KR102582731B1 - Method and system for analyzing brain image - Google Patents

Abstract

The present disclosure provides a brain image analysis method performed by at least one processor. This method includes acquiring a brain image including the dopaminergic nervous system and inputting the obtained brain image into a first machine learning model to obtain a region of interest extracted by the first machine learning model, Machine learning models can extract regions of interest associated with the dopaminergic nervous system from the entire brain image.

Description

Brain image analysis method and system {METHOD AND SYSTEM FOR ANALYZING BRAIN IMAGE}

The present disclosure relates to a method and system for analyzing brain images, and specifically, to a method and system for analyzing brain images associated with the dopaminergic nervous system using a machine learning model.

Dopamine is one of the neurotransmitters secreted to transmit signals between brain nerve cells and is produced by removing the carboxyl group from L-DOPA, a precursor synthesized in the brain and kidneys. Dopamine is involved in many functions of the brain and is closely related to behavior and cognition, voluntary movement, motivation, punishment and reward, inhibition of prolactin production, sleep, mood, attention, working memory, and learning. In addition, dopamine acts on the basal ganglia of the brain and plays an important role in allowing us to move our bodies precisely as we want.

Neurons that secrete dopamine are located in the ventral tegmental area of the midbrain, the substantia nigra, and the arcuate nucleus of the hypothalamus. Among these, the substantia nigra provides dopamine innervation to the striatum. This is called the nigrostriatal pathway.

Neurons that secrete dopamine synthesize L-DOPA from tyrosine and then produce dopamine from L-DOPA. The dopamine produced in this way moves to the synaptic vesicle by the vesicular monoamine transporter and is stored in the axon terminal. When the action potential of the neuron reaches the axon terminal, dopamine in the synaptic pocket is released into the synaptic cleft and acts by binding to the dopamine receptor expressed on the postsynaptic membrane. . Dopamine remaining in the synaptic gap without binding to the dopamine receptor is reabsorbed back into the axon terminal by the dopamine transporter, and is also decomposed by monoamine oxidase (MAO).

Parkinson's disease is characterized by motor symptoms such as hand tremors, bradykinesia (slow movement), rigidity (stiff muscles), upper body leaning forward, and postural instability, as well as neuropsychiatric diseases such as depression, anxiety, apathy, impulse control disorders, hallucinations, and psychosis. , is a representative degenerative disease in the field of neurology characterized by non-motor symptoms such as cognitive decline, autonomic nervous system abnormalities, sleep disorders, urinary disorders, and sensory abnormalities.

A characteristic of Parkinson's disease is the unexplained loss of dopamine neurons in the substantia nigra, resulting in a decrease in dopamine in the striatum. Symptoms of Parkinson's disease clearly appear after about 60-80% of the dopaminergic nerves in the substantia nigra are lost, and can be diagnosed through positron emission tomography (PET), a medical imaging test using radioisotopes, or single photon emission tomography (Single Photon Emission Tomography). -Photon emission computed tomography (SPECT) can be used to determine the degree of maintenance of the nigra-striatal pathway in Parkinson's patients.

In PET or SPECT imaging tests for the dopaminergic nervous system, the image pixel value is noticeably high in certain areas such as the striatum, where radiopharmaceuticals specifically bind, and the pixel value is low in other brain areas, making it difficult to accurately determine the structure of the brain. Additionally, if the uptake of radiopharmaceuticals in the striatum is low due to abnormalities in the dopaminergic nervous system, distinction and division of the striatal region becomes more difficult.

Accordingly, in the past, a method was used to manually draw a region of interest (ROI) on a PET or SPECT image, extract pixel values of the striatum and other brain regions, and determine the level of radiopharmaceutical intake through this. . Alternatively, use anatomical information from magnetic resonance imaging (MRI) or computed tomography (CT) to define the dopaminergic nervous system region, such as the striatum, as a region of interest, and then copy this onto the PET or SPECT image to determine the pixel value. was extracted.

However, the method of manually drawing the region of interest directly on PET and SPECT images is time-consuming and cumbersome, and the location, shape, and size of the region of interest differ depending on the operator drawing the region of interest, and the extracted pixel values vary accordingly. There were limits. In addition, methods using MRI or CT require additional tests and images, and spatial discrepancies between images may occur due to patient movement between MRI (or CT) and PET (or SPECT) imaging. This spatial inconsistency may cause errors in pixel values contained in PET or SPECT images extracted by copying the region of interest defined by MRI or CT.

Accordingly, there is a need for a method to automatically extract pixel values of local brain regions, including the striatum, from dopaminergic nervous system PET or SPECT images without the help of MRI or CT.

The present disclosure provides a brain image analysis method, a computer program stored in a recording medium, and a device (system) to solve the above problems.

The present disclosure may be implemented in various ways, including a method, a device (system), and/or a computer program stored in a computer-readable storage medium, and a computer-readable storage medium storing the computer program.

According to an embodiment of the present disclosure, a brain image analysis method performed by at least one processor includes acquiring a brain image including a dopaminergic nervous system and inputting the acquired brain image into a first machine learning model, 1 A step of obtaining a region of interest extracted by a machine learning model, wherein the first machine learning model can extract a region of interest associated with the dopaminergic nervous system from the entire region of the brain image.

Additionally, the first machine learning model can transform a brain image based on a brain template and extract a region of interest from the entire region of the transformed brain image.

Additionally, based on at least one of the size or shape of the brain template, at least one of the size or shape of the brain image may be modified.

Additionally, the first machine learning model can extract a region of interest from the entire area of the transformed brain image based on space of interest information included in the brain template.

In addition, the step of obtaining the extracted region of interest includes additionally inputting user information including at least one of gender, age, family history of brain disease, or disease of the user into the first machine learning model, and the first machine learning model The model can determine the location and size of the region of interest in the entire brain image area based on user information.

In addition, the brain image analysis method further includes the step of performing learning on a first machine learning model using a training set including user information for learning, brain images for learning, and a reference region before acquiring the brain image. And, the first machine learning model can be trained to minimize loss between the region of interest and the reference region output based on user information for learning and brain images for learning.

In addition, in the brain image analysis method, after the step of acquiring the region of interest, characteristic information about the region of interest is input into a second machine learning model, and from the second machine learning model, information about the occurrence of brain disease or progression of brain disease is obtained. It may further include obtaining at least one of the degrees, and the second machine learning model may determine brain disease based on feature information of the region of interest.

Additionally, the feature information about the region of interest may include at least one of the maximum pixel value, minimum pixel value, average pixel value or pixel distribution value, the range of pixel values, or the ratio between the first and second regions included in the region of interest. It can contain one.

Additionally, the first region may be associated with the first striatum included in the region of interest, and the second region may be associated with the second striatum included in the region of interest.

In order to execute the above-described brain image analysis method on a computer, a computer program stored in a computer-readable recording medium may be provided.

According to one embodiment of the present disclosure, an information processing system includes a memory and at least one processor connected to the memory and configured to execute at least one computer-readable program included in the memory, and the at least one program includes, Contains instructions for acquiring a brain image including the dopaminergic nervous system, inputting the acquired brain image into a first machine learning model, and obtaining a region of interest extracted by the first machine learning model, and a first machine learning model. can extract regions of interest associated with the dopaminergic nervous system from the entire area of the brain image.

According to some embodiments of the present disclosure, a region of interest can be accurately extracted from a brain image that is not combined with an MRI or CT image. Brain disease can be determined based on the acquired characteristic information about the region of interest, or brain disease progression information can be accurately determined.

According to some embodiments of the present disclosure, feature information about the region of interest can be obtained from a single brain image that is not combined with MRI or CT, so the use of expensive fusion imaging devices such as PET/MRI or PET/CT is necessary. In addition, costs can be reduced and speed can be improved in diagnosing brain diseases.

According to some embodiments of the present disclosure, since at least one region of interest included in a brain image is extracted using a machine learning model, image quantification errors resulting from poor manipulation can be minimized, and in addition, it is possible to minimize the error in brain image analysis. Accuracy can also be improved.

The effects of the present disclosure are not limited to the effects mentioned above, and other effects not mentioned are clear to a person skilled in the art (referred to as a ‘person skilled in the art’) in the technical field to which this disclosure pertains from the description of the claims. It will be understandable.

Embodiments of the present disclosure will be described with reference to the accompanying drawings described below, in which like reference numerals indicate like elements, but are not limited thereto.
1 is a schematic diagram showing a configuration in which an information processing system according to an embodiment of the present disclosure is connected to communicate with each of a photographing device and a user terminal.
Figure 2 is a block diagram showing the internal configuration of an information processing system according to an embodiment of the present disclosure.
Figure 3 is a block diagram showing the internal configuration of a processor included in an information processing system according to an embodiment of the present disclosure.
Figure 4 is a diagram illustrating data output through a first machine learning model according to an embodiment of the present disclosure.
Figure 5 is a diagram illustrating a process in which a first machine learning model is learned, according to an embodiment of the present disclosure.
Figure 6 is a diagram showing an example of a first brain image input to a first machine learning model and a second brain image output from the first machine learning model.
Figure 7 is a diagram showing an example of a third brain image input to the first machine learning model and a fourth brain image output from the first machine learning model.
Figure 8 is a diagram illustrating data output through a second machine learning model according to an embodiment of the present disclosure.
Figure 9 is a diagram illustrating a process in which a second machine learning model is learned, according to an embodiment of the present disclosure.
Figure 10 is a diagram showing an example of a fifth brain image input to a second machine learning model and a first region of interest output from the second machine learning model.
Figure 11 is a diagram showing an example of a sixth brain image input to a second machine learning model and a second region of interest output from the second machine learning model.
Figure 12 is a diagram illustrating data output through a third machine learning model according to an embodiment of the present disclosure.
Figure 13 is a diagram illustrating a process in which a third machine learning model is learned, according to an embodiment of the present disclosure.
FIG. 14 is a diagram illustrating an example of a machine learning model including an artificial neural network, according to an embodiment of the present disclosure.
Figure 15 is a graph comparing the average pixel value of the right basal ganglia obtained by a combined MRI/PET method with the average pixel value of the right basal ganglia obtained according to an embodiment of the present disclosure.
Figure 16 is a flowchart for explaining a brain image analysis method according to an embodiment of the present disclosure.

Hereinafter, specific details for implementing the present disclosure will be described in detail with reference to the attached drawings. However, in the following description, detailed descriptions of well-known functions or configurations will be omitted if there is a risk of unnecessarily obscuring the gist of the present disclosure.

In the accompanying drawings, identical or corresponding components are given the same reference numerals. Additionally, in the description of the following embodiments, overlapping descriptions of identical or corresponding components may be omitted. However, even if descriptions of components are omitted, it is not intended that such components are not included in any embodiment.

Advantages and features of the disclosed embodiments and methods for achieving them will become clear by referring to the embodiments described below in conjunction with the accompanying drawings. However, the present disclosure is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the present disclosure is complete and that the present disclosure does not convey the scope of the invention to those skilled in the art. It is provided only for complete information.

Terms used in this specification will be briefly described, and the disclosed embodiments will be described in detail. The terms used in this specification are general terms that are currently widely used as much as possible while considering the function in the present disclosure, but this may vary depending on the intention or precedent of a technician working in the related field, the emergence of new technology, etc. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description of the relevant invention. Accordingly, the terms used in this disclosure should be defined based on the meaning of the term and the overall content of the present disclosure, rather than simply the name of the term.

In this specification, singular expressions include plural expressions, unless the context clearly specifies the singular. Additionally, plural expressions include singular expressions, unless the context clearly specifies plural expressions. When it is said that a certain part includes a certain element throughout the specification, this does not mean excluding other elements, but may further include other elements, unless specifically stated to the contrary.

Additionally, the term 'module' or 'unit' used in the specification refers to a software or hardware component, and the 'module' or 'unit' performs certain roles. However, 'module' or 'unit' is not limited to software or hardware. A 'module' or 'unit' may be configured to reside on an addressable storage medium and may be configured to run on one or more processors. Thus, as an example, a 'module' or 'part' refers to components such as software components, object-oriented software components, class components and task components, processes, functions and properties. , procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, or variables. Components and 'modules' or 'parts' may be combined into smaller components and 'modules' or 'parts' or further components and 'modules' or 'parts'. Could be further separated.

According to an embodiment of the present disclosure, a 'module' or 'unit' may be implemented with a processor and memory. 'Processor' should be interpreted broadly to include general-purpose processors, central processing units (CPUs), microprocessors, digital signal processors (DSPs), controllers, microcontrollers, state machines, etc. In some contexts, 'processor' may refer to an application-specific integrated circuit (ASIC), programmable logic device (PLD), field programmable gate array (FPGA), etc. 'Processor' refers to a combination of processing devices, for example, a combination of a DSP and a microprocessor, a combination of a plurality of microprocessors, a combination of one or more microprocessors in combination with a DSP core, or any other such combination of configurations. You may. Additionally, 'memory' should be interpreted broadly to include any electronic component capable of storing electronic information. 'Memory' refers to random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), programmable read-only memory (PROM), erasable-programmable read-only memory (EPROM), May also refer to various types of processor-readable media, such as electrically erasable PROM (EEPROM), flash memory, magnetic or optical data storage, registers, etc. A memory is said to be in electronic communication with a processor if the processor can read information from and/or write information to the memory. The memory integrated into the processor is in electronic communication with the processor.

In the present disclosure, 'system' may include at least one of a server device and a cloud device, but is not limited thereto. For example, a system may consist of one or more server devices. As another example, a system may consist of one or more cloud devices. As another example, the system may be operated with a server device and a cloud device configured together.

In addition, terms such as first, second, A, B, (a), (b) used in the following embodiments are only used to distinguish one component from another component, and the terms The nature, sequence, or order of the relevant components are not limited.

Additionally, in the following embodiments, when a component is described as being 'connected', 'coupled' or 'connected' to another component, the component may be directly connected or connected to the other component. , it should be understood that another component may be 'connected', 'combined', or 'connected' between each component.

In addition, as used in the following embodiments, 'comprises' and/or 'comprising' means that the mentioned component, step, operation, and/or element includes one or more other components, steps, or operations. and/or the presence or addition of elements.

Hereinafter, various embodiments of the present disclosure will be described in detail with reference to the attached drawings.

Figure 1 is a schematic diagram showing a configuration in which the information processing system 110 according to an embodiment of the present disclosure is connected to communicate with each of the photographing device 120 and the user terminal 130. Here, a plurality of the information processing system 110, the photographing device 120, and the user terminal 130 may be located in the same location or may be located in different locations. Here, the network 140 may be a private network built in a hospital. As another example, the photographing device 120 may be directly connected to and communicate with the information processing system 110 through a dedicated line. As another example, the information processing system 110 may not communicate with at least one of the photographing device 120 or the user terminal 130 through the network 140. In this case, the information processing system 110 receives a file containing a brain image captured by the imaging device 120 through a storage means (e.g., USB memory, DVD, etc.), and the brain image included in the input file is input. Analysis of video can be performed.

The imaging device 120 may be a device capable of acquiring input image data, such as a medical image containing functional information, by imaging the patient's brain state. For example, the imaging device 120 uses positron emission tomography (PET) or single-photon emission computed tomography (SPECT), which are medical imaging tests using radioisotopes, to detect the brain. It may be a device that can capture video.

For example, the imaging device 120 includes at least one of a scintillation camera, a gamma camera, and a PET scanner, and includes one of the frontal cortex, posterior cingulate, lateral temporal lobe, parietal lobe, occipital cortex, caudate nucleus, central temporal lobe, and anterior cingulate in the patient's brain. You can film at least one part. At this time, the radioactive tracer is injected into the patient's body, and the imaging device 120 can image the patient's brain into which the radioactive tracer has been injected. Here, the radioactive tracer may be a substance that images areas related to the dopaminergic nervous system. Types of radioactive tracers may include, but are not limited to, at least one radioactive isotope among F-18, C-11, N-13, and O-15.

The user terminal 130 may receive brain image analysis results from the information processing system 110. The user terminal 130 may be a computing device owned by a user who has requested brain image analysis. The user terminal 130 may receive a brain image captured by the imaging device 120 and request analysis of the brain image from the information processing system 110. Additionally, the user terminal 130 may receive the analysis results of the brain image from the information processing system 110.

The network 140 may be configured to support communication for each device/terminal/system. Depending on the installation environment, the network 140 may be, for example, a wired network such as Ethernet, a wired home network (Power Line Communication), a telephone line communication device, and RS-serial communication, a mobile communication network, a wireless LAN (WLAN), It may consist of wireless networks such as Wi-Fi, Bluetooth, and ZigBee, or a combination thereof. The communication method is not limited, and may include not only communication methods utilizing communication networks that the network 140 may include (e.g., mobile communication networks, wired Internet, wireless Internet, broadcasting networks, satellite networks, etc.), but also short-range wireless communication between IoT devices. You can. According to one embodiment, the network 140 may be a private network established in a hospital.

The information processing system 110 is one or more server devices and/or databases capable of storing, providing and executing computer executable programs (e.g., downloadable applications) and data associated with brain image analysis, etc., or a cloud computing service-based It may include one or more distributed computing devices and/or distributed databases.

The information processing system 110 may receive brain images from the imaging device 120 or the user terminal 130. As another example, the information processing system 110 may receive a file containing a brain image captured by the imaging device 120 through a storage means (eg, USB memory, DVD, etc.). According to one embodiment, the information processing system 110 may input a brain image into a pre-trained machine learning model and obtain a region of interest of the brain image output from the machine learning model. Here, the region of interest may be a partial region associated with the dopaminergic nervous system among the entire brain image region.

The information processing system 110 can generate feature information about the region of interest, input this feature information into another machine learning model, and obtain disease analysis results for brain images from another machine learning model. In addition, the information processing system 110 may transmit the disease analysis results for the acquired brain image to the user terminal 130.

Figure 2 is a block diagram showing the internal configuration of the information processing system 110 according to an embodiment of the present disclosure. The information processing system 110 may include a memory 210, a processor 220, a communication module 230, and an input/output interface 240. As shown in FIG. 2, the information processing system 110 may be configured to communicate information and/or data over a network using a communication module 230.

Memory 210 may include any non-transitory computer-readable recording medium. According to one embodiment, the memory 210 may include a non-permanent mass storage device, such as a read only memory (ROM), a disk drive, a solid state drive (SSD), or flash memory. You can. As another example, non-perishable mass storage devices such as ROM, SSD, flash memory, disk drive, etc. may be included in the information processing system 110 as a separate persistent storage device that is distinct from memory. Additionally, the memory 210 may store an operating system and at least one program code (eg, code for brain image analysis, etc.). In FIG. 2, the memory 210 is shown as a single memory, but this is only for convenience of explanation, and the memory 210 may include a plurality of memories.

Software components may be loaded from a computer-readable recording medium separate from the memory 210. These separate computer-readable recording media may include recording media directly connectable to the information processing system 110, for example, floppy drives, disks, tapes, DVD/CD-ROM drives, memory cards, etc. It may include a computer-readable recording medium. As another example, software components may be loaded into the memory 210 through the communication module 230 rather than a computer-readable recording medium. For example, at least one program may be loaded into the memory 210 based on a computer program installed by files provided through the communication module 230 by developers or a file distribution system that distributes the installation file of the application. You can.

The communication module 230 may provide a configuration or function for the user terminal and/or external device and the information processing system 110 to communicate with each other through a network, and may provide a configuration or function for the information processing system 110 to communicate with the external system. Configuration or functionality can be provided.

In addition, the input/output interface 240 of the information processing system 110 is connected to the information processing system 110 or means for interfacing with a device (not shown) for input or output that the information processing system 110 may include. It can be. For example, the input/output interface 240 may include at least one of a PCI express interface and an Ethernet interface. In FIG. 2 , the input/output interface 240 is shown as an element configured separately from the processor 220, but the present invention is not limited thereto, and the input/output interface 240 may be included in the processor 220. Information processing system 110 may include more components than those in FIG. 2 . According to one embodiment, the information processing system 110 may receive a file containing a brain image, and in this case, the input/output interface 240 may acquire the brain image included in the file.

The processor 220 may be configured to process instructions of a computer program by performing basic arithmetic, logic, and input/output operations. Commands may be provided to the processor 220 by the memory 210 or the communication module 230. For example, processor 220 may be configured to execute received instructions according to program code stored in a recording device such as memory 210. In one embodiment, the memory 210 may store at least one program including instructions for executing methods according to various embodiments described later. Additionally, the processor 220 may be configured to execute at least one program.

According to one embodiment, the program may include instructions for acquiring a brain image including the dopaminergic nervous system, inputting the acquired brain image into a machine learning model, and obtaining a region of interest extracted by the machine learning model. .

Figure 3 is a block diagram showing the internal configuration of the processor 220 included in the information processing system 110 according to an embodiment of the present disclosure. As shown in FIG. 3 , the processor 220 may include a model learning unit 310, a region of interest extracting unit 320, and a disease analysis unit 330.

The model learning unit 310 may perform learning on at least one machine learning model using a training set. For example, the model learning unit 310 may use the first training set to train a first machine learning model that extracts a region of interest based on a template. As another example, the model learning unit 310 may train a second machine learning model that extracts a personalized region of interest using the second training set. As another example, the model learning unit 310 may use the third training set to learn a third machine learning model that analyzes brain disease based on feature information in the region of interest. The method by which the first machine learning model, the second machine learning model, and the third machine learning model are learned will be described later with reference to FIGS. 5, 9, and 13.

The region of interest extractor 320 acquires a brain image captured by an imaging device, inputs the acquired brain image into a first machine learning model or a second machine learning model, and selects the area required for brain disease analysis among all regions in the brain image. Areas of interest can be obtained. Here, the region of interest may include a cell region related to the dopaminergic nervous system and may be a partial region included in the brain image. For example, the region of interest may be a region associated with at least one of the basal ganglia, substantia nigra, and striatum.

According to one embodiment, the region of interest extractor 320 may preprocess the brain image and then input the preprocessed brain image into a first machine learning model or a second machine learning model. For example, the region of interest extractor 320 may perform preprocessing, such as smoothing processing, upscaling processing, and unnecessary object removal processing, on the brain image. Here, the unnecessary object is preprocessing to remove objects unnecessary for brain disease analysis. For example, the brain image may be preprocessed by removing objects related to bones from the brain image.

The region of interest extractor 320 may input a brain image into a first machine learning model or a second machine learning model, and obtain at least one region of interest output from the first machine learning model or the second machine learning model.

The disease analysis unit 330 can analyze brain disease using a third machine learning model. According to one embodiment, the disease analysis unit 330 generates feature information about the region of interest input to the third machine learning model based on the region of interest output from the first machine learning model or the second machine learning model. You can. Here, feature information about the region of interest may include pixel values and/or ratio values.

According to one embodiment, the disease analysis unit 330 determines at least one of the maximum pixel value, minimum pixel value, average pixel value, pixel distribution value, or range of pixel values of the region of interest, based on the pixel values included in the region of interest. can be calculated.

Additionally or alternatively, the disease analysis unit 330 may calculate the size ratio between the first region corresponding to the first striatum and the second region corresponding to the second striatum included in the region of interest. For example, the region of interest may include a plurality of striatum regions, and among these, the size ratio between the first region associated with the first striatum and the second region associated with the second striatum can be calculated. For example, the first area may be located on the left or front side of the brain image, and the second area may be located on the right or back side of the brain image.

The disease analysis unit 330 inputs feature information about the region of interest, including the calculated pixel value and/or ratio value, into a third machine learning model, and the brain disease analysis result output from the third machine learning model. can be obtained. Here, the feature information about the region of interest may include at least one of the maximum pixel value, minimum pixel value, average pixel value or pixel distribution value, range of pixel values, or ratio value between the first region and the second region of the region of interest. You can. Here, the brain disease analysis result may include at least one of information on whether the brain disease has occurred, the name of the brain disease if the brain disease has developed, or the degree of progression of the brain disease (eg, stage 1, stage 2, stage 3, etc.).

FIG. 4 is a diagram illustrating data output through the first machine learning model 410 according to an embodiment of the present disclosure. As illustrated in FIG. 4 , the first machine learning model 410 may be an artificial intelligence-based model configured to output a region of interest 430 based on a brain template. Here, the brain template is a reference brain image, has a predetermined size and shape, and may include at least one spatial information of interest. The spatial information of interest may be a relative position range or a pixel coordinate range with a specific point set in the brain template. Space of interest information may include a plurality of region information separated from each other.

The brain image 420 captured by the imaging device may be input to the first machine learning model 410. The first machine learning model 410 may perform calculations based on the brain image 420 and output a region of interest 430 based on the brain template. According to one embodiment, the first machine learning model 410 transforms the input brain image 420 to correspond to the size and/or shape of the brain template, and performs calculations based on the transformed brain image 420. , the region of interest 430 can be extracted from the brain image 420 and output. At this time, the first machine learning model 410 may extract the region of interest 430 from the entire area of the modified brain image 420 based on the space of interest information included in the brain template.

FIG. 5 is a diagram illustrating a process in which the first machine learning model 410 is learned, according to an embodiment of the present disclosure. Learning of the first machine learning model 410 may be performed by at least one processor included in the information processing system (eg, the model learning unit in FIG. 3). A first training set containing a plurality of learning data may be used to learn the first machine learning model 410. The first training set may include learning data including a brain image for learning and a reference region. Here, the reference area is a measure used as a ground truth in the learning process and may be an area associated with the dopaminergic nervous system in learning brain imaging. According to one embodiment, the reference region and learning brain image included in the first training set may be generated based on medical data of a patient who actually suffers from or has suffered from a brain disease (eg, Parkinson's disease).

Referring to FIG. 5 , the processor may input a learning brain image 510 and a brain template 520 into the first machine learning model 410. In some embodiments, the brain template 520 may not be input separately but may be embedded in the first machine learning model 410.

According to one embodiment, the first machine learning model 410 transforms the size and/or shape of the learning brain image 510 to correspond to the size and/or shape of the brain template 520, and the transformed learning brain A region of interest 530 may be extracted from the image 510 and the extracted region of interest 530 may be output.

The reference area 540 included in the learning data is extracted, and a loss value 550, which is the difference between the interest area 530 output from the first machine learning model 410 and the reference area 540, is calculated. It can be fed back to the first machine learning model 410. At this time, various loss functions, such as a perceptual loss function, can be used to calculate the loss value 550 between the region of interest 530 and the reference region 540.

The fed back loss value 550 may be reflected in the first machine learning model 410 to adjust the weight of at least one node included in the first machine learning model 410. Here, the node may be a node included in an artificial neural network.

When a plurality of learning data included in the first training set is repeatedly input to the first machine learning model 410 and repetitive learning is performed, the weight of the node included in the first machine learning model 410 is set to the optimal value. can converge.

FIG. 6 is a diagram illustrating an example of a first brain image 610 input to a first machine learning model and a second brain image 620 output from the first machine learning model. The black displayed area in the brain image input in FIG. 6 and the drawings described below is for convenience of explanation, and it is to be clarified that there may not be an area marked in a separate color in the brain image input to the machine learning model. .

The area marked in black in the first brain image 610 in FIG. 6 may be an area related to the dopaminergic nervous system. The first machine learning model may generate the second brain image 620 by modifying the first brain image 610 based on the brain template. The black area included in the second brain image 620 may be the area of interest.

According to one embodiment, the first machine learning model may extract and output only the region of interest (eg, black region) from the second brain image 620. According to some embodiments, the first machine learning model may output the second brain image 620 and the location range (eg, pixel coordinate range) for the region of interest.

FIG. 7 is a diagram illustrating examples of the third brain image 710 input to the first machine learning model and the fourth brain image 720 output from the first machine learning model. The area marked in black in the third brain image 710 in FIG. 7 may be an area related to the dopaminergic nervous system. The first machine learning model may generate the fourth brain image 720 by modifying the third brain image 710 based on the brain template. The black area included in the fourth brain image 720 may be the area of interest.

According to one embodiment, the first machine learning model may extract and output only the region of interest (eg, black region) from the fourth brain image 720. According to some embodiments, the first machine learning model may output the fourth brain image 720 and the location range (eg, pixel coordinate range) for the region of interest.

FIG. 8 is a diagram illustrating data output through a second machine learning model 810 according to an embodiment of the present disclosure. As illustrated in FIG. 8 , the second machine learning model 810 may be an artificial intelligence-based model configured to output a personalized region of interest 830. According to one embodiment, user information may be additionally input into the second machine learning model 810. Here, the user information may include at least one of gender, age, family history of brain disease, or the user's disease. Here, the brain disease family history may include the name of the brain disease that the user's family has suffered from or is suffering from. Additionally, the user's disease may include the name of the disease that the user has suffered from or is suffering from. Here, the user's disease may include other types of diseases in addition to brain disease.

The brain image 820 captured by the imaging device may be input to the second machine learning model 810. The second machine learning model 810 may perform calculations based on the brain image 820 and user information to extract and output a personalized (user) region of interest 830 from the brain image 820. According to one embodiment, the second machine learning model 810 determines the location and size of the region of interest in the entire brain image 820 based on user information, and based on the determined location and size, the second machine learning model 810 determines the location and size of the region of interest in the entire brain image 820. The region of interest 830 can be determined and extracted from the entire region 820.

According to one embodiment, the extracted region of interest may change depending on user information. That is, if the user information is different, the location and/or size of the area of interest may be determined differently.

FIG. 9 is a diagram illustrating a process in which the second machine learning model 810 is learned, according to an embodiment of the present disclosure. Learning for the second machine learning model 810 may be performed by at least one processor (eg, the model learning unit in FIG. 3) included in the information processing system. A second training set containing a plurality of learning data may be used to learn the second machine learning model 810. The second training set may include learning data including brain images for learning, user information, and a reference region. Here, the reference area is a measure used as a ground truth in the learning process and may be an area associated with the dopaminergic nervous system. Additionally, the user information may include at least one of gender, age, family history of brain disease, or the user's disease. According to one embodiment, the second training set may be generated based on medical data of a patient who actually suffers from or has suffered from a brain disease (eg, Parkinson's disease).

Referring to FIG. 9 , the processor may input the learning brain image 910 and user information 920 into the second machine learning model 810. According to one embodiment, the second machine learning model 810 determines the size and location of the area of interest based on the user information 920, and selects the area corresponding to the determined size and location as the area of interest 930. ) can be output.

The reference area 940 included in the learning data is extracted, and the loss value 950, which is the difference between the interest area 930 output from the second machine learning model 810 and the reference area 940, is calculated. It can be fed back to the second machine learning model 810. At this time, various loss functions, such as a perception-based loss function, can be used to calculate the loss value 950 between the region of interest 930 and the reference region 940.

The fed back loss value 950 may be reflected in the second machine learning model 810 to adjust the weight of at least one node included in the second machine learning model 810. Here, the node may be a node included in an artificial neural network.

When a plurality of learning data included in the second training set are repeatedly input to the second machine learning model 810 and repetitive learning is performed, the weight of the node included in the second machine learning model 810 is set to the optimal value. can converge.

FIG. 10 is a diagram illustrating an example of the fifth brain image 1010 input to the second machine learning model and the first region of interest 1020 output from the second machine learning model. The area marked in black in the fifth brain image 1010 in FIG. 10 may be an area related to the dopaminergic nervous system. The second machine learning model can determine the location and size of the object of interest in the input brain image 1010 based on user information, extract and output the region of interest 1020 corresponding to the determined location and size. Here, the region of interest 1020 may be related to the dopaminergic nervous system.

FIG. 11 is a diagram illustrating an example of a sixth brain image 1110 input to a second machine learning model and a second region of interest 1120 output from the second machine learning model. The second machine learning model can determine the location and size of the object of interest in the input brain image 1110 based on user information, extract and output the region of interest 1120 corresponding to the determined location and size. Here, the region of interest 1120 may be related to the dopaminergic nervous system.

FIG. 12 is a diagram illustrating data output through a third machine learning model 1210 according to an embodiment of the present disclosure. As illustrated in FIG. 12, the third machine learning model 1210 may be an artificial intelligence-based model configured to output disease analysis result information.

Feature information 1220 about the region of interest may be input into the third machine learning model 1210. The third machine learning model 1210 may perform calculations based on feature information 1220 for the region of interest and output the user's disease analysis result information 1230.

According to one embodiment, the third machine learning model 1210 is the maximum pixel value, minimum pixel value, average pixel value or pixel variance value, the range of pixel values, or the first region and the first region and the second pixel value included in the feature information for the region of interest. Based on at least one of the ratio values between the two areas, it can be determined whether the user has brain disease. Additionally, when a brain disease occurs, the third machine learning model 1210 can determine the name of the brain disease and the degree of progression of the brain disease (eg, stage 1, stage 2, stage 3, etc.). For example, the brain disease may be Parkinson's disease or another brain disease.

The third machine learning model 1210 includes information on whether a brain disease has occurred and/or, if a brain disease has developed, the name of the brain disease and the degree of progression of the brain disease (e.g., stage 1, stage 2, stage 3, etc.). Disease analysis result information (1230) can be output.

FIG. 13 is a diagram illustrating a process in which a third machine learning model 1210 is learned, according to an embodiment of the present disclosure. Learning for the third machine learning model 1210 may be performed by at least one processor included in the information processing system (eg, the model learning unit in FIG. 3). A third training set containing a plurality of learning data may be used to learn a third machine learning model 1210. The third training set may include feature information and reference results for learning. Here, the feature information for learning may include at least one of a maximum pixel value, a minimum pixel value, an average pixel value or a pixel distribution value, a range of pixel values, or a ratio value between the first area and the second area. Additionally, the reference result is a measure used as a ground truth in the learning process and may include information about brain disease associated with learning feature information. For example, the reference result may include information on whether a brain disease has occurred and/or, if a brain disease has developed, the name of the brain disease and the degree of progression of the brain disease (e.g., stage 1, stage 2, stage 3, etc.). there is. According to one embodiment, the learning feature information and reference results included in the third training set may be generated based on medical data of a patient who actually suffers from or has suffered from a brain disease (eg, Parkinson's disease).

Referring to FIG. 13, the processor may input learning feature information 1310 into the third machine learning model 1210. According to one embodiment, the third machine learning model 1310 may perform calculations based on the input learning feature information 1310 and output an analysis result 1320. Here, the analysis result 1320 may include information on whether a brain disease has occurred and/or, if a brain disease has developed, the name of the brain disease and the degree of progression of the brain disease (e.g., stage 1, stage 2, stage 3, etc.). there is

The reference result 1330 included in the learning data is extracted, and the loss value 1340, which is the difference between the analysis result 1320 output from the third machine learning model 1210 and the reference result 1330, is calculated. It can be fed back to the third machine learning model 410. According to one embodiment, various functions that can calculate the loss value are called, and the loss value 1340 can be calculated.

The fed back loss value 1340 may be reflected in the third machine learning model 1210 to adjust the weight of at least one node included in the first machine learning model 1210. Here, the node may be a node included in an artificial neural network.

When a plurality of learning data included in the third training set is repeatedly input to the third machine learning model 1210 and repetitive learning is performed, the weight of the node included in the third machine learning model 1210 is set to the optimal value. can converge.

FIG. 14 is a diagram illustrating an example of a machine learning model 1400 including an artificial neural network, according to an embodiment of the present disclosure. In machine learning technology and cognitive science, the machine learning model 1400 may refer to a statistical learning algorithm implemented based on the structure of a biological neural network or a structure that executes the algorithm. In other words, the machine learning model 1400 has nodes, which are artificial neurons that form a network by combining synapses, as in a biological neural network, repeatedly adjusting the weights of synapses to produce the correct output and inferred output corresponding to a specific input. It represents a machine learning model with problem-solving capabilities by learning to reduce the error between the two.

According to one embodiment, the machine learning model 1400 may be implemented as a multilayer perceptron (MLP) consisting of multiple layers of nodes and connections between them. The machine learning model 1400 according to this embodiment can be implemented using one of various artificial neural network structures including MLP. The machine learning model 1400 is located between an input layer that receives input signals or data from the outside, an output layer that outputs output signals or data corresponding to the input data, and an input layer and an output layer, and extracts characteristics by receiving signals from the input layer. It consists of n hidden layers that are passed to the output layer. Here, the output layer receives a signal from the hidden layer and outputs it to the outside. During the learning process, the weight associated with at least one node may be adjusted.

According to one embodiment, at least one of the first machine learning model, the second machine learning model, and the third machine learning model may be configured to include the artificial neural network illustrated in FIG. 14. As illustrated in FIG. 14 , when a brain image is input to the machine learning model 1400, a region of interest required for brain disease analysis may be output among the entire area of the brain image.

Figure 15 is a graph (1500) comparing the average pixel value of the right basal ganglia obtained by a combined MRI/PET method with the average pixel value of the right basal ganglia obtained according to an embodiment of the present disclosure. As illustrated in Figure 15, the average pixel value for the region of interest obtained by inputting PET images into a machine learning model without using MRI is almost identical to the average pixel value of the right basal ganglia obtained through MRI/PET combined method. You can see that Accordingly, according to the embodiment of the present disclosure, it can be seen that the accuracy of the average pixel value for the obtained region of interest is high.

FIG. 16 is a flowchart illustrating a brain image analysis method 1600 according to an embodiment of the present disclosure. The method shown in FIG. 16 is only an example to achieve the purpose of the present disclosure, and of course, some steps may be added or deleted as needed. Additionally, the method shown in FIG. 16 may be performed by at least one processor included in the information processing system shown in FIG. 2. For convenience of explanation, it will be described that each step shown in FIG. 16 is performed by a processor included in the information processing system shown in FIG. 2.

First, the processor can acquire a brain image including the dopaminergic nervous system (S1610). Here, the brain image may be PET-based brain image or SPECT-based brain image.

Afterwards, the processor may input the acquired brain image into the first machine learning model to obtain the region of interest extracted by the first machine learning model (S1620). According to one embodiment, the first machine learning model may extract a region of interest associated with the dopaminergic nervous system from the entire brain image.

According to one embodiment, the first machine learning model may transform a brain image based on a brain template and extract a region of interest from the entire region of the transformed brain image. At this time, based on at least one of the size or shape of the brain template, at least one of the size or shape of the brain image may be modified. Additionally, the first machine learning model can extract a region of interest from the entire area of the transformed brain image based on space of interest information included in the brain template.

According to some embodiments, the processor may additionally input user information including at least one of gender, age, family history of brain disease, or user's disease into the first machine learning model. In this case, the first machine learning model can determine the location and size of the region of interest in the entire brain image area based on user information. Additionally, before acquiring the brain image, the processor may perform learning on the first machine learning model using a training set including user information for learning, brain images for learning, and a reference region. In this case, the first machine learning model can be trained to minimize loss between the region of interest and the reference region output based on user information for learning and brain images for learning.

Thereafter, the processor may input feature information about the region of interest into the second machine learning model and obtain at least one of information on whether brain disease occurs or the degree of brain disease progression from the second machine learning model. Here, the second machine learning model can determine brain disease based on characteristic information of the region of interest.

According to one embodiment, the feature information about the region of interest includes the maximum pixel value, minimum pixel value, average pixel value or pixel distribution value, the range of pixel values, or the first and second regions included in the region of interest. It may include at least one of the ratio values between Here, the first region may be associated with the first striatum included in the region of interest, and the second region may be associated with the second striatum included in the region of interest. Taking the second brain image illustrated in FIG. 6 as an example, the processor can calculate the size ratio between the first region located on the left and marked in black and the second region located on the right and marked in black, and calculate The ratio may be included as feature information for the area of interest.

The above flowchart and above description are only examples and may be implemented differently in some embodiments. For example, in some embodiments, the order of each step may be changed, some steps may be performed repeatedly, some steps may be omitted, or some steps may be added.

The above-described method may be provided as a computer program stored in a computer-readable recording medium for execution on a computer. Media may be used to continuously store executable programs on a computer, or may be temporarily stored for execution or download. In addition, the medium may be a variety of recording or storage means in the form of a single or several pieces of hardware combined. It is not limited to a medium directly connected to a computer system and may be distributed over a network. Examples of media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and There may be something configured to store program instructions, including ROM, RAM, flash memory, etc. Additionally, examples of other media include recording or storage media managed by app stores that distribute applications, sites or servers that supply or distribute various other software, etc.

The methods, operations, or techniques of this disclosure may be implemented by various means. For example, these techniques may be implemented in hardware, firmware, software, or a combination thereof. Those skilled in the art will understand that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the disclosure herein may be implemented in electronic hardware, computer software, or combinations of both. To clearly illustrate this interchange of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and design requirements imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementations should not be interpreted as causing a departure from the scope of the present disclosure.

In a hardware implementation, the processing units used to perform the techniques may include one or more ASICs, DSPs, digital signal processing devices (DSPDs), programmable logic devices (PLDs). ), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, electronic devices, and other electronic units designed to perform the functions described in this disclosure. , a computer, or a combination thereof.

Accordingly, the various illustrative logical blocks, modules, and circuits described in connection with this disclosure may be general-purpose processors, DSPs, ASICs, FPGAs or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or It may be implemented or performed as any combination of those designed to perform the functions described in. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, such as a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other configuration.

For firmware and/or software implementations, techniques include random access memory (RAM), read-only memory (ROM), non-volatile random access memory (NVRAM), and PROM ( on computer-readable media such as programmable read-only memory (EPROM), electrically erasable PROM (EEPROM), flash memory, compact disc (CD), magnetic or optical data storage devices, etc. It may also be implemented as stored instructions. Instructions may be executable by one or more processors and may cause the processor(s) to perform certain aspects of the functionality described in this disclosure.

When implemented in software, the techniques described above may be stored on or transmitted through a computer-readable medium as one or more instructions or code. Computer-readable media includes both computer storage media and communication media, including any medium that facilitates transfer of a computer program from one place to another. Storage media may be any available media that can be accessed by a computer. By way of non-limiting example, such computer readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or the desired program code in the form of instructions or data structures. It can be used to transfer or store data and can include any other media that can be accessed by a computer. Any connection is also properly termed a computer-readable medium.

For example, if the Software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair cable, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, , fiber optic cable, twisted pair, digital subscriber line, or wireless technologies such as infrared, radio, and microwave are included within the definition of medium. As used herein, disk and disk include CD, laser disk, optical disk, digital versatile disc (DVD), floppy disk, and Blu-ray disk, where disks are usually magnetic. It reproduces data optically, while discs reproduce data optically using lasers. Combinations of the above should also be included within the scope of computer-readable media.

A software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known. An exemplary storage medium may be coupled to the processor such that the processor may read information from or write information to the storage medium. Alternatively, the storage medium may be integrated into the processor. The processor and storage medium may reside within an ASIC. ASIC may exist within the user terminal. Alternatively, the processor and storage medium may exist as separate components in the user terminal.

Although the above-described embodiments have been described as utilizing aspects of the presently disclosed subject matter in one or more standalone computer systems, the disclosure is not limited thereto and may also be implemented in conjunction with any computing environment, such as a network or distributed computing environment. . Furthermore, aspects of the subject matter of this disclosure may be implemented in multiple processing chips or devices, and storage may be similarly effected across the multiple devices. These devices may include PCs, network servers, and portable devices.

Although the present disclosure has been described in relation to some embodiments in this specification, various modifications and changes may be made without departing from the scope of the present disclosure as can be understood by a person skilled in the art to which the invention pertains. Additionally, such modifications and changes should be considered to fall within the scope of the claims appended hereto.

110: Information processing system
120: shooting device
130: user terminal
140: network

Claims (11)

In a brain image analysis method performed by at least one processor,
Performing learning on a first machine learning model using a training set including user information for learning, brain images for learning, and a reference region - the first machine learning model is output based on the user information for learning and brain images for learning. is learned to minimize loss between the region of interest and the reference region, and the user information for learning includes at least one of a family history of a brain disease for learning or a user disease for learning;
Obtaining a brain image including the dopaminergic nervous system;
Obtaining user information including at least one of a family history of brain disease or a user's disease;
Inputting the acquired brain image and the acquired user information into a first machine learning model to obtain a region of interest extracted by the first machine learning model; and
Characteristic information including the ratio of the first region and the second region included in the region of interest is input to a second machine learning model, and information about whether brain disease occurs or the degree of brain disease progression is obtained from the second machine learning model. Steps to obtain at least one
Including,
The first region is associated with the first striatum located on the left or front of the brain image, and the second region is associated with the second striatum located on the right or rear of the brain image,
The first machine learning model determines the location and size of the region of interest associated with the dopaminergic nervous system in the entire area of the brain image, based on the user information, and selects the region of interest corresponding to the determined location and size. Extracted from the entire brain image area,
The second machine learning model is a brain image analysis method that determines brain disease based on feature information including the ratio of the first area and the second area.
delete delete delete delete delete delete delete delete A computer program stored in a computer-readable recording medium for executing the method according to claim 1 on a computer.
As an information processing system,
Memory; and
At least one processor connected to the memory and configured to execute at least one computer-readable program included in the memory
Including,
The at least one program is,
A training set containing learning user information, learning brain images, and a reference region is used to learn a first machine learning model - the first machine learning model is output based on the learning user information and learning brain images. It is learned to minimize loss between the region of interest and the reference region, and the learning user information includes at least one of a family history of a brain disease for learning or a user disease for learning -,
Acquire brain images involving the dopaminergic nervous system,
Obtain user information including at least one of a family history of brain disease or the user's disease,
Inputting the acquired brain image and the acquired user information into a first machine learning model to obtain a region of interest extracted by the first machine learning model,
Characteristic information including the ratio of the first region and the second region included in the region of interest is input to a second machine learning model, and information about whether brain disease occurs or the degree of brain disease progression is obtained from the second machine learning model. Contains instructions for obtaining at least one,
The first region is associated with the first striatum located on the left or front of the brain image, and the second region is associated with the second striatum located on the right or rear of the brain image,
The first machine learning model determines the location and size of the region of interest associated with the dopaminergic nervous system in the entire area of the brain image, based on the user information, and selects the region of interest corresponding to the determined location and size. Extracted from the entire brain image area,
The second machine learning model is an information processing system that determines brain disease based on feature information including the ratio of the first area and the second area.

Images