KR102250954B1 - Apparatus and method for predicting dementia by dividing brain mri by brain region - Google Patents

Abstract

An apparatus and method for predicting dementia through segmentation of brain MRI images by brain regions are provided. The method for predicting dementia through segmentation of brain regions includes: obtaining a brain MRI image from a subject; Dividing the acquired brain MRI image into a plurality of regions; Measuring the capacity of the brain for each divided area; And predicting the likelihood of dementia based on the deep learning model, based on the measured brain capacity. Including, the step of dividing the brain MRI image into a plurality of regions may further include dividing the brain MRI image in pixel units using semantic segmentation.

Description

A device and method for predicting dementia through segmentation of brain MRI images by brain region {APPARATUS AND METHOD FOR PREDICTING DEMENTIA BY DIVIDING BRAIN MRI BY BRAIN REGION}

The present invention relates to an apparatus and method for predicting dementia through segmentation of brain MRI images by brain region, and more specifically, dividing the brain into a plurality of regions using semantic segmentation, and quantifying brain capacity for each region. The present invention relates to an apparatus and method for predicting dementia through segmentation of brain MRI images for predicting the probability of dementia.

Although research for the treatment of Dementia has been conducted worldwide for over 20 years, it is a symptom that has yet to be fully cured.

Dementia, one of the senile diseases, has increased rapidly with the global increase in the elderly population. In the United States, between 1999-2000 and 2005-2006, the mortality rate from Alzheimer's dementia doubled, and Korea's aging rate is 1.5 times faster than Japan and 5 times faster than France.

According to a prevalence survey by the Ministry of Health and Welfare, the number of dementia patients in Korea is 540,000 in 2012, and is expected to increase rapidly to 840,000 by 2020. Korea's dementia population is growing the fastest in the world, and the social cost for it is projected to reach 1.5% of GDP by 2050.

Various drugs can be used to treat dementia, which is expected to increase rapidly in patients, but these drugs are not the underlying treatment of dementia, but only have the effect of slowing the progression. However, it has a relatively high effect when it is prescribed and treated in the early stages of dementia.

Therefore, early prediction and early diagnosis of dementia can play a decisive role in relieving dementia symptoms. However, in the current method of examining dementia through MRI images, it is difficult to accurately and quickly examine dementia because it predicts dementia based on the experience of a doctor without using a quantitative method.

Unexamined Patent Publication No. 10-2018-0002234, 2018.01.08

The problem to be solved by the present invention is through segmentation of brain MRI images that can predict the probability of dementia for each brain region by measuring the capacity for each brain region and applying the measured brain capacity for each region to a deep learning model. It is to provide an apparatus and method for predicting dementia.

The problems to be solved by the present invention are not limited to the problems mentioned above, and other problems that are not mentioned will be clearly understood by those skilled in the art from the following description.

A method for predicting dementia for each brain region of a brain MRI image according to an embodiment of the present invention for solving the above-described problems includes: acquiring a brain MRI image from a subject; Dividing the acquired brain MRI image into a plurality of regions; Measuring the capacity of the brain for each divided area; And predicting the likelihood of dementia based on the deep learning model, based on the measured brain capacity. Including, the step of dividing the brain MRI image into a plurality of regions may further include dividing the brain MRI image in pixel units using semantic segmentation.

Here, the deep learning model is a model in which feature learning on brain capacity according to age is completed using an effective data set.

Here, semantic segmentation is characterized in that the brain MRI image is processed and segmented into 2D data.

In addition, the step of dividing the brain MRI image into a plurality of regions may include recognizing an object in the brain MRI image and dividing the region based on the recognized object; Including, the step of measuring the capacity of the brain for each region, calculating the number of pixels of the object existing in each region; includes.

In addition, the step of predicting the likelihood of dementia includes: predicting a time to develop dementia based on a deep learning model.

According to another embodiment of the present invention for solving the above-described problem, an apparatus for predicting dementia for each brain region of a brain MRI image includes: a learning unit that learns characteristics of brain capacity according to age using an effective data set; A measurement unit that divides the brain MRI image of the subject into a plurality of areas and measures brain capacity for each of the divided areas; A prediction unit that predicts the likelihood of dementia and a time to develop dementia based on the deep learning model, based on the measured brain capacity; And a communication unit for receiving an MRI image of the subject's brain from the user terminal, and providing a predicted possibility of dementia and a time to suffer from dementia to the user terminal. Including, the brain MRI image is processed into 2D data using semantic segmentation, characterized in that it is segmented in units of pixels.

The learning unit divides the valid data set into a training data set and a validation data set at a preset ratio, learns the training data set to generate a deep learning model, and verifies the deep learning model generated from the validation data set.

The measurement unit may recognize an object in the brain MRI image, divide a region based on the recognized object, and calculate the number of pixels of the object existing in each region. In addition, the measurement unit may divide the brain 　MRI 　 image of the subject for each region in which the object exists, and label each region so that it is mapped with the object.

The dementia prediction program for each brain region of a brain MRI image according to another embodiment of the present invention for solving the above-described problem is combined with hardware to execute the aforementioned method for predicting dementia for each brain region of the brain MRI image. Is stored in.

Other specific details of the present invention are included in the detailed description and drawings.

According to the present invention, it is possible to predict the probability of developing dementia for each brain region by measuring the capacity for each brain region and applying the measured brain capacity for each region to a deep learning model.

In addition, there is an effect of being able to predict when a subject may develop dementia by applying the measured brain capacity for each region to a deep learning model.

The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the following description.

1 is a diagram illustrating an apparatus for predicting dementia for each brain region according to an exemplary embodiment of the present invention.
2 is a flowchart illustrating a method for predicting dementia for each brain region according to an embodiment of the present invention.

Advantages and features of the present invention, and a method of achieving them will become apparent with reference to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in a variety of different forms. It is provided to fully inform the skilled person of the scope of the present invention, and the present invention is only defined by the scope of the claims.

The terms used in the present specification are for describing exemplary embodiments and are not intended to limit the present invention. In this specification, the singular form also includes the plural form unless specifically stated in the phrase. As used herein, “comprises” and/or “comprising” do not exclude the presence or addition of one or more other elements other than the mentioned elements. Throughout the specification, the same reference numerals refer to the same elements, and "and/or" includes each and all combinations of one or more of the mentioned elements. Although "first", "second", and the like are used to describe various elements, it goes without saying that these elements are not limited by these terms. These terms are only used to distinguish one component from another component. Therefore, it goes without saying that the first component mentioned below may be the second component within the technical idea of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used in the present specification may be used with meanings that can be commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, terms defined in a commonly used dictionary are not interpreted ideally or excessively unless explicitly defined specifically.

Spatially relative terms "below", "beneath", "lower", "above", "upper", etc. It can be used to easily describe the correlation between a component and other components. Spatially relative terms should be understood as terms including different directions of components during use or operation in addition to the directions shown in the drawings. For example, if a component shown in a drawing is turned over, a component described as "below" or "beneath" of another component will be placed "above" the other component. I can. Accordingly, the exemplary term “below” may include both directions below and above. Components may be oriented in other directions, and thus spatially relative terms may be interpreted according to the orientation.

In this specification,'deep learning' refers to a technology used to cluster or classify objects or data, and is a technology that inputs a lot of data into a computer and classifies similar ones. For example, if a picture similar to a stored dog picture is input, the computer can classify it as a dog picture.

The key to deep learning is prediction through classification. Computers divide data just as humans divide objects by discovering patterns in a lot of data. This discernment method is divided into'supervised learning' and'unsupervised learning'. Supervised learning is a method of teaching information to computers first. For example, you give a picture and tell it "this picture is a cat", which allows the computer to classify pictures of cats based on pre-learned results. On the other hand, unsupervised learning has no prior learning process. Without the process of learning that "this picture is a cat", the computer learns by itself that "this picture is a cat."

Deep learning is a field of machine learning, which includes research to express data in a form that can be processed by a computer, such as vectors or graphs, and to build a model that learns them. For specific learning goals, such as recognizing faces or facial expressions, deep learning focuses on building efficient models and better expressive methods for learning. Many of the expression methods of deep learning are inspired by neuroscience and are based on information processing or communication patterns in the nervous system. In other words, deep learning is a subset of machine learning that takes the role of raising the intelligence of computers to the next level, and implements a deep neural network using a large amount of data and computing technology. The basic principle of the deep neural network is to mimic the connectivity of the human and brain to classify the data and to find the correlation between the data.

A deep neural network is a concept used as a core model of deep learning and means an artificial neural network (ANN) that includes multiple hidden layers between an input layer and an output layer. . Deep neural networks can learn various nonlinear relationships including multiple hidden layers. However, problems such as overfitting and vanishing gradient problems in which errors increase with respect to actual data due to a large amount of computation for learning and over-learning may occur.

Deep neural networks include a deep trust neural network (DBN: Deep Belief Network) based on an unsupervised learning method, a deep autoencoder, etc., depending on the algorithm, and are synthesized for processing 2D data such as images. There are convolutional neural networks (CNNs) and recurrent neural networks (RNNs) for processing time series data.

In this specification, the'valid data set' is a set of data used for deep learning, and is divided into a training data set and a validation data set. The training data set is a set of data for establishing a learning model, and the validation data set is a set of data for verifying the performance of the established learning model.

In the present specification, the'brain MRI image' is a diagnostic method for examining abnormalities in brain tissue and blood vessels in 2D or 3D using a superconducting magnet, high frequency, and a computer. It has the advantage of good resolution and no exposure to X-rays.

In the present specification,'semantic segmentation' is an object recognition method in an image, and it is possible to classify a pixel region of an image and evaluate and visualize the segmentation result by using deep learning. In the present specification, the brain MRI image is divided into each area of the brain, that is, lateral ventricles, amygdala, hippocampus, and the like in pixel units using semantic segmentation.

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

1 is a diagram illustrating an apparatus for predicting dementia for each brain region according to an embodiment of the present invention.

As shown in FIG. 1, the device 100 for predicting dementia for each brain region (hereinafter, referred to as a device for predicting dementia) is connected to a user terminal through a network.

Here, the user terminal is a computer, a tablet PC, a personal digital assistant (PDA), a smartphone, etc. of the examinee or the subject, and transmits an MRI image of the subject's brain to the dementia prediction device 100 to request dementia prediction, and to predict dementia. Receives result information including the probability and timing of dementia from the device 100.

In addition, the network may include not only a data communication network including a local information communication network (LAN), a metropolitan area communication network (MAN), a wide area communication network (WAN), and the Internet, but also a broadcasting network, a telephone network, etc. Communication method can be used.

Referring to FIG. 1, the apparatus 100 for predicting dementia is illustrated as including a learning unit 120, a measurement unit 140, a prediction unit 160, and a communication unit 180, but is not limited thereto.

The learning unit 120 learns characteristics of brain capacity according to age by using the effective data set. Here, the effective data set is a plurality of brain MRI images for use in learning, and each brain MRI image is divided into a plurality of regions.

According to an embodiment, the learning unit 120 divides a valid data set into a training data set and a validation data set at a predetermined ratio (eg, 8 to 2), and training data The set is trained to generate a deep learning model, and the deep learning model created with the validation data set is verified. Accordingly, the learning unit 120 may generate a deep learning model with high accuracy.

In this case, the deep learning model is a model in which features of brain capacity according to age are learned, and is a model created by learning features of capacity, volume, and thickness of brain cortex by brain regions according to age.

The measurement unit 140 divides the brain MRI image of the subject into a plurality of regions, and measures brain capacity for each divided region. In this case, the brain MRI image is processed into 2D data using semantic segmentation and is divided into pixels. Semantic segmentation is one of object recognition techniques, and classifies the existence of an object or the type of object in units of pixels.

That is, the measurement unit 140 recognizes an object in the brain MRI image of the subject, divides an area based on the recognized object, calculates the number of pixels of the object existing in each area, and measures the capacity of the brain for each area. do..

According to an embodiment, the measurement unit 140 may divide a brain MRI image of a subject for each area in which an object exists, and label each area to be mapped with the object. For example, the measurement unit 140 may divide the brain MRI image of the subject into six regions and label each region as Right Lateral Ventricle, Left Lateral Ventricle, Right Amygdala, Left Amygdala, Right Hippocampus, and Left Hippocampus. .

According to an embodiment, the measurement unit 140 may measure a brain capacity, a volume, a brain cortex thickness, etc. for each divided brain region.

The prediction unit 160 predicts the likelihood of dementia and the time to suffer from dementia based on the deep learning model, based on the measured brain capacity.

According to an embodiment, the prediction unit 160 applies the brain capacity for each region of the subject measured by the measurement unit 140 to the deep learning model generated by the learning unit 120 to determine the characteristics of the brain capacity for each region of the subject. Extract. That is, when the brain capacity of each area of the subject is different from the brain capacity characteristics appropriate for the age group extracted by learning, for example, the predictor 160 assumes that the age of the subject is 62 years old, 60 learned through deep learning. If it is determined that the brain capacity of the subject is smaller than the brain capacity corresponding to the generation, the prediction unit 160 may determine that the subject's brain is more constricted than his age, and may predict dementia.

According to an exemplary embodiment, the prediction unit 160 may determine the degree of brain contraction of the subject based on the deep learning model to predict how much the probability of developing dementia is high and whether the patient will develop dementia in the next few years.

The communication unit 180 is a communication means for interconnecting the dementia prediction apparatus 100 to transmit and receive data to and from a user terminal through a network, and includes wireless communication modules such as mobile communication and satellite communication, wired communication modules such as the Internet, and Wi-Fi. It may include a short-range wireless communication module, such as.

According to an embodiment, the communication unit 180 receives a brain MRI image of a subject from a user terminal, and provides a predicted possibility of dementia and a time to suffer from dementia to the user terminal.

2 is a flowchart illustrating a method for predicting dementia according to an embodiment of the present invention.

The communication unit 180 acquires a brain MRI image of the subject from the user terminal (S10), and the measurement unit 140 divides the acquired brain MRI image into a plurality of regions (S20).

Brain MRI image is a diagnostic method for examining abnormalities in brain tissues and blood vessels in 2D or 3D using a superconducting magnet, high frequency, and a computer. The brain MRI image according to an embodiment of the present invention is semantic segmentation. It is processed as 2D data and divided into pixels.

According to an embodiment, a brain MRI image is divided into a plurality of regions in which an object is recognized, and each region is labeled to be mapped to an object within the region.

Thereafter, the measurement unit 140 measures the capacity of the brain for each divided area (S30). The measurement unit 140 may know the degree of contraction of the brain for each area through the measured brain capacity. That is, the number of pixels for an object existing in each area is calculated, and through this, it is possible to know how much the volume of the subject's brain is contracted for each area of the brain.

According to an embodiment, the measurement unit 140 may measure the capacity, volume, and thickness of the brain cortex for each divided brain region.

The prediction unit 160 predicts the possibility of dementia through the measured brain capacity based on the deep learning model (S40). That is, based on the deep learning model, the prediction unit 160 may determine the degree of brain contraction of the subject and predict how much the probability of developing dementia is high and whether the patient will develop dementia in the next few years.

In this case, the deep learning model is a model in which feature learning on brain capacity according to age has been completed using an effective data set. In other words, the deep learning model learns a plurality of brain MRI images, extracts brain capacity features by age, compares brain capacity features of normal people and dementia patients, and learns differences, specifically classifying brains by region. Thus, it is possible to learn characteristics of the brain capacity, volume, and thickness of the cerebral cortex for each brain region according to age.

The method for predicting dementia for each brain region according to an embodiment of the present invention described above may be implemented as a program (or application) to be executed in combination with a computer, which is hardware, and stored in a medium.

The above-described program includes C, C++, JAVA, Ruby, which can be read by a processor (CPU) of the computer through the device interface of the computer, in order for the computer to read the program and execute the methods implemented as a program. It may include code coded in a computer language such as machine language. Such code may include a functional code related to a function defining necessary functions for executing the methods, and a control code related to an execution procedure necessary for the processor of the computer to execute the functions according to a predetermined procedure. can do. In addition, such code may further include code related to a memory reference to which location (address address) of the internal or external memory of the computer or the media or additional information necessary for the processor of the computer to execute the functions. have. In addition, when the processor of the computer needs to communicate with any other computer or server in the remote in order to execute the functions, the code uses the communication module of the computer to determine how It may further include a communication-related code for whether to communicate or what information or media to transmit and receive during communication.

The stored medium is not a medium that stores data for a short moment, such as a register, cache, memory, etc., but a medium that stores data semi-permanently and can be read by a device. Specifically, examples of the storage medium include, but are not limited to, ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like. That is, the program may be stored in various recording media on various servers to which the computer can access, or on various recording media on the user's computer. In addition, the medium may be distributed over a computer system connected through a network, and computer-readable codes may be stored in a distributed manner.

In the above, embodiments of the present invention have been described with reference to the accompanying drawings, but those skilled in the art to which the present invention pertains can be implemented in other specific forms without changing the technical spirit or essential features. You will be able to understand. Therefore, the embodiments described above are illustrative in all respects, and should be understood as non-limiting.

100: dementia prediction device
120: Learning Department
140: measuring unit
160: prediction unit
180: communication department

Claims (10)

In the method for predicting dementia by the dementia prediction device,
Obtaining, by the dementia prediction apparatus, a brain MRI image from a subject;
Dividing, by the dementia prediction apparatus, the acquired brain MRI image into a plurality of regions;
Measuring, by the device for predicting dementia, the capacity of the brain for each divided region; And
Predicting, by the dementia prediction apparatus, a possibility of dementia based on a deep learning model, based on the measured brain capacity; Including,
In the step of dividing the brain MRI image into a plurality of regions by the dementia prediction apparatus, the brain MRI image is processed as 2D data using semantic segmentation to generate a plurality of 2D brain MRI images, and pixel Further comprising the step of dividing the plurality of 2D brain MRI images in units,
The step of predicting the likelihood of dementia includes predicting a time to develop dementia based on a deep learning model,
The capacity of the brain corresponds to the volume of each area of the brain,
The device for predicting dementia,
Determine the brain region in each 2D brain MRI image based on the sum of the pixels in the plurality of 2D brain MRI images, measure the volume of each region of the brain by measuring the volume for each region in each 2D brain MRI image, and ,
A method for predicting dementia for each brain region of a brain MRI image, for predicting the time to develop dementia based on the measured contraction of the volume of the brain for each region of the same age group. The method of claim 1,
The deep learning model,
It is a model that has completed learning about the capacity of the brain according to age using an effective data set.
The valid data set is a set of data used for deep learning, and the valid data set is a training data set, which is a set of data for generating a deep learning model, and validation, which is a set of data for verifying the performance of the deep learning model. A method for predicting dementia for each brain region of a brain MRI image, including a data set. delete The method of claim 1,
The step of dividing the brain MRI image into a plurality of regions includes, by the dementia prediction apparatus, recognizing an object in the brain MRI image and dividing the region based on the recognized object-the object corresponds to each region of the brain, and , The object includes at least one of the lateral ventricle (Lateral Ventricle), the amygdala (Amygdala) and the hippocampus (Hippocampus) -; Containing, dementia prediction method for each brain region of the brain MRI image. delete A learning unit that learns brain capacity according to age using an effective data set-the effective data set is a set of data used for deep learning -;
A measurement unit that divides the brain MRI image of the subject into a plurality of areas and measures brain capacity for each of the divided areas;
A prediction unit that predicts the likelihood of dementia and a time to develop dementia based on the deep learning model, based on the measured brain capacity; And
Including; a communication unit that receives the brain MRI image of the subject from the user terminal, and provides the predicted possibility of dementia and the time to develop dementia to the user terminal,
The brain MRI image is processed as 2D data using semantic segmentation to generate a plurality of 2D brain MRI images, and the plurality of 2D brain MRI images are segmented in pixel units,
The capacity of the brain corresponds to the volume of each area of the brain,
The measurement unit determines a brain region in each 2D brain MRI image based on the sum of pixels in the plurality of 2D brain MRI images, and measures the volume for each region in each 2D brain MRI image, thereby Is measured,
The prediction unit predicts the time to suffer from dementia based on the measured contraction of the volume of the brain by comparing the volume of the brain of a person of the same age. The method of claim 6,
The learning unit,
The valid data set includes a training data set, which is a set of data for generating a deep learning model, and a validation data set, which is a set of data for verifying the performance of the deep learning model. Dividing into the training data set and the validation data set, learning the training data set to generate a deep learning model, and verifying the deep learning model generated by the validation data set, dementia for each brain region of a brain MRI image Prediction device. The method of claim 6,
The measuring unit,
Recognizing an object in the brain MRI image, segmenting a region based on the recognized object,
Calculating the number of pixels of an object present in each region-The object corresponds to each region of the brain, and the object includes at least one of a lateral ventricle (Lateral Ventricle), an amygdala (Amygdala), and a hippocampus (Hippocampus). -, A device for predicting dementia by brain region of brain MRI image. delete A program for predicting dementia for each brain region of a brain MRI image, which is combined with a computer as hardware and stored in a medium to execute the method of claim 1.

Images