KR101995383B1 - Method for determining brain disorder based on feature ranking of medical image and apparatus using the same - Google Patents

Abstract

The present disclosure discloses a method for determining a brain disease using a medical image, and an apparatus using the same. Specifically, according to the method of the present invention, a computing apparatus acquires a brain image of a subject, uses a parcellation net to perform parcellation of a plurality of predetermined brain areas from the brain image as related to a brain disease, generates classification information on the brain disease based on result information of the parcellation, and updates the parcellation net based on the result information of the parcellation and the classification information.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for determining a brain disease based on a feature rank of a medical image,

Disclosed herein is a method for determining a brain disease using a medical image and an apparatus using the method. Specifically, according to the method of the present invention, a computing device acquires a brain image of a subject, and uses a compartment network to parcellate a plurality of brain regions previously designated as being related to brain diseases from the brain image , Generates classification information on brain diseases based on the result information of the compartment, and updates the compartment network based on the result information of the compartment and the classification information.

With respect to the determination of severity of dementia caused by brain diseases such as Alzheimer's disease, it is common to confirm whether or not a brain atrophy phenomenon, which is one of the causes of degenerative brain diseases, is observed. However, in order to classify the subject from the T1 image of MR (magnetic resonance) to Al (Alzheimer's disease) / MCI (mild cognitive impairment) and CN (Eg, brain atrophy or texture change in a specific brain region), there is not much that can be judged as AD, and end-to-end classification of normal and abnormal (eg 3D CNN; 3 -dimensional convolutional neural network) was not very good in terms of performance.

Therefore, in the conventional approach, a two-phase approach using a model that extracts significant information from T1 images by MR and classifies normal and abnormal is common.

However, the two-phase approach differs from the process of classifying the AD / MCI and CN (the second phase) by the process of feature ranking (the first phase) There was a limit to performance limitations.

In order to overcome the limitations of the conventional two-phase approach, the present inventors have proposed a method of updating the parameter by connecting the evaluation result of the second phase to the first phase by RL (reinforcement learning) We propose a method to improve performance by providing greater speed and accuracy.

The present invention overcomes the limitations of the two-phase approach using artificial intelligence as well as the limitations of conventional statistical based {e.g., age, sex, volume information, statistical analysis using handcrafted features, The purpose of this study is to improve the accuracy and efficiency of diagnosis of brain diseases by extracting image features (shape, texture, global position, structural features, etc.) Of 2D medical images by machine learning and using them for brain disease diagnosis and prediction.

Specifically, the present invention automatically analyzes the brain image through machine learning, thereby reducing diagnosis errors based on machine learning based diagnosis for cases in which a person is not seen or is difficult to distinguish, It is aimed at helping to improve the speed.

The characteristic configuration of the present invention for achieving the object of the present invention as described above and realizing the characteristic effects of the present invention described below is as follows.

According to an aspect of the present invention, there is provided a method of determining brain disease based on a feature ranking for a medical image, the method comprising: (a) obtaining a brain image of a subject Supporting another device associated with the computing device to acquire the brain image; (b) supporting the computing device to parcel the brain image from the brain image or to divide the other apparatuses into a plurality of brain regions that are previously specified to be related to brain diseases using the compartment network; (c) supporting the computing device to generate classification information on brain diseases based on the result information of the compartment or to cause the other device to generate using the disease prediction network; And (d) supporting the computing device to update the partition network or to update the other device based on the classification result information and the classification information using a reinforcement learning module .

According to another aspect of the present invention, there is also provided a computer program stored in a machine readable non-transitory medium, comprising instructions embodied to perform the method according to the invention.

According to still another aspect of the present invention, there is provided a computing device for determining a brain disease based on a feature ranking of a medical image, the apparatus comprising: a communication unit for obtaining a brain image of a subject; And (i) a process of implementing a parcel network for parcellating a plurality of brain regions previously designated as being related to brain diseases from the brain image; (ii) a process for implementing a disease prediction network that generates classification information on brain diseases based on the result information of the segment, and (iii) a process for updating the classification network based on the result information of the segment and the classification information And a processor for performing a process of implementing a reinforcement learning module.

According to an exemplary embodiment of the present disclosure, it is possible to overcome the limitations of conventional statistical-based disease judgment or prediction to extract the image features of a three-dimensional medical image by machine learning and utilize it for the determination and prediction of brain diseases, There is an effect.

In addition, according to the exemplary embodiment, there is an advantage that it is possible to apply an additional restriction to the parameter update by applying a series of reinforcement learning methodology as computed compensation as well as map learning to segment the original brain image into a segmentation network.

Thus, according to an exemplary embodiment, the partition action of the first phase and the normal / abnormal class action of the second phase are configured to be end-to-end learning as if performing one step without being strictly divided, There is an effect that a performance gain can be obtained.

In addition, according to the exemplary embodiment, the importance of the brain region that is not clinically known can be reflected in the machine learning, and the accuracy can be further improved.

Those skilled in the art will appreciate that the present invention can be applied to various types of 3-dimensional images of various modalities, and in particular, It will be understood that the present invention can be applied to various types of image systems and that the method of the present invention does not depend on a specific type of image or platform.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate only typical embodiments of the invention and are capable of illustration without departing from the spirit or scope of the invention.
1 is a conceptual diagram schematically showing an exemplary configuration of a computing device that performs a method of determining a brain disease based on a feature rank of a medical image according to the present invention (hereinafter referred to as a "feature-based brain disease determination method"). to be.
FIG. 2 is an exemplary block diagram illustrating hardware or software components of a computing device used in a feature-based brain disease determination method of the present invention. FIG. 3 is a block diagram of a machine And schematically shows a process of training and applying a learning model.
4 is a flow chart illustrating a method of determining a brain disease according to the present invention.
FIG. 5 is a conceptual diagram showing an example of a partition network that can be used for compartmentalization of a brain image in a feature-based brain disease determination method according to the present invention.
FIG. 6 is a conceptual diagram illustrating an example of a disease prediction network that can be used to generate classification information for classifying brain images as a brain disease disease or normal in the feature-based brain disease determination method of the present invention.

The following detailed description of the invention refers to the accompanying drawings, which illustrate, by way of example, specific embodiments in which the invention may be practiced in order to clarify the objects, technical solutions and advantages of the invention. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention.

The term "image" or "image data" used throughout the description and claims of the present invention refers to multidimensional data composed of discrete image elements (e.g., pixels for two dimensional images) Refers to a digital representation of an object that is visible (e.g., displayed on a video screen) or that object (such as a file corresponding to pixel output, e.g., CT, MRI detector, etc.).

For example, the term "image" or "image" may be represented by computed tomography (CT), magnetic resonance imaging (MRI), ultrasound, or any other medical imaging system known in the art And may be a medical image of the collected subject. The image may not necessarily be provided in a medical context but may be provided in a non-medical context, for example, an X-ray for security search.

For the sake of convenience of explanation, MRI image data is shown in the presented drawings as an exemplary image format (modality). However, those skilled in the art will appreciate that the image formats used in various embodiments of the present invention may be extended to three-dimensional images such as CT, PET (positron emission tomography), PET-CT, SPECT, SPECT-CT, MR- It is to be understood that the invention is not limited to the above-exemplified embodiments.

Throughout the detailed description and claims of the present invention, the term 'DICOM' (Digital Imaging and Communications in Medicine) is a generic term for various standards used in digital image representation and communication in medical devices, The standard is presented at the Joint Committee composed of the American Radiation Medical Association (ACR) and the American Electrical Manufacturers Association (NEMA).

In addition, the term 'Picture Archiving and Communication System (PACS)' refers to a system for storing, processing and transmitting according to the DICOM standard throughout the detailed description and claims of the present invention, , And MRI can be stored in the DICOM format and transmitted to a terminal inside or outside the hospital through the network, and the result of reading and the medical record can be added to the terminal.

Throughout the detailed description and claims of the present invention, 'learning' or 'learning' refers to performing machine learning through computing according to a procedure, It is not intended to refer to training, which is used in a generally accepted sense of machine learning.

Also, throughout the description and claims of this invention, the word 'comprise' and variations thereof are not intended to exclude other technical features, additions, elements or steps. Also, 'one' or 'one' is used in more than one meaning, and 'another' is limited to at least the second.

Other objects, advantages and features of the present invention will become apparent to those skilled in the art from this description, and in part from the practice of the invention. The following examples and figures are provided by way of illustration and are not intended to limit the invention. Accordingly, the details disclosed herein with respect to a particular structure or function are not to be construed in a limiting sense, but merely as being representative of the general inventive concept providing a guideline for carrying out the invention in various detail structures, It should be interpreted as basic data.

Moreover, the present invention encompasses all possible combinations of embodiments shown herein. It should be understood that the various embodiments of the present invention are different, but need not be mutually exclusive. For example, certain features, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in connection with one embodiment. It should also be understood that the position or arrangement of individual components within each disclosed embodiment may be varied without departing from the spirit and scope of the present invention. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is to be limited only by the appended claims, along with the full scope of equivalents to which such claims are entitled, if properly explained. In the drawings, like reference numerals refer to the same or similar functions throughout the several views.

Unless otherwise indicated herein or clearly contradicted by context, items referred to in the singular are intended to encompass a plurality unless otherwise specified in the context. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily carry out the present invention.

1 is a conceptual diagram schematically illustrating an exemplary configuration of a computing device that performs a feature-based brain disease determination method according to the present invention.

1, a computing device 100 according to an embodiment of the present invention includes a communication unit 110 and a processor 120. The communication unit 110 communicates with an external computing device (not shown) Communication is possible.

In particular, the computing device 100 may be implemented as a computer-readable medium, such as conventional computer hardware (e.g., a computer processor, memory, storage, input and output devices, Electronic communication devices, electronic information storage systems such as network-attached storage (NAS) and storage area networks (SAN), and computer software (i.e., computing devices that enable a computing device to function in a particular manner) Commands) to achieve the desired system performance.

The communication unit 110 of the computing device can send and receive requests and responses to and from other interworking computing devices. As an example, such requests and responses can be made by the same transmission control protocol (TCP) session But not limited to, a user datagram protocol (UDP) datagram, for example. In addition, in a broad sense, the communication unit 110 may include a keyboard, a mouse, an external input device, a printer, a display, and other external output devices for receiving commands or instructions.

The processor 120 of the computing device may also be a micro processing unit (MPU), a central processing unit (CPU), a graphics processing unit (GPU), a neural processing unit (NPU) or a tensor processing unit ), A data bus, and the like. It may further include a software configuration of an operating system and an application that performs a specific purpose.

FIG. 2 is an exemplary block diagram illustrating hardware or software components of a computing device used in a feature-based brain disease determination method of the present invention. FIG. 3 is a block diagram of a machine And schematically shows a process of training and applying a learning model.

Referring first to FIG. 2, a brief outline of a method and an apparatus according to the present invention, a computing device 100 may include an image acquisition unit 210 as a component thereof. The image acquisition unit 210 is configured to acquire data of a brain image to which the method according to the present invention is applied (S100). The individual modules shown in FIG. 2 include, for example, a communication unit It will be understood by those skilled in the art that the present invention can be implemented by interfacing the communication unit 110 and the processor 120 or the communication unit 110 and the processor 120. [

The brain image may be obtained, for example, from an external image storage system such as a radiographic apparatus or a medical image storage and transmission system (PACS) interlocked through the communication unit 110, but is not limited thereto. For example, the brain image may be acquired by the image acquisition unit 210 of the computing device 100 after the image captured by the medical imaging device is transmitted to the PACS according to the DICOM standard. Preferably, the brain image may be a three-dimensional image. For example, the brain image may be an MRI image, particularly an MR T1w image.

Next, the obtained brain image can be transmitted to the divisional network 220. Specifically, the divisional network 220 extracts a plurality of brain images related to brain diseases, for example, dementia caused by Alzheimer's disease And is configured to parcellate the brain region from the brain image.

When the compartment is completed, the disease prediction network 230 is configured to generate classification information on brain diseases based on the result information of the compartment.

The reinforcement learning unit 240 (242, 244, 246) is configured to update the partition network 220 based on the result information of the partition and the classification information. Accordingly, the partition performance of the partition network 220 can be continuously improved, and the individual components constituting the reinforcement learning unit 240, the compensation unit 242, the reinforcement learning cricity 244, (246) will be described later.

Meanwhile, the output unit 250 (not shown) may provide classification information on brain diseases to an external entity. Here, the external entity includes a user of the computing device 100 performing the method according to the present invention, a manager, a medical professional in charge of the subject, and the like. However, the external entity includes information on brain imaging, It is to be understood that any subject that needs information of whether or not it is included includes any subject. When the external entity is a human, the output unit 250 can provide classification information to an external entity through a user interface displayed on a predetermined output device, e.g., a display.

3, a computing device (referred to as a "recognition system") according to one embodiment of the present invention, including the above-described compartment network 220, disease prediction network 230, The machine learning module 300b of the machine learning module 300b may be transmitted or distributed from a separate learning system that trains the machine learning module 300a using data labeled with areas of the brain image and the brain image as training data . In other words, the machine learning module 300b used in the computing device 100 according to an embodiment of the present invention may be one that has been pre-trained by a separate learning system 300a.

Specific functions and effects of the respective components schematically described with reference to FIGS. 2 and 3 will be described later in detail with reference to FIGS. 4 to 6. FIG. Although the components shown in FIG. 2 are illustrated as being realized in one computing device for convenience of explanation, it will be understood that the computing device 100 performing the method of the present invention may be configured such that a plurality of devices are interlocked with each other . That is, each step of the method according to the present invention can be performed by one computing device directly or by supporting the one computing device to perform another computing device that is interlocked with the one computing device.

4 is a flow chart illustrating a method of determining a brain disease according to the present invention.

Referring to FIG. 4, a feature-based brain disease determination method according to the present invention includes: an image acquisition unit 210 implemented by a computing device 100 acquiring a brain image x of a subject; (S100) supporting another device, which is linked through the communication unit 110 of the mobile terminal 100, to acquire the brain image x. It will be understood that the present invention is not limited to the modality of the MRI image attached to the drawings, but can be applied to various image formats in general.

Next, a feature-based brain disease determination method according to the present invention is characterized in that a parcellation network 220 implemented by the computing device 100 divides a plurality of brain regions, x) or supporting the other device to partition (S200).

Here, the brain disease may be dementia, especially dementia caused by Alzheimer's disease.

Here, the predetermined number of brain regions may be 4 or more and less than 130 brain regions.

Specifically, in step S200, the compartment may be performed by a compartment net 220, which may include (i) a brain imaging of a number of patients with a brain disorder; And (ii) at least one of a plurality of brain images of a cognitively normal (CN) person and a human brain image of a plurality of mild cognitive impairment (MCI) . The three possible combinations here are: (i) a data set of brain disorders, cognitive normal, and mild to moderate disorders; (ii) brain disorders; cognitive normal data sets; or (iii) brain disorders, Data set. It goes without saying that various classes may be added to the data set for training besides brain diseases, cognitive normal, and hardness disorders.

FIG. 5 is a conceptual diagram showing an example of a partition network that can be used for compartmentalization of a brain image in a feature-based brain disease determination method according to the present invention.

Referring to FIG. 5, the partitioning network 220 for a partition is composed of a plurality of blocks for calculating a voxel-level segmentation from input volume volumes, Includes a plurality of layers.

In the neural network constituting the divided network 220, the characteristic of each image is automatically learned by learning a large amount of data in a structure of a multi-layer artificial neural network, and the value of the objective function, that is, the error of the classification result, , And has a technical feature that can extract and classify various levels of features from low-level features such as points, lines, and faces to complex and meaningful high-level features.

The segmentation network 220 can perform segmentation of the brain image by extracting image features including the shape, texture, global position, and structural characteristics of the input brain image. Those of ordinary skill in the art will be able to contemplate various embodiments of a compartment network that exerts the effects that the present invention is intended to achieve.

Advantages of the partition network according to the present invention are that, in the conventional pixel-based image processing technique before using the depth neural network, the co-registration consumes the processing time in about 10 minutes, and the sophisticated segment algorithm (For example, in the case of the conventional Freesurfer), it can be performed in a short time without consuming an enormous amount of computation, such as consuming processing time on the start of the brain. The brain disease prediction, which takes from several tens of minutes to several hours, Which can be performed within a relatively short period of time.

In FIG. 4, a feature map calculated by the segment of the segment network 220 is shown as x ', which does not refer directly to the result of segmentation but also to the intermediate layer of the segment network Refers to the derived value.

Next, a feature-based brain disease determination method according to the present invention is characterized in that a disease prediction network 230 implemented by the computing device 100 is classified into classification based on brain information (S300) to generate information or to cause the other device to generate the information.

Here, the classification information may be information for classifying the subject as (i) a disorder of a brain disease disease or a mild disorder and (ii) a normal state. For example, using the x 'illustrated in FIG. 4, whether or not a brain disease disease (y') can be derived.

Preferably, in step S300, the risk of developing the brain disease may be calculated together with the classification information.

FIG. 6 is a conceptual diagram illustrating an example of a disease prediction network that can be used to generate classification information for classifying brain images as a brain disease disease or normal in the feature-based brain disease determination method of the present invention.

Advantageously, step S300 may be performed by a neural network trained to determine the state of interconnections between brain regions of the brain image that are the result of the segment. For example, the disease prediction network 230 may include a 3D CNN (3-dimensional convolutional neural network), and in particular, a graph convolutional neural network (GCN).

6, an output layer value is calculated from an input layer value of a disease prediction network, and hidden layer activation between hidden layers is calculated And how the feature values in the hidden layer are distributed according to the state of interconnections between brain regions. The advantage of this model is that the interconnection, coupling activity, volume distribution, etc. of each brain region of the brain image, which is the result of the preceding compartment, can be reflected in the classification of brain disease incidence.

4, a feature-based brain disease determination method according to the present invention is characterized in that a reinforcement learning module 240 implemented by the computing device 100 includes a result information of the compartment and the classification information (S400) to update the partition network (220) or to update the other device based on the update information (S400).

The step of performing reinforcement learning by the reinforcement learning unit (S400) will be described in more detail as follows.

First, in the step S400, the compensating unit 242 belonging to the reinforcement learning unit 240 calculates a compensation value (i.e., a compensation value) based on the parcellation output volume for each brain region of the brain image, (step S420) of calculating the reward (r) or supporting the other device to calculate (step S420).

For example, the compensation value may be calculated by performing an independent t-test based on the volume and the classification information, and ranking by using the resultant p-value .

As a specific example, for a data set calculated with y '= 0 (no brain disease, normal) and a data set calculated with y' = 1 (brain disease prevalence), t- , And the p-value is obtained and ranked according to the p-value, which can be regarded as a 'significant degree of the corresponding data', which can be defined as the compensation value (r) of the reinforcement learning.

Next, in step S400, after step S420, the reinforcement learning critic 244 belonging to the reinforcement learning unit 240 determines whether the reinforcement learning critic 244 belongs to the reinforcement learning critic (Step S440) (step S440) of evaluating an action of the reinforcement learning actor 246 belonging to the learning unit 240 or assisting the other device to evaluate the action . The error used here may be a TD (Temporal Difference) error.

Here, the action is about which value is more importantly accepted in x ', and the reinforcement learning unit 240 can optimize the actions for maximizing the accumulated value of r.

After step S440, the step S400 is a step in which the enhanced learning actor 246 generates an advantage value to be applied to the output loss of the divisional network 220 or causes the other device to generate (S460; not shown); And a step S480 (not shown) of further training the partition network 220 through policy slope updating according to the advantage value or supporting the other device to further train.

The function of the policy slope update and the advantage value in these steps (S460 and S480) can be given, for example, by the following Equation (1).

Where θ is the policy gradient, which is the gradient of the RL policy network parameter. Also, t refers to the current time step, s _t refers to the state at the current time step, and w refers to the critic network parameter. And v (s, w) refers to the value (critical) function at state s mediated by w, and π refers to policy (network). A _t refers to the action in t time steps, and π (A _t | S _t , θ) refers to the policy when taking A _t as an action in state s _t , mediated by θ.

As described above, the method according to the present invention connects the classification information, which is a result evaluated by the disease prediction network 230, to the compartment network 220 by reinforcement learning, and updates the parameters of the compartment network, The classification performance can be obtained. In particular, there is an advantage that the importance of the brain region that is not clinically known can be reflected in the learning.

Since the parameter updating of the divided network 220 by the reinforcement learning unit can be repeatedly performed on a plurality of brain images, the conditions for terminating the repeated execution can be easily understood by a general technician, And will not be described because it may obscure the gist of the present invention.

If the classification information is generated by the disease prediction network 230, the feature-based brain disease determination method according to the present invention may be applied to the use of an external entity, (Step S400 ') in which the terminal 250 (not shown) provides the generated classification information to an external entity or provides the other device with the classification information.

As described so far, the present invention has the effect of greatly improving the speed and accuracy of diagnosis of dementia caused by brain diseases, particularly Alzheimer's disease, over all of the embodiments and modifications.

Based on the description of the embodiments above, those skilled in the art will recognize that the methods and / or processes of the present invention and their steps may be implemented in hardware, software, or any combination of hardware and software suitable for the particular application Points can be clearly understood. The hardware may include special features or components of a general purpose computer and / or a dedicated computing device or a specific computing device or a particular computing device. The processes may be realized by one or more microprocessors, microcontrollers, embedded microcontrollers, programmable digital signal processors or other programmable devices having internal and / or external memory. Additionally or alternatively, the processes can be configured to process application specific integrated circuits (ASICs), programmable gate arrays, programmable array logic (PAL) Or any other device or combination of devices. Furthermore, the objects of the technical solution of the present invention, or portions contributing to the prior art, may be implemented in the form of program instructions that can be executed through various computer components and recorded on a machine-readable recording medium. The machine-readable recording medium may include program commands, data files, data structures, and the like, alone or in combination. The program instructions recorded on the machine-readable recording medium may be those specially designed and constructed for the present invention or may be those known to those of ordinary skill in the computer software arts. Examples of the machine-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tape, optical recording media such as CD-ROM, DVD, Blu-ray, magneto-optical media such as floptical disks magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include, but are not limited to, any of the above devices, as well as a heterogeneous combination of processors, processor architectures or combinations of different hardware and software, Which may be constructed using a structured programming language such as C, an object-oriented programming language such as C ++ or an advanced or low-level programming language (assembly language, hardware description languages and database programming languages and techniques) This includes not only bytecode, but also high-level language code that can be executed by a computer using an interpreter or the like.

Thus, in one aspect of the present invention, when the methods and combinations described above are performed by one or more computing devices, combinations of the methods and methods may be implemented as executable code that performs each of the steps. In another aspect, the method may be implemented as systems for performing the steps, and the methods may be distributed in various ways throughout the devices, or all functions may be integrated into one dedicated, stand-alone device, or other hardware. In yet another aspect, the means for performing the steps associated with the processes described above may include any of the hardware and / or software described above. All such sequential combinations and combinations are intended to be within the scope of this disclosure.

For example, the hardware device may be configured to operate as one or more software modules to perform processing in accordance with the present invention, and vice versa. The hardware device may include a processor, such as an MPU, CPU, GPU, TPU, coupled to a memory, such as ROM / RAM, for storing program instructions and configured to execute instructions stored in the memory, And a communication unit capable of receiving and sending data. In addition, the hardware device may include a keyboard, a mouse, and other external input devices for receiving commands generated by the developers.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, Those skilled in the art will appreciate that various modifications and changes may be made thereto without departing from the scope of the present invention.

Therefore, the spirit of the present invention should not be construed as being limited to the above-described embodiments, and all of the equivalents or equivalents of the claims, as well as the following claims, I will say.

Such equally or equivalently modified means include, for example, a logically equivalent method which can produce the same result as the method according to the present invention, Should not be limited by the foregoing examples, but should be understood in the broadest sense permissible by law.

Claims (10)

A method for determining a brain disease based on a feature rank of a medical image,
(a) supporting a computing device to acquire a brain image of a subject or to acquire the brain image by another device associated with the computing device;
(b) supporting the computing device to parcel the brain image from the brain image or to divide the other apparatuses into a plurality of brain regions that are previously specified to be related to brain diseases using the compartment network;
(c) supporting the computing device to generate classification information related to brain diseases displayed on the brain image based on the result information of the compartment or to cause the other device to generate using the disease prediction network; And
(d) the computing device updates the partition network or supports the other device to update based on the classification result information and the classification information, using a reinforcement learning module
, ≪ / RTI &
The step (d)
(d1) The computing device calculates a reward based on a parcellation output volume of each brain region of the brain image that is a result of the compartment, using the compensation calculating unit of the reinforcement learning unit Supporting the other device to calculate;
(d2) the computing device is configured to use the reinforcement learning criterion of the reinforcement learning unit to calculate, based on an error of the accumulated value of the compensation value, the reinforcement learning criterion of the reinforcement learning actor of the reinforcement learning unit Evaluating an action or assisting the other device to evaluate it;
(d3) the computing device generates, by using the reinforcement learning actor of the reinforcement learning unit, an advantage value to be applied to an output loss of the partition network or to support the other device to generate step; And
(d4) the computing device further trains the partition network by policy gradient update according to the advantage value or supports the other device to further train
/ RTI > The method of claim 1, The method according to claim 1,
The step (b)
The partition network includes:
(i) brain imaging of multiple Alzheimer's patients; And
(ii) at least one of a brain image of a person with a plurality of cognitively normal (CN) and a human brain image with a plurality of mild cognitive impairment (MCI)
Wherein the at least one of the at least two of the at least two of the at least two of the at least two of the at least one of the at least two of the at least one of the at least two of the at least one of the at least two. 3. The method of claim 2,
Wherein the segmentation network performs the segmentation by extracting image features including shape, texture, global location, and structural features of the input brain image. The method according to claim 1,
Wherein the disease prediction network comprises:
(i) trained to reflect the interconnectivity between the brain regions of the brain image resulting from the compartment, (ii) the volume distribution of each brain region of the brain image resulting from the compartment is reflected Characterized by at least one of the following. delete The method according to claim 1,
The compensation value,
And calculating a ranking based on the volume and the classification information by performing an independent t-test on the basis of the volume and the classification information and using the resultant p-value. . The method according to claim 1,
Wherein the error is a Temporal Difference (TD) error. The method according to claim 1,
After the step (c)
(d ') the computing device providing the classification information to an external entity or supporting the other device to provide
Further comprising the steps of: A computer program product, stored in a machine readable non-volatile storage medium, comprising instructions embodied in a computer-readable medium for causing a computing device to perform the method of any one of claims 1 to 4 and 6 to 8, . 1. A computing device for determining a brain disease based on a feature rank for a medical image,
A communication unit for acquiring a brain image of the subject; And
(i) a process for implementing a parcel network for parcellating a plurality of brain regions previously designated as being related to brain diseases from the brain image; (ii) a process of generating a disease prediction network that generates classification information on brain diseases displayed on the brain image based on the result information of the compartment; and (iii) A processor for performing a process of implementing a reinforcement learning module for updating a network,
, ≪ / RTI &
The process (iii)
(iii-1) calculating a reward based on a parcellation output volume for each brain region of the brain image resulting from the segment; (iii-2) evaluating an action of a reinforcement learning actor of the reinforcement learning unit based on an error with respect to an accumulated value of the calculated compensation value; (iii-3) generating an advantage value to be applied to an output loss of the partition network; And (iii-4) further training the compartment network by policy gradient update based on the advantage value.

Images