KR102605720B1 - Brain disease prediction device and method, learning device for predicting brain disease - Google Patents

Abstract

A brain disease prediction device according to an embodiment includes a data collection unit that collects a plurality of brain images at different viewpoints, a data processing unit that extracts information on brain tissue images of a region of interest among the brain images, and a plurality of brain images at different viewpoints. The progression of the brain disease, the current progression rate of the brain disease, the future progression rate of the brain disease, and the brain disease by using the information of a plurality of brain tissue images of the region of interest extracted from the brain image as input to a pre-trained learning model. It may include a diagnosis unit that predicts at least one piece of information among clinical diagnostic evaluation scores according to the condition.

Description

Brain disease prediction device and method, learning device for predicting brain disease

The present invention relates to a deep learning-based prediction device and method for brain diseases, such as Alzheimer's disease.

Alzheimer's disease is a neurodegenerative brain disease that affects many older people and is clinically characterized by loss of cognitive function. As the average lifespan increases due to the development of medicine, Alzheimer's disease is on the rise, and as a result, the need for early diagnosis and technology to predict the outcome of the progression of Alzheimer's disease is increasing for appropriate measures such as slowing down the progression of Alzheimer's disease. there is.

MRI, a non-invasive method among brain imaging methods, has no artificial shading effect from the skull compared to computed tomography (CT), which is the same non-invasive method, allowing for microdiagnosis of cerebral perfusion status and structural improvement of the brain. , various diseases that occur due to functional changes can be tested.

Early detection of Alzheimer's disease can be a critical factor in providing treatment before the patient's cognitive function rapidly declines and daily life becomes difficult.

Since Alzheimer's disease causes symptoms in which the patient's brain atrophies over time, research is being conducted to detect and diagnose the patient's condition from brain imaging. When diagnosing diseases using brain imaging, since the brain structures among patients are all different, in order to uniformly check the degree of common change, a template is used to convert the structure into the same structure and the degree of change in each anatomical region is checked. method can be used.

Clinical examples of Alzheimer's disease diagnosis methods include Mini-Mental State Examination (MMSE), Alzheimer's Disease Assessment Score (ADAS), Clinical Dementia Rating (CDR), and blood tests. A method of comprehensively judging the results can be used.

Recently, with the development of computer image classification technology through deep learning techniques, a type of machine learning, research is being actively conducted to diagnose clinical conditions by extracting features from medical images. However, the existing deep learning-based single brain image-based diagnosis system has difficulty in early diagnosis of Alzheimer's disease because the morphological differences in brain images are also subtle when the disease progression is minimal.

Domestic Public Patent No. 10-2019-0105452 (published on September 17, 2019)

The purpose of embodiments of the present invention is to provide a brain disease prediction device and method for more reliably predicting the progression of brain diseases such as Alzheimer's disease, and a learning device for predicting brain diseases.

However, the technical problems to be achieved by the embodiments of the present invention are not limited to the problems mentioned above, and various technical problems can be derived from the contents described below within the range obvious to those skilled in the art.

A brain disease prediction device according to an embodiment of the present invention includes a data collection unit that collects a plurality of brain images at different times, a data processing unit that extracts information on brain tissue images of a region of interest among the brain images, and Information on a plurality of brain tissue images of the region of interest extracted from a plurality of brain images at a certain point of time is input to a pre-trained learning model to determine the progression of the brain disease, the current progression rate of the brain disease, the future progression rate of the brain disease, and It may include a diagnosis unit that predicts at least one piece of information among clinical diagnostic evaluation scores according to the brain disease state.

The data processing unit includes a brain position alignment unit that processes the plurality of brain images so that the positions of the brains in the plurality of brain images are aligned, and a brain tissue image of the region of interest in the plurality of brain images in which the brain positions are aligned. a brain tissue image extraction unit that extracts, a removal unit that removes a region excluding the brain tissue image of the region of interest from the plurality of brain images, and information on the brain tissue using the brain tissue image of the region of interest. It may include an information extraction unit that extracts.

The brain tissue image of the region of interest in the data processing unit may include at least one of a middle temporal gyrus image, an entorhinal cortex image, a fusiform gyrus, and a hippocampus image. there is.

The learning model may include a plurality of learning models each learned for a normal state, a mild cognitive impairment state, and an Alzheimer's disease state according to information in the brain tissue image.

The learning data for pre-training the learning model in the diagnostic unit includes first learning data including first brain tissue image information acquired at a first time point and a first diagnostic evaluation score for the first brain tissue image information; , second learning data including second brain tissue image information acquired at a second time point and a second diagnostic evaluation score for the second brain tissue image information, and third brain tissue image information acquired at a third time point, and It may include third learning data including a third diagnostic evaluation score for the third brain tissue image information.

The training data for pre-training the learning model in the diagnosis unit may include each of the first training data, the second training data, and the third training data, or a combination of two or more training data.

The data collection unit further collects clinical diagnostic evaluation score information according to the state of the brain tissue, and the diagnostic unit may use the information on the brain tissue image and the clinical diagnostic evaluation score information as learning data for the learning model. there is.

The diagnostic unit may predict information on the brain tissue image at a time earlier than the current time or information on the brain tissue image at a later time than the current time using the learning model.

The diagnostic unit may reuse information on the brain tissue image at a time earlier than the current time or information on the brain tissue image at a later time than the current time and use it as input data for the learning model.

The learning model includes an encoding module unit and a decoding module unit, and the diagnosis unit uses information on the brain tissue image as an input to the encoding module and includes volume change amount, shape change amount, shape change amount, and position change amount information of the brain tissue image. Extract a feature, and use the feature as an input to the decoding module to predict at least one information of the progression of the brain disease, the current progression rate of the brain disease, the future progression rate of the brain disease, and the clinical diagnostic evaluation score. You can.

In addition, the brain disease prediction method of the brain disease prediction device according to an embodiment of the present invention includes collecting data from a plurality of brain images at different viewpoints, and extracting information on a brain tissue image of a region of interest among the brain images. The progression of the brain disease, the current progression rate of the brain disease, and the brain by using the information on the stage and the plurality of brain tissue images of the region of interest extracted from the plurality of brain images at different viewpoints as input to the pre-trained learning model. It may include predicting at least one piece of information among the future progression rate of the disease and a clinical diagnostic evaluation score according to the brain disease state.

In addition, the learning device according to an embodiment of the present invention includes a learning data collection unit that collects learning data, the progress of the brain disease, the current progression rate of the brain disease, the future progression rate of the brain disease, and the state of the brain disease. It may include a learning model that is learned using the learning data to predict at least one piece of information among clinical diagnostic evaluation scores.

The learning model may be any one of RNN, CNN, and DNN, or a combination thereof.

The learning data collection unit collects first learning data including first brain tissue image information acquired at a first time point and a first diagnostic evaluation score for the first brain tissue image information, and second brain tissue image information obtained at a second time point. Second learning data including tissue image information and a second diagnostic evaluation score for the second brain tissue image information, and third brain tissue image information acquired at a third time point and a second diagnostic evaluation score for the third brain tissue image information 3 It may include third learning data including diagnostic evaluation scores.

The learning data collection unit may collect data of each of the first learning data, the second learning data, and the third learning data, or a combination of two or more learning data.

Embodiments of the present invention have the effect of diagnosing Alzheimer's disease more reliably using past brain imaging data and providing meaningful information on the analysis and basis of the diagnosis.

An embodiment of the present invention is a diagnostic method that can help clinicians' decision-making. In addition to magnetic resonance imaging, a reliable diagnosis is made by combining the probability of progression of Alzheimer's disease and the prediction of the clinical diagnostic evaluation score through clinical diagnostic evaluation scores. This has a possible effect.

Figure 1 is a block diagram showing the configuration of a brain disease device according to an embodiment of the present invention.
Figure 2 is a block diagram showing information collected in the data collection unit of the brain disease device according to an embodiment of the present invention.
Figure 3 is a diagram showing brain images collected in the data collection unit of the brain disease device according to an embodiment of the present invention.
Figure 4 is a diagram showing diagnostic evaluation score information collected from the data collection unit of the brain disease device according to an embodiment of the present invention.
Figure 5 is a block diagram showing the configuration of a data processing unit of a brain disease device according to an embodiment of the present invention.
Figure 6 is a diagram showing brain tissue extracted from the data processing unit of the brain disease device according to an embodiment of the present invention.
Figure 7 is a diagram showing a learning model of a brain disease device according to an embodiment of the present invention.
Figure 8 is a diagram showing a plurality of learning models of a brain disease device according to an embodiment of the present invention.
Figure 9 is a diagram showing mathematical modeling of a learning model of a brain disease device according to an embodiment of the present invention.
Figure 10 is a diagram showing the performance results of a brain disease device according to an embodiment of the present invention.
Figure 11 is a block diagram showing a brain disease method according to an embodiment of the present invention.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and can be implemented in various forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to those skilled in the art to which the present invention pertains. It is provided to fully inform the person of the scope of the invention, and the scope of the invention is only defined by the claims.

In describing the embodiments of the present invention, detailed descriptions of well-known functions or configurations will be omitted except when actually necessary. The terms described below are terms defined in consideration of functions in the embodiments of the present invention, and may vary depending on the intention or custom of the user or operator. Therefore, the definition should be made based on the contents throughout this specification.

The functional blocks shown in the drawings and described below are only examples of possible implementations. Other functional blocks may be used in other implementations without departing from the spirit and scope of the detailed description. Additionally, although one or more functional blocks of the present invention are shown as individual blocks, one or more of the functional blocks of the present invention may be a combination of various hardware and software components that perform the same function.

Additionally, the expression that it includes certain components is an open expression and simply refers to the existence of the components, and should not be understood as excluding additional components.

Furthermore, when it is mentioned that a component is connected or connected to another component, it should be understood that although it may be directly connected or connected to the other component, other components may exist in between.

In addition, expressions such as 'first, second', etc. are expressions used only to distinguish a plurality of components, and do not limit the order or other characteristics between the components.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Figure 1 is a block diagram showing the configuration of a brain disease device according to an embodiment of the present invention, Figure 2 is a block diagram showing information collected in the data collection unit of the brain disease device according to an embodiment of the present invention, and Figure 3 is a diagram showing brain images collected from the data collection unit of the brain disease device according to an embodiment of the present invention, and Figure 4 shows diagnostic evaluation score information collected from the data collection unit of the brain disease device according to an embodiment of the present invention. Figure 5 is a block diagram showing the configuration of the data processing unit of the brain disease device according to an embodiment of the present invention, and Figure 6 shows brain tissue extracted from the data processing unit of the brain disease device according to an embodiment of the present invention. 7 is a diagram showing a learning model of a brain disease device according to an embodiment of the present invention, FIG. 8 is a diagram showing a plurality of learning models of a brain disease device according to an embodiment of the present invention, and FIG. 9 is a diagram showing the mathematical modeling of the learning model of the brain disease device according to an embodiment of the present invention, and Figure 10 is a diagram showing the performance results of the brain disease device according to an embodiment of the present invention.

Referring to FIG. 1 , the brain disease prediction apparatus 1000 according to an embodiment of the present invention may include a data collection unit 100.

The data collection unit 100 may collect brain images taken at different viewpoints for the same subject. As shown in FIG. 2, the data collection unit 100 may include brain image information 110 and diagnostic evaluation score information 120. The data collection unit 100 may only collect brain image information 110.

As shown in FIG. 3, the brain image may be an image taken from MRI. The brain image may include a brain image 10 taken at a first viewpoint, a brain image 20 taken at a second viewpoint, and a brain image 30 taken at a third viewpoint, but the number is limited. It doesn't work. The first point in time may be one year ago, the second point in time may be two years ago, and the third point in time may be three years ago. Here, a brain image at the current time may be further included.

As shown in FIG. 4, diagnostic evaluation score information 120 may be a diagnostic evaluation score for the corresponding brain image. Here, the diagnostic evaluation score may include pMCI-early, pMCI-late, and sMCI, and may further include MMSE, ADAS-cog 11, and ADAS-cog13, but is not limited thereto.

Returning to FIG. 1, the brain disease prediction apparatus 1000 according to the embodiment may further include a data processing unit 200.

The data processing unit 200 may perform image processing to obtain brain volume information through a matching process and a brain tissue segmentation process. The data processing unit 200 divides the anatomical region from the brain volume information, obtains voxel values normalized to the entire brain volume for each region of interest, and then extracts a sample by obtaining a random sample set. .

The data processing unit 200 adjusts the symmetry of the input image and then removes brain regions that are not required for analysis. In order to compensate for errors in the obtained brain images, the brain tissue is separated after shift correction. After converting all brain tissues into the same template space, a density map of the brain is obtained through a registration process. The obtained brain density map is divided into regions of interest for diagnosis by anatomical region, and then normalized to match the distribution of data values.

For example, when a magnetic resonance image is input, the data processing unit 200 aligns the brain location, removes the skull and cerebellum, separates the brain tissue, and registers them to separate a region of interest with brain volume information. The preprocessor extracts samples by extracting volume information using a random sampling technique from the normalized value of brain volume information separated by region of interest to the total brain volume.

Here, the data processing unit 200 can separate a region of interest with brain volume information using the distribution of brightness levels of the brain template information. For example, a region of interest having brain volume information can be separated through the distribution of brightness based on template information such as location information of brain tissue and brightness information for each brain tissue.

More specifically, as shown in FIG. 5, the data processing unit 200 includes a brain position alignment unit 210, a brain tissue image extraction unit 220, a removal unit 230, and an information extraction unit 240. It can be included.

The brain position alignment unit 210 may process the brain image so that the brain position in the brain image is aligned. The brain position alignment unit 210 may align the brain position through a brain image registration process. The brain position alignment unit 210 can align the brain position by matching the symmetry of the brain image.

The brain tissue image extraction unit 220 may extract a brain tissue image of a region of interest from a brain image. Brain tissue may be a biomarker related to Alzheimer's disease. According to a recent study, in relation to Alzheimer's disease, an increase in Amyloid β and abnormal tau levels was observed in neocortical regions, the middle temporal gyrus tissue (42) of the brain (40) in Figure 6A, and the inner olfactory bulb in Figure 6B. A decrease in volumes was observed in the entorhinal cortex tissue (44), fusiform gyrus tissue (46), and hippocampus tissue (48) in Figure 6c.

Therefore, the brain tissue image extraction unit 220 may extract one or more of the middle temporal gyrus tissue image, entorhinal cortex tissue image, fusiform gyrus tissue image, and hippocampus tissue image. can be extracted.

The removal unit 230 may remove a region of the brain image excluding the brain tissue image of the region of interest. That is, the removal unit 230 can remove brain tissue images such as the skull and cerebellum.

The information extraction unit 240 may extract information about brain tissue using a brain tissue image of the region of interest. Here, the information may include various information about brain tissue, and may include volume information as an example. This makes it possible to obtain images of the brain tissue of the specimen.

Returning to FIG. 1 , the brain disease prediction apparatus 1000 according to the embodiment may further include a diagnostic unit 400.

The diagnostic unit 400 uses the learning model 300, which has been learned in advance by inputting information from brain tissue images, to determine the progression of the brain disease, the current progression rate of the brain disease, the future progression rate of the brain disease, and clinical clinical information. At least one piece of information among the diagnostic evaluation scores can be predicted. The diagnosis unit 400 can diagnose Alzheimer's disease using this information.

The learning model 300 may be placed inside the diagnosis unit 400 or may be placed in a space separate from the diagnosis unit 400.

The learning model 300 is trained by learning data and labels using one or a combination of neural networks such as Recurrent Neural Network (RNN), Convolution Neural Network (CNN), and Deep Neural Network (DNN), and then performs a diagnostic unit. It can be used by (400).

The learning model 300 may include a learning data collection unit that collects learning data, but is not limited to this. The learning model 300 may be a separate learning device that includes a learning data collection unit.

As shown in FIG. 7, the learning model 300 may include an encoding module 310 and a decoding module 320.

The encoding module 310 may extract features including information on the brain tissue image by using information on the brain tissue image as input. Here, features such as volume change, shape change, shape change, and position change of brain tissue can be extracted. Here, the brain tissue image information uses brain tissue images acquired over time, so the amount of change can be measured.

The learning model 300 may be learned using brain images of a predetermined period of the same subject and clinical diagnostic evaluation scores obtained from the same subject on the same day when each brain image was acquired.

For example, the learning model 300 may include first learning data including a first brain tissue image acquired at a first time point and a first diagnostic evaluation score for the first brain tissue image, and Second learning data including a second brain tissue image and a second diagnostic evaluation score for the second brain tissue image, and a third brain tissue image acquired at a third time point and a third diagnostic evaluation score for the third brain tissue image. It can be learned using third learning data including diagnostic evaluation scores. Here, a plurality of learning models 300 may be provided and learned using respective learning data.

In some cases, in order to improve the reliability or accuracy of the learning model 300, it is based on the fact that the first learning data at the first time point, the second learning data at the second time point, and the third learning data at the third time point are each time series data. Thus, the value obtained from the third learning data can be learned by the second learning model as the second learning data, and the value obtained by the second learning data can be learned by the first learning model as the first learning data. there is.

Additionally, as shown in FIG. 8, the learning model 300 may include a plurality of learning models. The learning model 300 may include a first learning model 300a, a second learning model 300b, and a third learning model 300c.

The first learning model 300a may be learned using information from brain tissue images in a normal state. The second learning model 300b may be learned using information on mild cognitive impairment, which is an early onset state of brain disease. The third learning model 300c can be learned using information on Alzheimer's disease, which is a brain disease invention. Of course, the three data can be used in combination to improve learning accuracy.

는 시간차 정보가 반영된 값을 의미하고, Δ_t,Δ_t+1,Δ_t+2는 시간차 정보를 의미하고, Mt, Mt+1, Mt+2는 마스킹 벡터를 의미하고, Z_t는 변수 간 상관관계 정보가 반영된 값을 의미하고,

는 시간차 정보 및 변수 간 상관관계 정보가 반영된 값을 의미하고,

는 행렬곱 연산자를 의미하고,

는 연결 연산자를 의미할 수 있다.As shown in FIG. 9, the learning model 300 can be mathematically modeled. Here, C ₀ , C ₁ , C ₂ mean the cell state, ht-1, ht, ht+1 mean the hidden state, Xt means the input data,

means the value reflecting the time difference information, Δ _t ,Δ _t+1 ,Δ _t+2 means the time difference information, Mt, Mt+1, Mt+2 mean the masking vector, and Z _t is the It refers to a value that reflects correlation information,

means the value reflecting time difference information and correlation information between variables,

means the matrix multiplication operator,

may mean a connection operator.

Here, each value can be calculated by Equation 1.

[Equation 1]

Here, (1) process is an equation that reflects time difference information, (2) process is an equation that performs a matrix multiplication operation on the time difference information and masking vector on the input data, and (3) process is an equation that reflects correlation information between variables. The (5) process combines the time difference information and the correlation information between variables, and the (6) process uses the input value through the (5) process if there is a missing value, and if there is no missing value, the input value is used. It can be calculated using data.

Returning to FIG. 1, when using the patient's current brain tissue image and clinical diagnostic evaluation score information as input data, the diagnostic unit 400 uses the current brain tissue image and clinical diagnostic evaluation score information as input data. Thus, by using the learning model 300, the brain image information at a time before the current time or the brain image information at a time after the current time can be predicted.

In addition, when the diagnosis unit 400 uses the learning model 300 with brain image information from a time before the current time or the brain image information from a time after the current time as the input data, the progression of the brain disease, It is possible to predict at least one of the current progression rate of the brain disease, future progression rate of the brain disease, and clinical diagnostic evaluation score.

In addition, the diagnostic unit 400 can simultaneously predict the progression stage of the brain disease and the clinical diagnostic evaluation score from each time feature obtained from the features extracted by the learning model 300 or predict it regardless of the order. In order to simultaneously predict the disease progression stage and clinical diagnostic evaluation score from each time characteristic, a final diagnosis method is performed through the diagnosis unit 400, and a method of discriminating between classes and a prediction model suitable for the clinical diagnostic evaluation score are created. You can use a deep learning model to configure it.

As described above, if the clinical diagnostic evaluation score is not input by the learning model 300, the diagnostic unit 400 performs the clinical diagnostic evaluation using the features extracted based on the brain image by the learning model 300. The score can be predicted. In this case, the diagnosis unit 400 may diagnose the brain disease by integrating the probability of each time progression stage and the predicted clinical diagnostic evaluation score.

As shown in Figure 10, the data used for performance results used the ADNI-based TADPOLE challenge dataset, and 655 subjects were used. Of these, 395 were used as learning data, 131 as validation data, and 129 as test data.

Performance results can be composed of CN (Cognitively Normal), MCI (Mild Cognitive Impairment), and AD (Alzheimer's Disease) labels, and performance evaluation was evaluated using 5-fold cross validation.

In the case of classification, the mAUC (multi-class AUC) indicator was used, and in the case of regression, the mean absolute error (MAE) can be used as an indicator to estimate the volume value, and the prediction value of the cognitive/clinical evaluation score In this case, MAE and RMSE (root mean square error) were used.

Calculation of mAUC can be calculated by Equation 2.

[Equation 2]

The calculation of MAE and RMSE can be calculated by Equation 3.

[Equation 3]

As shown in Figure 10, in the case of mAUC, when clinical diagnostic evaluation scores are used (with Clinical scores) and when not used (w/o Clinical scores) are compared, and clinical diagnostic evaluation scores are used. It can be seen that the performance (with clinical scores) is further improved. Therefore, it can be confirmed that using clinical diagnostic evaluation scores contributes to improving classification performance.

In addition, as shown in Figure 10, as a prediction result for the brain tissue image information and the clinical diagnostic evaluation score, the brain tissue image information at the next time point estimated from the brain tissue image information at the current time point and the brain tissue at the actual next time point It represents the value calculated by image information and MAE corresponding to Equation 3. Additionally, in the case of ADAS11 and ADAS13, which correspond to clinical diagnostic evaluation scores, the values calculated by MAE and RMSE corresponding to Equation 3 are shown. Therefore, it can be seen that the brain disease prediction in the example is significantly improved compared to the prior art.

Hereinafter, a brain disease method according to an embodiment will be described with reference to FIG. 11.

Figure 11 is a block diagram showing a brain disease method according to an embodiment. Here, the brain disease method according to the embodiment may be performed by the brain disease device described above.

As shown in FIG. 11, the brain disease method according to the embodiment includes collecting a plurality of brain image data at different time points (S100) and extracting information on the brain tissue image of the region of interest among the brain images. (S200), the progression of the brain disease, the current progression rate of the brain disease, by using the information of the plurality of brain tissue images of the region of interest extracted from the plurality of brain images at different viewpoints as input to a pre-trained learning model, It may include a step (S300) of predicting at least one piece of information among the future progression rate of the brain disease and a clinical diagnostic evaluation score according to the state of the brain disease.

The step (S100) of collecting a plurality of brain image data at different time points may be performed in a data collection unit.

The step (S200) of extracting information on a brain tissue image of a region of interest among brain images may be performed in a data processing unit.

The progression of the brain disease, the current progression rate of the brain disease, and the future of the brain disease are determined by using the information of the plurality of brain tissue images of the region of interest extracted from the plurality of brain images at different viewpoints as input to a pre-trained learning model. The step (S300) of predicting at least one piece of information among the progression rate and the clinical diagnostic evaluation score according to the brain disease state may be performed in the diagnosis unit.

Embodiments of the present invention described above can be implemented through various means. For example, embodiments of the present invention may be implemented by hardware, firmware, software, or a combination thereof.

In the case of implementation by hardware, the method according to embodiments of the present invention includes one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), and digital signal processing elements ( It can be implemented by DSPDs (Digital Signal Processing Devices), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, etc.

In the case of implementation by firmware or software, the method according to embodiments of the present invention may be implemented in the form of a module, procedure, or function that performs the functions or operations described above. A computer program in which software codes, etc. are recorded may be stored in a computer-readable recording medium or memory unit and driven by a processor. The memory unit is located inside or outside the processor and can exchange data with the processor through various known means.

Additionally, the present invention may be expressed as a block diagram including a plurality of blocks or a flow diagram including a plurality of steps, and combinations of each block of the block diagram and each step of the flow diagram may be performed by computer program instructions. Since these computer program instructions can be mounted on the encoding processor of a general-purpose computer, special-purpose computer, or other programmable data processing equipment, the instructions performed through the encoding processor of the computer or other programmable data processing equipment are included in each block or block of the block diagram. Each step of the flowchart creates a means to perform the functions described. These computer program instructions may also be stored in computer-usable or computer-readable memory that can be directed to a computer or other programmable data processing equipment to implement a function in a particular way, so that the computer-usable or computer-readable memory The instructions stored in can also produce manufactured items containing instruction means that perform the functions described in each block of the block diagram or each step of the flow diagram. Computer program instructions can also be mounted on a computer or other programmable data processing equipment, so that a series of operational steps are performed on the computer or other programmable data processing equipment to create a process that is executed by the computer, thereby generating a process that is executed by the computer or other programmable data processing equipment. Instructions that perform processing equipment may also provide steps for executing functions described in each block of the block diagram and each step of the flow diagram.

In addition, each block or each step may represent a module, segment, or portion of code that includes one or more executable instructions for executing a specified logical function. It should also be noted that in some alternative embodiments it is possible for the functions mentioned in blocks or steps to occur out of order. For example, two blocks or steps shown in succession may in fact be performed substantially simultaneously, or the blocks or steps may sometimes be performed in reverse order depending on the corresponding function.

As such, a person skilled in the art to which the present invention pertains will understand that the present invention can be implemented in other specific forms without changing its technical idea or essential features. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. The scope of the present invention is indicated by the claims described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. .

Claims (15)

A data collection unit that collects multiple brain images at different time points;
a data processing unit that extracts information on a brain tissue image of a region of interest among the brain images; and
The progression of the brain disease, the current progression rate of the brain disease, and the future of the brain disease are determined by using the information of the plurality of brain tissue images of the region of interest extracted from the plurality of brain images at different viewpoints as input to a pre-trained learning model. It includes a diagnostic unit that predicts at least one information of the progression rate and the clinical diagnostic evaluation score according to the brain disease state,
The data collection unit further collects clinical diagnostic evaluation score information according to the state of the brain tissue,
The diagnostic unit uses information on the brain tissue image at the current time among the different time points and the clinical diagnostic evaluation score information at the current time as learning data for the learning model,
A brain disease prediction device wherein the diagnostic unit predicts information on the brain tissue image at a time before the current time or information on the brain tissue image at a later time than the current time using the learning model. According to paragraph 1,
The data processing unit,
a brain position alignment unit that processes the plurality of brain images so that the positions of the brains in the plurality of brain images are aligned;
a brain tissue image extraction unit that extracts a brain tissue image of the region of interest from the plurality of brain images in which the brain locations are aligned;
a removal unit that removes a region excluding the brain tissue image of the region of interest from the plurality of brain images;
An information extraction unit that extracts information about the brain tissue using the brain tissue image of the region of interest. According to paragraph 2,
In the data processing unit,
A brain disease prediction device wherein the brain tissue image of the region of interest includes at least one of a middle temporal gyrus image, entorhinal cortex image, fusiform gyrus, and hippocampus image. According to paragraph 1,
The learning model is a brain disease prediction device including a plurality of learning models each learned for a normal state, a mild cognitive impairment state, and an Alzheimer's disease state according to information on the brain tissue image. According to paragraph 1,
In the diagnostic department,
The learning data for pre-training the learning model includes first learning data including first brain tissue image information acquired at a first time point and a first diagnostic evaluation score for the first brain tissue image information, and a second time point. Second learning data including second brain tissue image information acquired from and a second diagnostic evaluation score for the second brain tissue image information, third brain tissue image information acquired at a third time point, and the third brain A brain disease prediction device including third learning data including a third diagnostic evaluation score for tissue image information. According to clause 5,
In the diagnostic department,
The training data for pre-training the learning model includes each of the first training data, the second training data, and the third training data, or a combination of two or more training data. According to paragraph 1,
A brain disease prediction device wherein the clinical diagnostic evaluation score information includes diagnostic evaluation scores for a plurality of brain images at different time points. delete According to paragraph 1,
The diagnostic department,
A brain disease prediction device that reuses information on the brain tissue image at a time earlier than the current time or information on the brain tissue image at a later time than the current time and uses it as input data for the learning model. According to paragraph 1,
The learning model is,
Includes an encoding module unit and a decoding module unit,
The diagnostic department,
Using the information of the brain tissue image as input to the encoding module, features including volume change amount, shape change amount, shape change amount, and position change amount information of the brain tissue image are extracted,
A brain disease prediction device that uses the characteristic as an input to the decoding module unit to predict at least one information of the progression of the brain disease, the current progression rate of the brain disease, the future progression rate of the brain disease, and the clinical diagnostic evaluation score. In the brain disease prediction method of the brain disease prediction device,
Collecting data from a plurality of brain images at different time points;
extracting information on a brain tissue image of a region of interest from the brain image; and
The progression of the brain disease, the current progression rate of the brain disease, and the future of the brain disease are determined by using the information of the plurality of brain tissue images of the region of interest extracted from the plurality of brain images at different viewpoints as input to a pre-trained learning model. Predicting at least one information of progression rate and clinical diagnostic evaluation score according to the brain disease state,
The learning model uses information on the brain tissue image at the current time among the different time points and the clinical diagnostic evaluation score information at the current time as learning data for the learning model,
The predicting step uses the learning model to predict information on the brain tissue image at a time earlier than the current time or information on the brain tissue image at a later time than the current time. A learning data collection unit that collects learning data; and
Learned using the learning data to predict at least one information of the progression of the brain disease, the current progression rate of the brain disease, the future progression rate of the brain disease, and the clinical diagnostic evaluation score according to the brain disease state. including a learning model;
The learning model uses information on the brain tissue image at the current time among different time points and the clinical diagnostic evaluation score information at the current time as learning data for the learning model,
The predicting step is a learning device for predicting a brain disease in which information on the brain tissue image at a time before the current time or information on the brain tissue image at a time after the current time is predicted using the learning model. According to clause 12,
A learning device for predicting brain disease in which the learning model is any one of RNN, CNN, and DNN or a combination thereof. According to clause 12,
The learning data collection unit first learning data including first brain tissue image information obtained at a first time point and a first diagnostic evaluation score for the first brain tissue image information, and a second time point Second learning data including second brain tissue image information acquired from and a second diagnostic evaluation score for the second brain tissue image information, third brain tissue image information acquired at a third time point, and the third brain A learning device for predicting a brain disease including third learning data including a third diagnostic evaluation score for tissue image information. According to clause 14,
The learning data collection unit is a learning device for predicting a brain disease wherein the first learning data, the second learning data, and the third learning data are collected separately or a combination of two or more learning data.

Images