosure.

[0022] FIG. 4 is an exemplary diagram of a method for controlling a computing device that acquires an augmented MR image for a brain MR image using a CDF mapping function according to one embodiment of the present disclosure.

[0023] FIGS. 5a to 5c are exemplary diagrams of a report generated based on numerical information regarding the volume of each region contained in the brain according to one embodiment of the present disclosure.

[0024] FIG. 6 is an exemplary diagram illustrating a method for generating a CDF mapping function according to one embodiment of the present disclosure.

[0025] FIG. 7 is a detailed block diagram of a computing device for analyzing brain MR images based on a deep learning model according to another embodiment of the present disclosure.

[0026] Embodiments of the present disclosure are described below with reference to the attached drawings so that those skilled in the art (hereinafter, those skilled in the art) can easily implement them. The embodiments presented in the present disclosure are provided to enable those skilled in the art to use or implement the contents of the present disclosure. Accordingly, various modifications to the embodiments of the present disclosure will be apparent to those skilled in the art. That is, the present disclosure may be embodied in various different forms and is not limited to the embodiments below.

[0027] Throughout the specification of the present disclosure, identical or similar reference numerals refer to identical or similar components. Additionally, to clearly explain the present disclosure, reference numerals in the drawings that are unrelated to the description of the present disclosure may be omitted.

[0028] The term “or” as used in this disclosure is intended to mean an implicit “or” rather than an exclusive “or.” That is, unless otherwise specified in this disclosure or its meaning is not clear from the context, “X uses A or B” should be understood to mean one of the natural implicit substitutions. For example, unless otherwise specified in this disclosure or its meaning is not clear from the context, “X uses A or B” may be interpreted as X using A, X using B, or X using both A and B.

[0029] The term “and / or” as used in this disclosure should be understood to refer to and include all possible combinations of one or more of the enumerated related concepts.

[0030] The terms “comprising” and / or “comprising” as used in this disclosure should be understood to mean the presence of certain features and / or components. However, the terms “comprising” and / or “comprising” should be understood not to exclude the presence or addition of one or more other features, other components and / or combinations thereof.

[0031] Where not otherwise specified in the present disclosure or where it is not clear from the context that the singular form indicates, the singular should generally be interpreted as including “one or more.”

[0032] The term "the N (N is a natural number)" used in this disclosure may be understood as an expression used to distinguish the components of this disclosure from one another according to certain criteria, such as functional perspectives, structural perspectives, or convenience of explanation. For example, components performing different functional roles in this disclosure may be distinguished as a first component or a second component. However, components that are substantially identical within the technical scope of this disclosure but need to be distinguished for the convenience of explanation may also be distinguished as a first component or a second component.

[0033] The term “acquisition” as used in this disclosure may be understood to mean not only receiving data through a wired or wireless communication network with an external device or system, but also generating data in an on-device form.

[0034] Meanwhile, the terms "module" or "unit" used in this disclosure may be understood as referring to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, or a combination of software and hardware. In this case, "module" or "unit" may be a unit composed of a single element, or a unit expressed as a combination or set of multiple elements. For example, in a narrow sense, "module" or "unit" may refer to a hardware element of a computing device or a set thereof, an application program that performs a specific function of software, a procedure implemented through software execution, or a set of instructions for program execution. Furthermore, in a broad sense, "module" or "unit" may refer to the computing device itself that constitutes the system, or an application executed on the computing device. However, since the above-described concept is merely an example, the concepts of "module" or "part" may be defined in various ways within the scope understandable to those skilled in the art based on the contents of this disclosure.

[0035] As used in this disclosure, the term "model" may be understood as a system implemented using mathematical concepts and language to solve a specific problem, a set of software units to solve a specific problem, or an abstract model regarding a processing process to solve a specific problem. For example, a neural network "model" may refer to an overall system implemented as a neural network that possesses problem-solving capabilities through learning. In this case, the neural network may possess problem-solving capabilities by optimizing parameters connecting nodes or neurons through learning. A neural network "model" may include a single neural network or a set of neural networks composed of multiple neural networks.

[0036] The term “data” as used in this disclosure may include images, signals, etc. The term “image” as used in this disclosure may refer to multidimensional data composed of discrete image elements. In other words, “image” may be understood as a term referring to a digital representation of an object visible to the human eye. For example, “image” may refer to multidimensional data composed of elements corresponding to pixels in a two-dimensional image. “Image” may refer to multidimensional data composed of elements corresponding to voxels in a three-dimensional image.

[0037] As used in this disclosure, the term “image” may refer to multi-dimensional data composed of discrete image elements (e.g., pixels in a two-dimensional image and voxels in a three-dimensional image). For example, an image may include, but is not limited to, medical images acquired by a medical imaging device such as a magnetic resonance imaging device, a computed tomography (CT) device, an ultrasound device, or an X-ray device.

[0038] As used in this disclosure, the term "medical image" is a general concept encompassing all forms of images that include medical knowledge, and may include images acquired through various modalities such as visible light cameras, IR cameras, ultrasound, X-ray, CT, MRI, PET, etc.

[0039] As used in this disclosure, the term "picture archiving and communication system (PACS)" may refer to a system that stores, processes, and transmits medical images in accordance with the Digital Imaging and Communications in Medicine (DICOM) standard. For example, the "picture archiving and communication system" may be linked with digital medical imaging equipment to store medical images, such as magnetic resonance imaging (MRI) and computed tomography (CT) images, in accordance with the Digital Imaging and Communications in Medicine standard. The "picture archiving and communication system" may transmit medical images to terminals inside or outside the hospital via a communication network. At this time, meta information, such as interpretation results and medical records, may be added to the medical images.

[0040] As used in this disclosure, the term "object" refers to a subject of imaging and may include a person, an animal, or a part thereof. For example, an object may include a part of the body (such as an organ or tissue) or a phantom. A phantom refers to a substance having a volume that is very close to the density and effective atomic number of a living organism, and may include a spherical phantom having properties similar to those of the body.

[0041] A Magnetic Resonance Imaging (MRI) system is a system that acquires images of cross-sectional areas of an object by expressing the intensity of a Magnetic Resonance (MR) signal in response to a Radio Frequency (RF) signal generated in a magnetic field of a specific intensity as contrast.

[0042] The MRI system causes the main magnet to form a static magnetic field and aligns the direction of the magnetic dipole moment of specific atomic nuclei of an object located within the static field with the direction of the static field. A gradient field coil applies a gradient signal to the static field to form a gradient field, thereby inducing different resonance frequencies for different parts of the object. An RF coil can irradiate magnetic resonance signals in accordance with the resonance frequency of the area where image acquisition is desired. Additionally, as the gradient field is formed, the RF coil can receive magnetic resonance signals of different resonance frequencies radiated from various parts of the object. The MRI system acquires images by applying image reconstruction techniques to the magnetic resonance signals received through these steps. Furthermore, the MRI system may reconstruct multiple magnetic resonance signals into image data by performing serial or parallel signal processing on multiple magnetic resonance signals received by a multi-channel RF coil.

[0043] The explanation of the foregoing terms is intended to aid in understanding the present disclosure. Accordingly, it should be noted that unless a foregoing term is explicitly stated as a matter limiting the content of the present disclosure, it is not to be used in the sense of limiting the technical concept of the content of the present disclosure.

[0044] FIG. 1 is an exemplary diagram of a computing device (100) for acquiring a magnetic resonance image based on a deep learning model according to one embodiment of the present disclosure.

[0045] A computing device (100) for analyzing brain MR images based on a deep learning model according to one embodiment of the present disclosure may be a hardware device or part of a hardware device that performs comprehensive processing and computation of data, or it may be a software-based computing environment connected to a communication network. For example, the computing device (100) may be a server that performs intensive data processing functions and is an entity that shares resources, or it may be a client that shares resources through interaction with the server. Additionally, the computing device (100) may be a cloud system that enables multiple servers and clients to interact to comprehensively process data. Since the above description is merely one example regarding the type of computing device (100), the type of computing device (100) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure. For example, the computing device (100) may include a smartphone, tablet PC, PC, smart TV, server (e.g., micro server, cloud server), etc., that processes magnetic resonance images or performs processing functions. As another example, the computing device (100) may be a magnetic resonance imaging (MRI) device that directly acquires magnetic resonance images.

[0046] The computing device (100) can acquire a brain MR image of a subject. The brain MR image may be acquired from a medical image storage and transmission system (PACS) that is network-connected to the computing device (100) described above, or it may be acquired by the computing device (100) directly performing an MR scan of the subject. At this time, the computing device (100) can analyze the brain MR image by inputting the acquired brain MR image into a pre-trained neural network model (10). In particular, the computing device (100) can use the pre-trained neural network model (10) to divide the subject's brain included in the brain MR image into multiple regions and identify the volume of each divided region. The computing device (100) can determine a quantitative value for the volume of each divided region. Furthermore, the computing device (100) can analyze the state of the subject's brain based on the quantitative value and diagnose brain-related diseases. Multiple regions of the brain may be regions distinguished according to specific anatomical structures of the brain, or regions classified organizationally or functionally.

[0047] Meanwhile, the computing device (100) can train a neural network model (10) used for the analysis of brain MR images using training data (50). The computing device (100) can train the neural network model (10) to divide the brain included in the brain MR images (30) into multiple regions using training data (50) that includes multiple brain MR images (30). To this end, the computing device (100) can compose the training data (50) using multiple brain MR images (30) obtained by photographing the brains of different subjects. Meanwhile, each brain MR image (30) has different contrast and brightness depending on the MRI device that captured the subject's brain or the scan conditions set on the MRI device. Obtaining brain MR images (30) with various contrasts and brightness is important to increase the diversity of data during the learning process of the neural network model (10) to prevent overfitting and to improve the generalization performance of the neural network model (10) for MR images obtained under various conditions. This is because it can increase the accuracy and reliability of the neural network model (10) and ensure consistency of quantitative analysis results.

[0048] However, obtaining multiple brain MR images (30) acquired under various scanning conditions is a difficult task to implement because it requires repeatedly capturing images of the same subject under different scanning conditions. However, a computing device according to one embodiment of the present disclosure can effectively generate an MR image identical to that acquired under various scanning conditions for the same subject by utilizing a CDF mapping function to augment each brain MR image (30). Hereinafter, an embodiment of the present disclosure related thereto will be described in detail with reference to FIGS. 2 to 5.

[0049] FIG. 2 is a block diagram of a computing device (100) for analyzing a brain MR image (30) based on a deep learning model according to one embodiment of the present disclosure. FIG. 3 is a flowchart of a method for controlling a computing device (100) for analyzing a brain MR image (30) based on a deep learning model according to one embodiment of the present disclosure.

[0050] Referring to FIG. 2, the computing device (100) includes at least one processor (110, hereinafter referred to as processor) and memory (120). However, since FIG. 2 is merely an example, the computing device (100) may include other configurations for implementing a computing environment. Additionally, only some of the disclosed configurations may be included in the computing device (100).

[0051] A processor (110) according to one embodiment of the present disclosure may be understood as a constituent unit comprising hardware and / or software for performing computing operations. For example, the processor (110) may read a computer program and perform data processing for machine learning. The processor (110) may process computational processes such as processing input data for machine learning, extracting features for machine learning, and calculating errors based on backpropagation. A processor (110) for performing such data processing may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA). Since the above-described types of processors (110) are merely examples, the types of processors (110) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0052] The processor (110) is electrically connected to other components of the computing device (100) (i.e., memory (120)) to control the overall operation of the computing device (100).

[0053] A memory (120) according to one embodiment of the present disclosure may be understood as a configuration unit comprising hardware and / or software for storing and managing data processed by a computing device (100). That is, the memory (120) may store data of any form generated or determined by a processor (110) and data of any form received by a communication interface. For example, the memory (120) may include at least one type of storage medium among a flash memory type (120), a hard disk type, a multimedia card micro type, a card type memory, RAM (random access memory), SRAM (static random access memory), ROM (read-only memory), EEPROM (electrically erasable programmable read-only memory), PROM (programmable read-only memory), magnetic memory, a magnetic disk, and an optical disk. Additionally, the memory (120) may include a database system that controls and manages data in a predetermined system. Since the above-described type of memory (120) is merely an example, the type of memory (120) can be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0054] The memory (120) can structure and organize data, combinations of data, and program code executable on the processor (110) that are necessary for the processor (110) to perform calculations. For example, the memory (120) may store a neural network model (10) and training data (50). Additionally, the memory (120) may store program code that operates the neural network model (10) to perform learning based on the training data (50), program code that operates the neural network model (10) to receive a brain MR image (30) as input and perform inference according to the purpose of use of the computing device (100), and processed data generated as the program code is executed.

[0055] Referring to FIG. 3, the processor (110) can train a neural network model (10) based on training data (50) that includes a plurality of brain MR images (30) (S310). Specifically, the processor (110) can construct training data (50) using a plurality of brain MR images (30) obtained by photographing the brains of different subjects, and train a neural network model (10) based on the training data (50). For example, the neural network model (10) may include network models such as a Deep Neural Network (DNN), Recurrent Neural Network (RNN), Bidirectional Recurrent Deep Neural Network (BRDNN), Multilayer Perceptron (MLP), and Convolutional Neural Network (CNN).

[0056] In particular, the processor (110) can train the neural network model (10) to divide the brain of an object included in the brain MR image (30) into multiple regions. To this end, the processor (110) can train the neural network model (10) to identify a specific region of the brain to which each pixel included in the brain MR image (30) belongs by utilizing each brain MR image (30) included in the training data (50) and label data matching each brain MR image (30) (i.e., brain MR images (30) including a segmentation mask labeled for each region). In this process, the neural network model (10) can be trained to analyze the features of the input image at the pixel level and to automatically divide the brain of an object included in the input image into multiple regions. That is, the neural network model (10) can be trained to divide the input image corresponding to multiple regions of the brain and to identify each region.

[0057] FIG. 4 is an exemplary diagram of a method for controlling a computing device (100) that acquires an augmented MR image (40) for a brain MR image (30) using a CDF mapping function according to one embodiment of the present disclosure.

[0058] In training a neural network model (10) according to one embodiment of the present disclosure, the processor (110) can obtain an augmented MR image (40) corresponding to each brain MR image (30) by applying a CDF mapping function set based on scan conditions to each of the plurality of brain MR images (30) included in the training data (50). Then, the processor (110) can train the neural network model (10) based on the training data (50) including the augmented MR image (40). Here, the CDF mapping function is generated based on a Cumulative Distribution Function (CDF) and may be a function set to correct the contrast and brightness of the MR image according to the scan conditions applied to each brain MR image (30). Specifically, the CDF mapping function can match the contrast and brightness between images of different scan conditions by converting the pixel value distribution of each MR image into a Cumulative Distribution Function (CDF) and then re-mapping the pixel values ​​based on this.

[0059] The processor (110) can obtain a histogram of the pixel value distribution of the brain MR image (30), apply a CDF mapping function to obtain a transformed histogram, and ultimately obtain an augmented MR image (40) corresponding to the transformed histogram. More specifically, the processor can calculate the cumulative distribution of pixel values ​​from the histogram of the pixel value distribution of the brain MR image (30), apply a CDF mapping function to transform the calculated cumulative distribution into a cumulative distribution of scan conditions corresponding to the CDF mapping function. Then, the processor (110) can obtain an augmented MR image (40) with a histogram corresponding to the transformed cumulative distribution.

[0060] Meanwhile, according to one embodiment of the present disclosure, the CDF mapping function may include a plurality of CDF mapping functions corresponding to different scan conditions. Accordingly, the processor (110) may identify the scan conditions applied to each brain MR image (30), select a CDF mapping function corresponding to a scan condition different from the scan conditions applied to each brain MR image (30) for each brain MR image (30), and apply the selected CDF mapping function for each brain MR image (30) to each brain MR image (30). For example, referring to FIG. 4, the processor (110) may apply a first CDF mapping function to a first brain MR image (30-1) obtained under a first scan condition to convert it into a third brain MR image (40-3) under a third scan condition. On the other hand, the processor (110) can apply a second CDF mapping function to a second brain MR image (30-2) obtained under a second scan condition to convert it into a fourth brain MR image (40-4) under a third scan condition. That is, although the third scan condition is the same, different CDF mapping functions may be applied depending on the scan condition of the MR image corresponding to the source data. Similarly, the processor (110) can apply a third CDF mapping function to a first brain MR image (30-1) obtained under a first scan condition to convert it into a fifth brain MR image (40-5) under a fourth scan condition, and apply a fourth CDF mapping function to a second brain MR image (30-2) obtained under a second scan condition to convert it into a sixth brain MR image (40-6) under a fourth scan condition.

[0061] Accordingly, the processor (110) can obtain multiple MR brain images, such as those obtained under various scan conditions, using a CDF mapping function without separate additional shooting. That is, the processor (110) can obtain an augmented MR image (40) obtained under scan conditions different from the scan conditions applied to each brain MR image (30) in correspondence with each brain MR image (30).

[0062] Meanwhile, a scan condition according to one embodiment of the present disclosure may be determined according to a set value of a scan parameter including at least one of a pulse sequence, magnetic field strength, slice thickness, and flip angle. That is, the scan condition is determined by a combination of scan parameters of pulse sequence, magnetic field strength, slice thickness, and flip angle, and may also be determined according to a set value of each scan parameter. That is, even if the same combination of scan parameters is used, if the set values ​​of each scan parameter are different, they may be classified as different scan conditions. Meanwhile, a plurality of scan conditions may be set in advance, and the processor (110) may generate a CDF mapping function corresponding to each scan condition in advance and store it in memory (120).

[0063] The augmented MR image (40) and the brain MR image (30) corresponding to the augmented MR image (40) may have different contrast and brightness for the same brain. Scan parameters such as pulse sequence, magnetic field strength, slice thickness, and flip angle are factors that determine the contrast and brightness of the MR image, and each setting value can have a significant impact on the structural characteristics of the image. For example, repetition time (TR) and tissue relaxation time (T1, T2) control the difference in signal strength between tissues, which can result in emphasizing or weakening the brightness or contrast of a specific tissue. In addition, the signal-to-noise ratio (SNR) depends on the magnetic field strength and slice thickness, which affects the clarity and detail representation of the image. Therefore, even when performing MR imaging on the same brain, the MR images may exhibit different contrast and brightness depending on the scan parameter settings. In this way, by obtaining multiple brain MR images (30) with different contrast and brightness, and providing training data (50) including various contrast and brightness when training the neural network model (10), the processor (110) can improve the generalization performance of the neural network model (10).

[0064] Referring again to FIG. 3, when the training of the neural network model (10) is completed, the processor (110) inputs a brain MR image obtained from the subject into the trained neural network model (10) to identify multiple regions included in the subject's brain and to identify the volume of the multiple regions. Specifically, the processor (110) can identify multiple regions included in the subject's brain as a result of segmentation through the neural network model (10), calculate the number of pixels included in each region, and calculate the three-dimensional volume of each region by multiplying the voxel size of the MR image (e.g., 0 0 mm³) by the number of pixels. Through this, the processor (110) can quantitatively determine the volume of multiple regions included in the subject's brain. In the volume calculation process, the processor (110) considers the voxel size of the MR image to reflect the physical space actually occupied by each pixel, and based on this, can accurately measure the volume of each region. For example, if the segmented region represents a structure such as gray matter, white matter, or ventricles of the brain, the processor (110) can calculate the total volume of the brain structure by summing the volumes of all pixels belonging to that region.

[0065] And, the processor (110) can provide numerical information regarding the volume of each region included in the brain of the subject (S330). At this time, the processor (110) can provide numerical information including the volume value of each region included in the brain of the subject, the ratio of the total volume of the brain to the volume of the skull region of the subject, and the ratio of the volume of each region to the volume of the skull region of the subject.

[0066] FIGS. 5a to 5c are exemplary diagrams of a report generated based on numerical information regarding the volume of each region contained in the brain according to one embodiment of the present disclosure.

[0067] In particular, the processor (110) can generate a report on the brain condition of the identified subject by utilizing volume information of multiple regions of the subject's brain. At this time, the report may include volume values ​​for the subject's entire brain and major regions. For example, referring to FIG. 5a, the first region of the report may include volume values ​​for the subject's whole brain, sub-regions of the cerebral cortex (frontal lobe, temporal lobe, parietal lobe, occipital lobe), superior lateral ventricles, inferior lateral ventricles, and hippocampus. Additionally, the volume values ​​of each region may be converted into a ratio relative to the total volume of the skull.

[0068] Additionally, referring to FIG. 5b, another page of the report may provide information on volume values ​​for total intracranial volume, whole brain, cortical gray matter, and cerebral white matter. Through this, the processor (110) can provide the user with important criteria for evaluating the progression of neurodegenerative diseases such as dementia. In particular, since a reduction in volume of specific brain regions, such as the hippocampus, may be prominent in the early stages of dementia, tracking changes in hippocampal volume can play a key role in diagnosis and prognosis assessment. Additionally, the report may provide a reference range (5th - 95th percentile) based on the above volume values, along with a value (% of TIV) calculated as a ratio of the total intracranial volume (TIV). These reference ranges are based on average data from other subjects of the same age and gender and can be used to determine whether the volume of a specific area of ​​the subject falls within the normal range.

[0069] In particular, the processor (110) can analyze the volume change and trend of change of each region by comparing it with past data of the same subject and include this in the report. For example, it can specifically present how much the volume of the subject's hippocampus has decreased over a certain period, or how the degree of atrophy in the cerebral cortex region has changed over time. This provides important information for quantitatively evaluating the progression of dementia diseases such as Alzheimer's and monitoring the effectiveness of treatment.

[0070] Additionally, the processor (110) can visualize multiple regions of the brain identified based on brain MR images as graphic objects, thereby distinguishing and displaying each region of the brain. This allows for a clear visual understanding of the subject's brain condition and helps medical staff and patients understand the condition. The report integrates this quantitative data and visual data to comprehensively evaluate the subject's brain condition and provides important data for the diagnosis and treatment planning of neurodegenerative diseases such as dementia.

[0071] Hereinafter, a method for generating a CDF mapping function for each scan condition according to one embodiment of the present disclosure will be described.

[0072] According to one embodiment of the present disclosure, a processor (110) may identify a first histogram regarding the pixel value distribution of a first reference MR image corresponding to a first scan condition and a second histogram regarding the pixel value distribution of a second reference MR image corresponding to a second scan condition, and generate a CDF mapping function between the first scan condition and the second scan condition. Here, the reference MR image is an MR image representing each scan condition and may be an image that characteristically reflects the pixel value distribution of MR images acquired under the corresponding scan condition. The reference MR image may be used to generate a cumulative distribution function (CDF) for a specific scan condition or to generate a CDF mapping function through comparison with an MR image under a different scan condition.

[0073] The processor (110) can generate a CDF mapping function according to each scan condition. To do this, the processor (110) analyzes the pixel value distribution of a reference MR image obtained under a specific scan condition, generates a cumulative distribution function (CDF) based on the distribution, compares the pixel value distribution between reference MR images obtained under different scan conditions, and maps them to generate a CDF mapping function that matches contrast and brightness.

[0074] At this time, the processor (110) can identify a first CDF function corresponding to a first reference MR image based on a first histogram and identify a second CDF function corresponding to a second reference MR image based on a second histogram. Then, the processor (110) can obtain a first CDF mapping function that converts from a first scan condition to a second scan condition and a second CDF mapping function that converts from a second scan condition to a first scan condition based on the first CDF function and the second CDF function.

[0075] Specifically, the processor (110) can generate a first histogram based on the pixel value distribution of a first reference MR image obtained under a first scan condition, and generate a second histogram based on the pixel value distribution of a second reference MR image obtained under a second scan condition. These histograms visually represent the distribution of pixel values ​​and may serve as indicators that quantitatively express the contrast and brightness characteristics of each image. Furthermore, the processor (110) generates a first CDF function by calculating the cumulative distribution from the first histogram and generates a second CDF function in the same manner from the second histogram. In this process, each CDF function can represent the overall distribution characteristics of the image by cumulatively calculating the pixel value distribution of the corresponding image. The processor (110) can generate a CDF mapping function by comparing these two CDF functions and deriving a mapping rule necessary for converting from the first scan condition to the second scan condition, or vice versa.

[0076] FIG. 6 is an exemplary diagram illustrating a method for generating a CDF mapping function according to one embodiment of the present disclosure.

[0077] Referring to FIG. 6, if the first scan condition is one in which the flip angle included in the scan parameters is set to 12 degrees and the second scan condition is one in which the flip angle is set to 8 degrees, the processor (110) can generate a histogram of pixel values ​​of the brain MR image acquired at a flip angle of 12 degrees and the brain MR image acquired at a flip angle of 8 degrees, and generate a first CDF mapping function that converts from the first scan condition to the second scan condition and a second CDF mapping function that converts from the second scan condition to the first scan condition. At this time, when the brain MR image obtained under the first scan condition is converted based on the first CDF mapping function, the CDF of the converted brain MR image is similar to the CDF of the brain MR image obtained under the second scan condition (flip angle 8 degrees), and when the brain MR image obtained under the second scan condition is converted based on the second CDF mapping function, the CDF of the converted brain MR image is similar to the CDF of the brain MR image obtained under the first scan condition (flip angle 12 degrees).

[0078] Meanwhile, according to one embodiment of the present disclosure, the processor (110) identifies a scan condition applied to a brain MR image obtained from a subject, and if the identified scan condition corresponds to a new scan condition that does not correspond to a plurality of CDF mapping functions, it can generate a new CDF mapping function based on the new scan condition. Meanwhile, the processor (110) identifies a scan condition for a specific brain MR image, and if the scan condition is a new scan condition that is not included in the previously generated CDF mapping function, it can generate a new CDF mapping function. In this process, the processor (110) analyzes the pixel value distribution of the MR image obtained under the new scan condition to generate a new histogram and CDF function, and can additionally derive a new mapping function by comparing this with other existing scan conditions. FIG. 7 is a detailed block diagram of a computing device (100) that analyzes a brain MR image based on a deep learning model (10) according to another embodiment of the present disclosure. Referring to FIG. 7, the computing device (700) may include one or more processors (710), memory (720), communication interface (730), image processing unit (740), display (750), user interface (760), and output interface (770). A detailed description of the configurations shown in FIG. 7 that overlap with the configuration shown in FIG. 2 (one or more processors (710) and memory (720)) is omitted.

[0079] A communication interface (730) according to one embodiment of the present disclosure may be understood as a configuration unit that transmits and receives data through any known wired or wireless communication system. For example, the communication interface (730) may perform data transmission and reception using a wired or wireless communication system such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), 5th generation mobile communication (5G), ultrawide-band wireless communication, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity (Wi-Fi), near field communication (NFC), or Bluetooth. Since the communication systems described above are merely examples, wired or wireless communication systems for SS data transmission and reception of the communication interface (730) may be applied in various ways other than those described above.

[0080] The communication interface (730) can receive data necessary for the processor (110) to perform calculations through wired or wireless communication with any system or any client. Additionally, the communication interface (730) can transmit data generated through the calculations of the processor (110) through wired or wireless communication with any system or any client. For example, the communication interface (730) can receive brain MR images (and multiple brain MR images included in the training data (50)) through communication with a cloud server or other computing device that performs tasks such as PACS, databases, and standardization of medical data within a hospital environment. The communication interface (730) can transmit output data of the neural network model (10), and intermediate data and processed data derived during the calculation process of the processor (110), through communication with the aforementioned database, server, or other computing device.

[0081] The image processing unit (740) can obtain a brain MR image corresponding to the magnetic resonance signal (20) by performing image processing (e.g., inverse Fourier operation, etc.) on the magnetic resonance signal (20) obtained through the scanning unit or the magnetic resonance signal (20) obtained through the communication interface. Alternatively, the image processing unit (740) may restore the quality of the obtained brain MR image using a pre-trained neural network model.

[0082] The display (750) can display various images. Here, the images include both still images and videos. In particular, the display (750) can display acquired or restored brain MR images, and can also provide information related to brain MR images (e.g., reports, etc.) to a user or subject. The display (750) can be implemented as various types of displays (750), such as LCD (Liquid Crystal Display Panel), OLED (Organic Light Emitting Diodes), LCoS (Liquid Crystal on Silicon), DLP (Digital Light Processing), etc. Additionally, the display (750) may also include a driving circuit, a backlight unit, etc., which can be implemented in forms such as a-si TFT, LTPS (low temperature poly silicon) TFT, OTFT (organic TFT), etc.

[0083] Meanwhile, the display (750) may be implemented as a touch screen by combining it with a touch panel, and in this case, the display (750) can perform the function of an input interface that receives user touch input as well as an output interface that outputs an image through the touch screen.

[0084] The user interface (760) can receive control commands regarding the overall operation of the computing device (100) from the user. For example, the user interface (760) can receive information regarding object information, parameter information, scan conditions, pulse sequences, etc. from the user. To this end, the user interface (760) can be implemented using a keyboard, mouse, microphone, etc.

[0085] The output interface (770) can output information acquired by the computing device (100) to the outside. To this end, the output interface (770) can be implemented as a speaker, etc. The speaker can output voice messages related to the brain MR image reconstruction scenario.

[0086] Meanwhile, according to one embodiment of the present disclosure, the computing device (100) may directly scan a subject (e.g., a patient) to obtain a magnetic resonance image (i.e., a brain MR image). To this end, the computing device (100) may further include a scanning unit, and the scanning unit includes a static magnetic field unit, a gradient magnetic field unit, and an RF coil unit.

[0087] The scanning unit may be implemented in a form that allows an object (e.g., a patient) to be inserted into the empty internal space of the scanning unit. To this end, the scanning unit may further include a table. The scanning unit may form a static magnetic field and a gradient magnetic field in the internal space and irradiate an RF signal. Specifically, the static magnetic field unit may form a static magnetic field to align the direction of the magnetic dipole moments of the atomic nuclei contained in the object with the direction of the static magnetic field. To this end, the static magnetic field unit may be implemented as a permanent magnet or as a superconducting magnet using a cooling coil.

[0088] The gradient magnetic field unit can form a gradient magnetic field by applying a gradient to a static magnetic field according to a control signal from a processor. The gradient magnetic field unit includes X, Y, and Z coils that form gradient magnetic fields in the mutually orthogonal X, Y, and Z axis directions, and generates a gradient signal according to the shooting position so that resonance frequencies can be induced differently for each part of the object.

[0089] The RF coil unit can irradiate an RF signal (e.g., an RF pulse sequence) onto a target according to a control signal from a processor. Additionally, the RF coil unit can receive a magnetic resonance signal (MR signal) emitted from the target. The RF coil unit can transmit an RF signal with a frequency identical to the frequency of precession to the target, directed toward a precessing atomic nucleus, stop transmitting the RF signal, and then receive the magnetic resonance signal emitted from the target.

[0090] The RF coil section may be implemented as a transmitting RF coil that generates electromagnetic waves having a radio frequency corresponding to the type of atomic nucleus and a receiving RF coil that receives electromagnetic waves radiated from the atomic nucleus, respectively, or as a single RF transmitting and receiving coil having both transmitting and receiving functions.

[0091] The various embodiments of the present disclosure described above may be combined with additional embodiments and modified to the extent understandable to those skilled in the art in light of the detailed description above. The embodiments of the present disclosure are illustrative in all respects and should be understood as not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form. Accordingly, all modifications or variations derived from the meaning, scope, and equivalents of the claims of the present disclosure should be interpreted as being included within the scope of the present disclosure.

Claims

1. A method for analyzing brain MR images based on a deep learning model, performed by a computing device comprising at least one processor, wherein A step of training a neural network model based on training data including multiple brain MR images; When the above learning is completed, a step of inputting a brain MR image acquired from a subject into the learned neural network model to identify a plurality of regions included in the subject's brain and to identify the volume of the plurality of regions; and The method includes the step of providing numerical information regarding the volume of each region included in the brain of the subject; The above-mentioned training step is, A step of obtaining an augmented MR image corresponding to each brain MR image by applying a CDF mapping function set based on scan conditions to each of the plurality of brain MR images included in the training data; and A step of training the neural network model based on training data including the augmented MR image; comprising method.

2. In Paragraph 1, The above CDF mapping function is, It includes multiple CDF mapping functions corresponding to different scan conditions, and The step of acquiring the above-mentioned augmented MR image is, A step comprising: identifying the scan conditions applied to each of the above-mentioned brain MR images, selecting a CDF mapping function corresponding to a scan condition different from the scan conditions applied to each of the above-mentioned brain MR images for each of the above-mentioned brain MR images, and applying the CDF mapping function selected for each of the above-mentioned brain MR images to each of the above-mentioned brain MR images; method.

3. In Paragraph 2, The step of acquiring the above-mentioned augmented MR image is, A step comprising: acquiring the augmented MR image obtained under scan conditions different from the scan conditions applied to each of the brain MR images, corresponding to each of the brain MR images; method.

4. In Paragraph 3, The above-mentioned augmented MR image and the brain MR image corresponding to the above-mentioned augmented MR image are characterized by having different contrasts and brightnesses for the same brain. method.

5. In Paragraph 2, Identifying scan conditions applied to a brain MR image acquired from the above subject, and if the identified scan condition corresponds to a new scan condition that does not correspond to the plurality of CDF mapping functions, newly generating the CDF mapping function based on the new scan condition. method.

6. In Paragraph 1, The above scan conditions are, Determined according to a set value of a scan parameter including at least one of a pulse sequence, magnetic field strength, slice thickness, and flip angle, method.

7. In Paragraph 1, A step comprising: identifying a first histogram regarding the pixel value distribution of a first reference MR image corresponding to a first scan condition and a second histogram regarding the pixel value distribution of a second reference MR image corresponding to a second scan condition, and generating a CDF mapping function between the first scan condition and the second scan condition; method.

8. In Paragraph 7, The step of generating the above CDF mapping function is, A step of identifying a first CDF function corresponding to the first reference MR image based on the first histogram; A step of identifying a second CDF function corresponding to the second reference MR image based on the second histogram; and Based on the first CDF function and the second CDF function, the method comprises the step of obtaining a first CDF mapping function that converts from the first scan condition to the second scan condition and a second CDF mapping function that converts from the second scan condition to the first scan condition. method.

9. In Paragraph 1, The step of providing the above numerical information is, The method comprises the step of providing numerical information including the volume value of each of the above regions, the ratio of the total volume of the brain to the volume of the skull region of the subject, and the ratio of the volume of each of the above regions to the volume of the skull region of the subject. method.

10. In a computing device for analyzing brain MR images based on a deep learning model, Memory for storing at least one instruction; and By executing at least one instruction, a neural network model is trained based on training data including a plurality of brain MR images, and when the training is completed, a brain MR image obtained from a subject is input into the trained neural network model to identify a plurality of regions included in the subject's brain, identify the volumes of the plurality of regions, and provide numerical information regarding the volume of each region included in the subject's brain. The at least one processor applies a CDF mapping function set based on scan conditions to each of the plurality of brain MR images included in the training data, thereby acquiring an augmented MR image corresponding to each brain MR image, and trains the neural network model based on the training data including the augmented MR image. Computing device.

11. A computer program stored on a computer-readable storage medium, wherein, when executed on one or more processors, the computer program performs operations for analyzing brain MR images based on a deep learning model, and said operations are, The operation of training a neural network model based on training data including multiple brain MR images; When the above learning is completed, the operation of inputting a brain MR image acquired from a subject into the learned neural network model to identify a plurality of regions included in the subject's brain and to identify the volume of the plurality of regions; and The operation of providing numerical information regarding the volume of each region included in the brain of the above-mentioned subject; is included, The above-mentioned learning operation is, The operation of acquiring an augmented MR image corresponding to each brain MR image by applying a CDF mapping function set based on scan conditions to each of the plurality of brain MR images included in the training data; and The operation of training the neural network model based on training data including the augmented MR image; Computer program.

Images