WO2023229384A1 - 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 - Google Patents
학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 Download PDFInfo
- Publication number
- WO2023229384A1 WO2023229384A1 PCT/KR2023/007108 KR2023007108W WO2023229384A1 WO 2023229384 A1 WO2023229384 A1 WO 2023229384A1 KR 2023007108 W KR2023007108 W KR 2023007108W WO 2023229384 A1 WO2023229384 A1 WO 2023229384A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- noise
- generating
- learning
- input
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/60—Image enhancement or restoration using machine learning, e.g. neural networks
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/20—ICT specially adapted for the handling or processing of medical images for handling medical images, e.g. DICOM, HL7 or PACS
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H30/00—ICT specially adapted for the handling or processing of medical images
- G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Definitions
- the present disclosure relates to a method, computer program, and device for generating learning data, and specifically relates to a method, computer program, and device for generating learning data using medical images.
- a medical imaging device is a device that obtains a patient's body information and provides images.
- Medical imaging devices include X-ray imaging devices, ultrasound diagnostic devices, computed tomography devices, and magnetic resonance imaging (MRI) imaging devices.
- MRI magnetic resonance imaging
- Magnetic resonance imaging devices are equipment that requires considerable imaging time. Therefore, accelerated imaging technology to shorten the imaging time of magnetic resonance imaging is very important in the medical industry. In order for an accelerated magnetic resonance image to be used in the medical field, it must contain all information about the subject being imaged, while minimizing noise that affects the interpretation of the information. In response to this need, technology has recently been developed based on artificial intelligence to restore low-quality, accelerated MRI images to a high-quality, non-accelerated state.
- the present disclosure is intended to solve the problems of the prior art described above, and relates to a learning data generation method, computer program, and device for generating a learning image and a label image using only one input image .
- a method of generating learning data performed by a computing device includes generating random noise corresponding to noise of an input image, generating first noise and second noise based on the random noise, and generating first noise based on the input image and the first noise. Generating an image and generating a second image based on the input image and the second noise, wherein the first image and the second image have a noise-independent relationship with each other .
- generating the first noise and the second noise may include generating the first noise by applying a first coefficient to the random noise, and generating a second noise that is dependent on the first coefficient to the random noise. and generating the second noise by applying a coefficient.
- the first image is generated by applying the first noise to the input image
- the second image is generated by applying the second noise to the input image
- the second image further includes generating a learning image and a label image that are input to an artificial neural network model that outputs a high-quality medical image based on a low-quality medical image. It is characterized by
- generating the learning image and the label image may include determining weights corresponding to each of the first image and the second image based on a noise reduction target set in the artificial neural network model, and and generating the learning image and the label image by combining the first image and the second image based on a weight.
- the step of determining the weight determines the weight corresponding to each of the first image and the second image such that the sum of the weight of the first image and the weight of the second image is 1.
- the input image, the first image and the second image are each k-space data.
- generating k-space training data and k-space label data by combining the first image and the second image, and performing Fourier transform on the k-space training data and the k-space label data, respectively.
- the method further includes generating learning images and label images that are input to an artificial neural network model that outputs high-quality medical images based on low-quality medical images.
- the step of generating the random noise may include calculating the noise size based on a standard deviation of pixel values that are a certain distance away from the center of the input image and generating the random noise corresponding to the noise size. It includes steps to:
- the random noise includes noise that follows a complex Gaussian distribution.
- the input image, the first image and the second image are each magnetic resonance images.
- generating the random noise may include distinguishing a background of the input image, calculating the noise size based on a standard deviation of pixel values of the background, and the random noise corresponding to the noise size. It includes the step of generating.
- the random noise includes noise following a Rician Distribution or Noncentral chi distribution.
- a learning data generating device is disclosed according to an embodiment of the present disclosure for realizing the above-described problem.
- the device generates a memory for storing an input image and random noise corresponding to the noise of the input image, and generates first noise and second noise based on the random noise, and generates the input image and the first noise. and a processor that generates a first image based on and generates a second image based on the input image and the second noise, wherein the first image and the second image are noise independent of each other. .
- the processor generates the first noise and the second noise by applying a first coefficient and a second coefficient to the random noise, respectively, and applies the first noise and the second noise to the input image.
- the first image and the second image are generated by applying each image, and the first coefficient and the second coefficient are dependent on each other.
- the processor generates a training image and a label image for learning the artificial neural network model by combining the first image and the second image based on a noise reduction goal set in the artificial neural network model, and
- the artificial neural network model is characterized by being learned to output high-quality medical images based on low-quality medical images.
- a computer program stored in a computer-readable storage medium When the computer program is executed on one or more processors, it performs the following operations. At this time, the operations include generating random noise corresponding to the noise of the input image, generating first noise and second noise based on the random noise, and generating random noise based on the input image and the first noise. Generating a first image and generating a second image based on the input image and the second noise, wherein the first image and the second image are independent of noise.
- the present disclosure can generate a learning image and a label image using one input image, and can train an artificial neural network even without a high-quality label image.
- the present disclosure can generate learning images and label images based on the noise reduction goal of the artificial neural network model, so that the noise reduction goal of the artificial neural network model can be adjusted quantitatively and in detail. there is.
- FIG. 1 is a block diagram of a computing device according to an embodiment of the present disclosure.
- Figure 2 is a block diagram of an artificial neural network model learning system according to an embodiment of the present disclosure.
- Figure 3 is a flowchart showing a method of operating a learning data generating device according to an embodiment of the present disclosure.
- Figure 4 is a flowchart showing a method of operating a learning data generating device according to an embodiment of the present disclosure.
- Figure 5 is a flowchart showing a method of operating an artificial neural network model according to an embodiment of the present disclosure.
- Figure 6 is a block diagram of an artificial neural network model learning system according to an embodiment of the present disclosure.
- the term “or” is intended to mean an inclusive “or” and not an exclusive “or.” That is, unless otherwise specified in the present disclosure or the meaning is not clear from the context, “X uses A or B” should be understood to mean one of natural implicit substitutions. For example, unless otherwise specified in the present disclosure or the meaning is not clear from context, “X uses A or B” means that It can be interpreted as one of the cases where all B is used.
- N is a natural number
- N is a natural number
- components performing different functional roles may be distinguished as first components or second components.
- components that are substantially the same within the technical spirit of the present disclosure but must be distinguished for convenience of explanation may also be distinguished as first components or second components.
- module refers to a computer-related entity, firmware, software or part thereof, hardware or part thereof.
- the “module” or “unit” can be understood as a term referring to an independent functional unit that processes computing resources, such as a combination of software and hardware.
- the “module” or “unit” may be a unit composed of a single element, or may be a unit expressed as a combination or set of multiple elements.
- a “module” or “part” in the narrow sense is a hardware element or set of components of a computing device, an application program that performs a specific function of software, a process implemented through the execution of software, or a program. It can refer to a set of instructions for execution, etc.
- module or “unit” may refer to the computing device itself constituting the system, or an application running on the computing device.
- module or “unit” may be defined in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.
- model refers to a system implemented using mathematical concepts and language to solve a specific problem, a set of software units to solve a specific problem, or a process to solve a specific problem. It can be understood as an abstract model of a process.
- a neural network “model” may refer to an overall system implemented as a neural network that has problem-solving capabilities through learning. At this time, the neural network can have problem-solving capabilities by optimizing parameters connecting nodes or neurons through learning.
- a neural network “model” may include a single neural network or a neural network set in which multiple neural networks are combined.
- image refers to multi-dimensional data consisting of discrete image elements (e.g., pixels in a two-dimensional image and voxels in a three-dimensional image). can do.
- the image may include, but is not limited to, a medical image acquired by a medical imaging device such as a magnetic resonance imaging device, a computed tomography (CT) device, an ultrasonic imaging device, or an X-ray imaging device.
- a medical imaging device such as a magnetic resonance imaging device, a computed tomography (CT) device, an ultrasonic imaging device, or an X-ray imaging device.
- medical image used in this disclosure is a concept that collectively refers to all forms of images encompassing medical knowledge and includes various modalities such as visible light cameras, IR cameras, ultrasound, X-ray, CT, MRI, and PET. It may include images acquired through .
- the term “medical image archiving and communication system” refers to the storage, processing, and processing of medical images in accordance with the DICOM (digital imaging and communications in medicine) standard. It can refer to a transmitting system.
- the “medical image storage and transmission system” is linked with digital medical imaging equipment to produce medical images such as magnetic resonance imaging (MRI) and computed tomography (CT) images, and other digital medical images. It can be saved according to communication standards.
- the “medical image storage and transmission system” can transmit medical images to terminals inside and outside the hospital through a communication network. At this time, meta information such as reading results and medical records may be added to the medical image.
- object used in the present disclosure refers to a subject of photography and may include a person, an animal, or a part thereof.
- the object may include a part of the body (organ, etc.) or a phantom.
- a phantom refers to a substance with a volume very close to the density and effective atomic number of a living organism, and may include a spherical phantom with properties similar to the body.
- the Magnetic Resonance Image (MRI) system expresses the strength of the Magnetic Resonance (MR) signal in contrast to the RF (Radio Frequency) signal generated in a magnetic field of a certain strength, providing information on the tomographic area of the object. It is a system for acquiring images.
- MR Magnetic Resonance
- RF Radio Frequency
- the MRI system causes the main magnet to form a static magnetic field and aligns the magnetic dipole moment direction of a specific atomic nucleus of an object located in the static magnetic field in the direction of the static magnetic field.
- the gradient magnetic field coil can apply a gradient signal to a static field to form a gradient magnetic field and induce different resonance frequencies for each part of the object.
- the RF coil can irradiate magnetic resonance signals according to the resonance frequency of the area for which images are desired to be acquired. Additionally, as a gradient magnetic field is formed, the RF coil can receive magnetic resonance signals of different resonance frequencies radiated from various parts of the object.
- the MRI system acquires images by applying image restoration techniques to the magnetic resonance signals received through these steps. Additionally, the MRI system may perform serial or parallel signal processing on a plurality of magnetic resonance signals received by a multi-channel RF coil to reconstruct the plurality of magnetic resonance signals into image data.
- FIG. 1 is a block diagram of a computing device according to an embodiment of the present disclosure.
- the computing device 100 may be a hardware device or part of a hardware device that performs comprehensive processing and calculation of data, or may be a software-based computing environment connected to a communication network.
- the computing device 100 may be a server that performs intensive data processing functions and shares resources, or it may be a client that shares resources through interaction with the server.
- the computing device 100 may be a cloud system that allows a plurality of servers and clients to interact and comprehensively process data. Since the above description is only an example related to the type of computing device 100, the type of computing device 100 may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.
- a computing device 100 may include a processor 110, a memory 120, and a network unit 130. there is. However, since FIG. 1 is only an example, the computing device 100 may include other components for implementing a computing environment. Additionally, only some of the configurations disclosed above may be included in computing device 100.
- the processor 110 may be understood as a structural unit including hardware and/or software for performing computing operations.
- the processor 110 may read a computer program and perform data processing for machine learning.
- the processor 110 may process computational processes such as processing input data for machine learning, extracting features for machine learning, and calculating errors based on backpropagation.
- the processor 110 for performing such data processing includes a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), and a custom processing unit (TPU). It may include a semiconductor (ASIC: application specific integrated circuit), or a field programmable gate array (FPGA: field programmable gate array). Since the type of processor 110 described above is only an example, the type of processor 110 may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.
- the processor 110 may generate learning data input to an artificial neural network model.
- the artificial neural network model is trained with training data and can output low-quality images as high-quality images.
- an artificial neural network model can create high-quality images by removing noise from low-quality input images.
- An artificial neural network model may include at least one neural network.
- Neural networks may include network models such as Deep Neural Network (DNN), Recurrent Neural Network (RNN), Bidirectional Recurrent Deep Neural Network (BRDNN), Multilayer Perceptron (MLP), and Convolutional Neural Network (CNN). It is not limited to this.
- DNN Deep Neural Network
- RNN Recurrent Neural Network
- BPDNN Bidirectional Recurrent Deep Neural Network
- MLP Multilayer Perceptron
- CNN Convolutional Neural Network
- the image may be a medical image, for example a magnetic resonance image.
- the magnetic resonance image may have been acquired through accelerated imaging.
- Accelerated shooting can be understood as a shooting technique that shortens shooting time by increasing the acceleration factor compared to regular shooting.
- the acceleration index is a term used in parallel imaging techniques, and can be understood as the number of signal lines fully sampled in the k-space region divided by the number of signal lines sampled through imaging.
- the processor 110 may generate learning data based on one input image.
- the input image may be a medical image, magnetic resonance image, or k-space data.
- Training data may include a training image and a label image corresponding to the training image. Label images may be of higher quality than training images.
- the learning image and the label image may be independent of noise.
- the learning image and the label image may have a noise-independent relationship by a preset value, and may have a noise-dependent relationship by a preset value.
- the artificial neural network model learned with the training image and label image can be trained to remove noise by a preset value.
- the processor 110 may generate two images that are noise-independent from each other using one input image and generate learning data using the two images. Therefore, an artificial neural network can be trained without high-quality labeled images. Additionally, the noise reduction goal of the artificial neural network model can be adjusted in detail by setting the degree of noise independence or noise dependence of the two images.
- the processor 110 may train an artificial neural network model using a training image and a label image generated based on one input image.
- the processor 110 may learn an artificial neural network model to improve image quality.
- the image may be, for example, a magnetic resonance image, or may have been acquired by accelerated imaging.
- Accelerated shooting can be understood as a shooting technique that shortens shooting time by increasing the acceleration factor compared to regular shooting.
- the acceleration index is a term used in parallel imaging techniques, and can be understood as the number of signal lines fully sampled in the K-space region divided by the number of signal lines sampled through imaging.
- an acceleration index of 2 can be understood as acquiring half the number of signal lines compared to the number of fully sampled signal lines when acquiring a line by sampling a magnetic resonance signal in the phase encoding direction. Therefore, as the acceleration index increases, the capturing time of the magnetic resonance image may decrease proportionally. That is, if the acceleration index is increased when taking a magnetic resonance image, accelerated imaging with a shortened magnetic resonance image taking time can be implemented.
- the processor 110 may create a user interface that provides an environment for interaction with a user of the computing device 100 or a user of an arbitrary client.
- the processor 110 may create a user interface for receiving an input image and a noise reduction goal of an artificial neural network model.
- the processor 110 may create a user interface that allows functions such as output, modification, change, or addition of data to be implemented based on an external input signal applied from the user. Since the role of the user interface described above is only an example, the role of the user interface may be defined in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.
- the memory 120 may be understood as a structural unit including hardware and/or software for storing and managing data processed in the computing device 100. That is, the memory 120 can store any type of data generated or determined by the processor 110 and any type of data received by the network unit 130.
- the memory 120 may be a flash memory type, hard disk type, multimedia card micro type, card type memory, or random access memory (RAM). ), SRAM (static random access memory), ROM (read-only memory), EEPROM (electrically erasable programmable read-only memory), PROM (programmable read-only memory), magnetic memory , a magnetic disk, or an optical disk may include at least one type of storage medium.
- the memory 120 may include a database system that controls and manages data in a predetermined system. Since the type of memory 120 described above is only an example, the type of memory 120 may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.
- the memory 120 can structure, organize, and manage data necessary for the processor 110 to perform operations, combinations of data, and program codes executable on the processor 110.
- the memory 120 may store an input image received through the network unit 130, which will be described later. Additionally, the memory 120 may store learning images and label images generated from input images by the processor 110. Additionally, the memory 120 may store program code that operates the processor 110 to generate learning data.
- the network unit 130 may be understood as a structural unit that transmits and receives data through any type of known wired or wireless communication system.
- the network unit 130 is a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), and WiBro (wireless).
- broadband internet 5th generation mobile communication (5G), ultra wide-band wireless communication, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity ), data transmission and reception can be performed using a wired or wireless communication system such as near field communication (NFC), or Bluetooth. Since the above-described communication systems are only examples, the wired and wireless communication systems for data transmission and reception of the network unit 130 may be applied in various ways other than the above-described examples.
- the network unit 130 may receive data necessary for the processor 110 to perform calculations through wired or wireless communication with any system or client. Additionally, the network unit 130 may transmit data generated through the calculation of the processor 110 through wired or wireless communication with any system or any client.
- the network unit 130 can receive medical images through communication with a medical image storage and transmission system, a cloud server that performs tasks such as improving the resolution of medical images and standardization, or the computing device 100. there is.
- the network unit 130 may transmit the generated learning image and label image through communication with the above-described system, server, or computing device 100.
- the network unit 130 may provide the user interface created by the processor 110 to any system or client through wired or wireless communication with any system or client.
- the network unit 130 provides a user interface for visualizing data processed by the processor 110 through communication with a medical image storage and transmission system or a computing device 100 including a display. It can be provided to a system or device.
- the network unit 130 may receive an external input signal applied from a user through a user interface from the above-described system or device and transmit it to the processor 110.
- the processor 110 may operate to implement functions such as output, modification, change, or addition of data based on the external input signal transmitted from the network unit 130.
- a system or device provided with a user interface may operate on its own to implement functions such as output, modification, change, or addition of data according to an external input signal applied from the user. there is.
- Figure 2 is a block diagram of an artificial neural network model learning system according to an embodiment of the present disclosure.
- the artificial neural network model learning system 1 includes a learning data generating device 10 and an artificial neural network model 500.
- the learning data generating device 10 receives one input image 200 and generates learning data including a learning image 410 and a label image 420.
- the artificial neural network model 500 is trained to output high-quality images from low-quality images using learning data.
- the learning data generating device 10 can measure the noise size of the input image 200.
- the noise size may mean the standard deviation of pixel values included in the input image 200.
- the background of the input image 200 is segmented, and this refers to the standard deviation of pixel values of the segmented background.
- the learning data generating device 10 may generate random noise having the same size as the measured noise size. Random noise may mean noise that follows a complex Gaussian distribution, Rician distribution, or noncentral chi distribution. Additionally, the random noise may be an image that has the same size as the input image. At this time, noise for each pixel of random noise, which is in the form of an image, can be generated independently.
- the noise size can be set for each pixel of the input image 200. Additionally, it can be set to be proportional to the structure factor (g-factor) value.
- the learning data generating device 10 generates first noise by multiplying random noise by a first coefficient, and generates second noise by multiplying random noise by a second coefficient. At this time, when one of the first coefficient and the second coefficient is determined, the other value can be determined.
- the learning data generating device 10 generates a first image 310 based on the input image 200 and the first noise, and generates a second image 320 based on the input image 200 and the second noise. do. For example, the learning data generating device 10 generates a first image 310 by adding first noise to the input image 200, and creates a second image 320 by adding second noise to the input image 200. ) is created.
- the first coefficient and the second coefficient are determined so that the first image and the second image have a noise-independent relationship with each other. That is, the first coefficient and the second coefficient are determined so that the noise of the input image 200 plus the first noise and the noise of the input image 200 plus the second noise have a noise-independent relationship.
- the learning data generating device 10 generates a learning image 410 and a label image 420 by combining the first image 310 and the second image 320. Combining methods may vary. For example, the learning data generating device 10 may output the first image 310 and the second image 320, respectively, based on the noise reduction goal set in the artificial neural network model 500. Determine the corresponding weight. For example, the weight means a coefficient of a linear combination of the first image 310 and the second image 320.
- the learning image 410 and the label image 420 may be generated according to Equation 1, for example.
- Equation 1 T refers to the learning image 410, L refers to the label image 420, X refers to the first image 310, and Y refers to the second image 320.
- a means training the artificial neural network model 500 to achieve a noise reduction goal, for example, to reduce noise by a times or to increase the signal-to-noise ratio (SNR) by a times.
- SNR signal-to-noise ratio
- the signal sizes of the learning image 410 and the label image 420 are the same, and the label image 420 includes noise that is dependent on the learning image 410 by 1/a. Since the artificial neural network model 500 does not learn independent noise, it is learned to increase the SNR by a times. For example, if a is infinite, the training image 410 is the first image 310 and the label image 420 is the second image 320. At this time, since the first image 310 and the second image 320 are noise-independent, the artificial neural network model 500 is trained to increase the SNR infinitely. Meanwhile, Equation 1 is an example, and the method of generating the learning image 410 and the label image 420 by combining the first image 310 and the second image 320 is not limited thereto.
- the learning data generating device 10 may provide learning data including the generated learning image 410 and label image 420 to the artificial neural network model 500.
- the configuration shown in FIG. 1 is an example, and the learning data generating device 10 and the artificial neural network model 500 may be included in different systems, where the learning data generating device 10 and the artificial neural network model ( 500) can transmit and receive learning data through the network.
- Figure 3 is a flowchart showing a method of operating a learning data generating device according to an embodiment of the present disclosure.
- the learning data generating device may be an embodiment of the learning data generating device 10 of FIG. 1 or the learning data generating device 11 of FIG. 6.
- the learning data generating device generates random noise corresponding to the noise of the input image (S110). Specifically, random noise having the same size as the noise size of the input image is generated.
- the learning data generating device calculates the noise size based on the standard deviation of pixel values that are a certain distance away from the center of the input image. Then, it generates noise that has the calculated size and follows a complex Gaussian distribution.
- the learning data generating device segments the background of the input image and calculates the noise size based on the standard deviation of pixel values of the background. Then, noise is generated that follows the Rician Distribution or Noncentral Chi distribution with the calculated size.
- the learning data generating device generates first noise and second noise based on random noise (S120).
- the first noise is generated by applying the first coefficient to random noise
- the second noise is generated by applying the second coefficient to random noise
- the second coefficient is based on a preset value and the first coefficient. It is decided.
- the first coefficient and the second coefficient are determined so that the noise of the input image plus the first noise and the noise of the input image plus the second noise have a noise-independent relationship with each other. That is, the first coefficient and the second coefficient are dependent on each other.
- the learning data generating device generates first noise by multiplying random noise by a first coefficient, and generates second noise by multiplying random noise by a second coefficient.
- the learning data generating device generates a first image based on the input image and the first noise (S130) and generates a second image based on the input image and the second noise (S140).
- the order of steps S130 and S140 is not limited thereto and may be performed simultaneously.
- the first image is created by adding first noise to the input image
- the second image is created by adding second noise to the input image.
- the first image and the second image are noise independent of each other.
- the learning data generating device generates a learning image and a label image that are input to the artificial neural network model based on the first image and the second image (S150).
- preprocessing of the first image and the second image may be performed. For example, an operation of converting the domains of the first image and the second image may be performed.
- step S150 is not necessarily performed and can be omitted.
- Figure 4 is a flowchart showing a method of operating a learning data generating device according to an embodiment of the present disclosure.
- the learning data generating device may be an embodiment of the learning data generating device 10 of FIG. 1 or the learning data generating device 11 of FIG. 6.
- the learning data generating device can set a noise reduction goal for the artificial neural network model (S210).
- the learning data generating device determines weights corresponding to the first image and the second image based on the noise reduction goal (S220). For example, the learning data generating device may ensure that the sum of the weight of the first image and the weight of the second image is 1.
- the learning data generating device generates a learning image and a label image by combining the first image and the second image based on the weight (S230).
- Figure 5 is a flowchart showing a method of operating an artificial neural network model according to an embodiment of the present disclosure.
- the artificial neural network model may be an example of the artificial neural network model 500 of FIG. 1 or FIG. 6 .
- the artificial neural network model receives learning images and label images each generated based on medical images (S310).
- the learning image and the label image are generated by combining the first image and the second image generated based on the medical image, and the first image and the second image are noise independent of each other.
- the first image is generated based on the medical image and first noise
- the second image is generated based on the medical image and second noise.
- the first noise and the second noise are generated based on random noise of the same size as the noise size of the medical image.
- the artificial neural network model is trained to output high-quality medical images based on low-quality medical images using learning images and label images (S320).
- Figure 6 is a block diagram of an artificial neural network model learning system according to an embodiment of the present disclosure.
- the artificial neural network model learning system 2 is similar to the artificial neural network model learning system 1 of FIG. 2, so common points are omitted.
- the learning image 410 and the label image 420 may be magnetic resonance images in the image domain.
- the input image 600 may be an image in the k-space domain, that is, k-space data.
- the learning data generating device 11 measures the noise size of the input image 600.
- the noise size refers to the standard deviation of pixel values that are a certain distance away from the center of the input image 600.
- the learning data generating device 11 may generate random noise having the same size as the measured noise size. Random noise may mean noise that follows a complex Gaussian distribution.
- the learning data generating device 11 generates first noise by multiplying random noise by a first coefficient, and generates second noise by multiplying random noise by a second coefficient.
- the learning data generating device 11 generates a first image 710 based on the input image 600 and the first noise, and generates a second image 720 based on the input image 600 and the second noise. do.
- the first image 710 and the second image 720 are independent of noise.
- the learning data generating device 11 generates k-space learning data 810 and k-space label data 820 based on the first image 710 and the second image 720. Since this is similar to the method described above in Equation 1 of FIG. 2 and FIG. 4, the following description will be omitted.
- the learning data generation device 11 performs Fourier transformation on the k-space learning data 810 and the k-space label data 820, respectively, to generate a learning image 410 and a label image 420, and generates an artificial neural network model ( 500).
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Biophysics (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Pathology (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
- High Energy & Nuclear Physics (AREA)
- Heart & Thoracic Surgery (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
Abstract
Description
Claims (17)
- 적어도 하나의 프로세서를 포함하는 컴퓨팅 장치에 의해 수행되는, 학습 데이터 생성 방법으로서,입력 이미지의 노이즈와 대응되는 랜덤 노이즈를 생성하는 단계;상기 랜덤 노이즈를 기초로, 제1 노이즈 및 제2 노이즈를 생성하는 단계; 및상기 입력 이미지 및 상기 제1 노이즈를 기초로 제1 이미지를 생성하고, 상기 입력 이미지 및 상기 제2 노이즈를 기초로 제2 이미지를 생성하는 단계;를 포함하고,상기 제1 이미지와 상기 제2 이미지는 서로 노이즈 독립 관계인 것을 특징으로 하는 방법.
- 제1항에 있어서,상기 제1 노이즈 및 상기 제2 노이즈를 생성하는 단계는,상기 랜덤 노이즈에 제1 계수를 적용하여 상기 제1 노이즈를 생성하고, 상기 랜덤 노이즈에 상기 제1 계수와 종속 관계인 제2 계수를 적용하여 상기 제2 노이즈를 생성하는 단계;를 포함하는 방법.
- 제1항에 있어서,상기 제1 이미지는상기 입력 이미지에 상기 제1 노이즈를 적용하여 생성된 것이고,상기 제2 이미지는상기 입력 이미지에 상기 제2 노이즈를 적용하여 생성된 것인 방법.
- 제1항에 있어서,상기 제1 이미지 및 상기 제2 이미지를 기초로, 저품질의 의료 이미지를 기초로 고품질의 의료 이미지를 출력하는 인공 신경망 모델에 입력되는 학습 이미지 및 라벨 이미지를 생성하는 단계;를 더 포함하는 것을 특징으로 하는 방법.
- 제4항에 있어서,상기 학습 이미지 및 상기 라벨 이미지를 생성하는 단계는,상기 인공 신경망 모델에 설정된 노이즈 감소 목표를 기초로, 상기 제1 이미지 및 상기 제2 이미지에 각각 대응되는 가중치를 결정하는 단계; 및상기 가중치를 기초로 상기 제1 이미지 및 상기 제2 이미지를 조합하여 상기 학습 이미지 및 상기 라벨 이미지를 생성하는 단계;를 포함하는 방법.
- 제5항에 있어서,상기 가중치를 결정하는 단계는,상기 제1 이미지의 가중치와 상기 제2 이미지의 가중치의 합이 1이 되도록 상기 제1 이미지 및 상기 제2 이미지에 각각 대응되는 가중치를 결정하는 방법.
- 제1항에 있어서,상기 입력 이미지, 상기 제1 이미지 및 상기 제2 이미지는 각각 k-스페이스 데이터인 방법.
- 제7항에 있어서,상기 제1 이미지 및 상기 제2 이미지를 조합하여 k-스페이스 학습 데이터 및 k-스페이스 라벨 데이터를 생성하는 단계; 및상기 k-스페이스 학습 데이터 및 상기 k-스페이스 라벨 데이터를 각각 푸리에 변환하여 저품질의 의료 이미지를 기초로 고품질의 의료 이미지를 출력하는 인공 신경망 모델에 입력되는 학습 이미지 및 라벨 이미지를 생성하는 단계;를 더 포함하는 것을 특징으로 하는 방법.
- 제7항에 있어서,상기 랜덤 노이즈를 생성하는 단계는,상기 입력 이미지의 중심으로부터 일정 거리만큼 떨어진 픽셀값들의 표준 편차를 기초로 상기 노이즈 크기를 계산하는 단계; 및상기 노이즈 크기에 대응되는 상기 랜덤 노이즈를 생성하는 단계;를 포함하는 방법.
- 제7항에 있어서,상기 랜덤 노이즈는 복소 가우시안 분포를 따르는 노이즈를 포함하는 방법.
- 제1항에 있어서,상기 입력 이미지, 상기 제1 이미지 및 상기 제2 이미지는 각각 자기 공명 영상인 방법.
- 제11항에 있어서,상기 랜덤 노이즈를 생성하는 단계는,상기 입력 이미지의 배경을 구분하고, 상기 배경의 픽셀값들의 표준 편차를 기초로 상기 노이즈 크기를 계산하는 단계; 및상기 노이즈 크기에 대응되는 상기 랜덤 노이즈를 생성하는 단계;를 포함하는 방법.
- 제11항에 있어서,상기 랜덤 노이즈는 라이시안 분포(Rician Distribution) 또는 비중심 카이 분포(Noncentral chi distribution)를 따르는 노이즈를 포함하는 방법.
- 학습 데이터 생성 장치로서,입력 이미지를 저장하는 메모리; 및상기 입력 이미지의 노이즈와 대응되는 랜덤 노이즈를 생성하고, 상기 랜덤 노이즈를 기초로, 제1 노이즈 및 제2 노이즈를 생성하고, 상기 입력 이미지 및 상기 제1 노이즈를 기초로 제1 이미지를 생성하고, 상기 입력 이미지 및 상기 제2 노이즈를 기초로 제2 이미지를 생성하는 프로세서;를 포함하고,상기 제1 이미지와 상기 제2 이미지는 서로 노이즈 독립 관계인 것을 특징으로 하는 장치.
- 제14항에 있어서,상기 프로세서는,상기 랜덤 노이즈에 제1 계수 및 제2 계수를 각각 적용하여 상기 제1 노이즈 및 상기 제2 노이즈를 생성하고, 상기 입력 이미지에 상기 제1 노이즈 및 상기 제2 노이즈를 각각 적용하여 상기 제1 이미지 및 상기 제2 이미지를 생성하고,상기 제1 계수와 상기 제2 계수는 서로 종속 관계인 것을 특징으로 하는 장치.
- 제14항에 있어서,상기 프로세서는,인공 신경망 모델에 설정된 노이즈 감소 목표를 기초로, 상기 제1 이미지 및 상기 제2 이미지를 조합하여 상기 인공 신경망 모델을 학습하기 위한 학습 이미지 및 라벨 이미지를 생성하고,상기 인공 신경망 모델은,저품질의 의료 이미지를 기초로 고품질의 의료 이미지를 출력하도록 학습된 것을 특징으로 하는 장치.
- 컴퓨터 판독가능 저장 매체 저장된 컴퓨터 프로그램(program)으로서, 상기 컴퓨터 프로그램은 하나 이상의 프로세서(processor)에서 실행되는 경우, 하기 위한 동작들을 수행하도록 하며,상기 동작들은,입력 이미지의 노이즈와 대응되는 랜덤 노이즈를 생성하는 동작;상기 랜덤 노이즈를 기초로, 제1 노이즈 및 제2 노이즈를 생성하는 동작; 및상기 입력 이미지 및 상기 제1 노이즈를 기초로 제1 이미지를 생성하고, 상기 입력 이미지 및 상기 제2 노이즈를 기초로 제2 이미지를 생성하는 동작을 포함하고,상기 제1 이미지와 상기 제2 이미지는 서로 노이즈 독립 관계인 것을 특징으로 하는 컴퓨터 프로그램.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| EP23812158.6A EP4503049A4 (en) | 2022-05-25 | 2023-05-24 | METHOD FOR GENERING TRAINING DATA, COMPUTER PROGRAM AND DEVICE |
| US18/721,768 US20250061546A1 (en) | 2022-05-25 | 2023-05-24 | Training data generation method, computer program and device |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020220064298A KR102513218B1 (ko) | 2022-05-25 | 2022-05-25 | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 |
| KR10-2022-0064298 | 2022-05-25 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2023229384A1 true WO2023229384A1 (ko) | 2023-11-30 |
Family
ID=86101679
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/KR2023/007108 Ceased WO2023229384A1 (ko) | 2022-05-25 | 2023-05-24 | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US20250061546A1 (ko) |
| EP (1) | EP4503049A4 (ko) |
| KR (1) | KR102513218B1 (ko) |
| WO (1) | WO2023229384A1 (ko) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102513218B1 (ko) * | 2022-05-25 | 2023-04-25 | 주식회사 에어스 메디컬 | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 |
| US12584985B1 (en) * | 2024-10-28 | 2026-03-24 | Airs Medical Inc. | Apparatus for obtaining magnetic resonance images base on deep learning model and method of controlling the same |
| CN120707431A (zh) * | 2025-08-29 | 2025-09-26 | 烟台大学 | 一种HumanNeRF图像降噪方法、系统和设备 |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102513218B1 (ko) * | 2022-05-25 | 2023-04-25 | 주식회사 에어스 메디컬 | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 |
Family Cites Families (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10803555B2 (en) * | 2017-08-31 | 2020-10-13 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a trained neural network model for scattering correction |
| US10592779B2 (en) * | 2017-12-21 | 2020-03-17 | International Business Machines Corporation | Generative adversarial network medical image generation for training of a classifier |
| US10430708B1 (en) * | 2018-08-17 | 2019-10-01 | Aivitae LLC | System and method for noise-based training of a prediction model |
| KR102097741B1 (ko) | 2019-07-25 | 2020-04-06 | 주식회사 딥노이드 | 인공지능 학습용 의료 영상 데이터 정제 시스템 및 그 구동방법 |
| CN111340699B (zh) * | 2020-02-17 | 2022-02-18 | 西北大学 | 基于非局部先验和稀疏表示的磁共振图像去噪方法及装置 |
| US11408954B2 (en) * | 2020-03-24 | 2022-08-09 | GE Precision Healthcare LLC | Systems and methods of reducing noise and artifacts in magnetic resonance imaging |
| US12228629B2 (en) * | 2020-10-07 | 2025-02-18 | Hyperfine Operations, Inc. | Deep learning methods for noise suppression in medical imaging |
| US12530745B2 (en) * | 2021-07-26 | 2026-01-20 | GE Precision Healthcare LLC | Systems and methods to reduce unstructured and structured noise in image data |
| US12118720B2 (en) * | 2021-12-17 | 2024-10-15 | GE Precision Healthcare LLC | Systems and methods of magnetic resonance image processing using neural networks having reduced dimensionality |
| US12287385B2 (en) * | 2022-04-22 | 2025-04-29 | GE Precision Healthcare LLC | Systems and methods of noise reduction in magnetic resonance images |
| KR102533000B1 (ko) * | 2022-08-18 | 2023-05-16 | 주식회사 에어스메디컬 | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 |
| US20240296527A1 (en) * | 2023-03-02 | 2024-09-05 | GE Precision Healthcare LLC | Reducing noise in ct images using synthetic data |
| US20240378726A1 (en) * | 2023-05-12 | 2024-11-14 | GE Precision Healthcare LLC | Deep learning based medical imaging system and method |
| KR20250042426A (ko) * | 2023-09-20 | 2025-03-27 | 삼성전자주식회사 | 뉴럴 디퓨전 기반의 영상 처리 방법 및 장치 |
| US20250217938A1 (en) * | 2023-12-29 | 2025-07-03 | Google Llc | Weighting Functions and Adaptive Noise Schedule for Training Noise-Based Machine-Learned Models |
| US12584985B1 (en) * | 2024-10-28 | 2026-03-24 | Airs Medical Inc. | Apparatus for obtaining magnetic resonance images base on deep learning model and method of controlling the same |
-
2022
- 2022-05-25 KR KR1020220064298A patent/KR102513218B1/ko active Active
-
2023
- 2023-05-24 WO PCT/KR2023/007108 patent/WO2023229384A1/ko not_active Ceased
- 2023-05-24 EP EP23812158.6A patent/EP4503049A4/en active Pending
- 2023-05-24 US US18/721,768 patent/US20250061546A1/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102513218B1 (ko) * | 2022-05-25 | 2023-04-25 | 주식회사 에어스 메디컬 | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 |
Non-Patent Citations (5)
| Title |
|---|
| DESAI ARJUN D, OZTURKLER BATU M, SANDINO CHRISTOPHER M, BOUTIN ROBERT, WILLIS MARC, VASANAWALA SHREYAS, HARGREAVES BRIAN A, RÉ CHR: "Noise2Recon: Enabling Joint MRI Reconstruction and Denoising with Semi-Supervised and Self-Supervised Learning", ARXIV (CORNELL UNIVERSITY), CORNELL UNIVERSITY LIBRARY, ARXIV.ORG, ITHACA, 7 October 2022 (2022-10-07), Ithaca, XP093111651, [retrieved on 20231213], DOI: 10.48550/arxiv.2110.00075 * |
| See also references of EP4503049A4 * |
| SONG HWANJUN, KIM MINSEOK, PARK DONGMIN, SHIN YOOJU, LEE JAE-GIL: "Learning From Noisy Labels With Deep Neural Networks: A Survey", IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, IEEE, USA, vol. 34, no. 11, 1 November 2023 (2023-11-01), USA, pages 8135 - 8153, XP093111654, ISSN: 2162-237X, DOI: 10.1109/TNNLS.2022.3152527 * |
| TIAN QIYUAN; LI ZIYU; FAN QIUYUN; POLIMENI JONATHAN R.; BILGIC BERKIN; SALAT DAVID H.; HUANG SUSIE Y.: "SDnDTI: Self-supervised deep learning-based denoising for diffusion tensor MRI", NEUROIMAGE, ELSEVIER, AMSTERDAM, NL, vol. 253, 1 March 2022 (2022-03-01), AMSTERDAM, NL , XP087022574, ISSN: 1053-8119, DOI: 10.1016/j.neuroimage.2022.119033 * |
| VARADARAJAN DIVYA; HALDAR JUSTIN P.: "A Majorize-Minimize Framework for Rician and Non-Central Chi MR Images", IEEE TRANSACTIONS ON MEDICAL IMAGING, IEEE, USA, vol. 34, no. 10, 1 October 2015 (2015-10-01), USA, pages 2191 - 2202, XP011670407, ISSN: 0278-0062, DOI: 10.1109/TMI.2015.2427157 * |
Also Published As
| Publication number | Publication date |
|---|---|
| EP4503049A1 (en) | 2025-02-05 |
| US20250061546A1 (en) | 2025-02-20 |
| EP4503049A4 (en) | 2026-03-11 |
| KR102513218B1 (ko) | 2023-04-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2023229384A1 (ko) | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 | |
| KR102622283B1 (ko) | 맥스웰 병렬 이미징 | |
| WO2024039043A1 (ko) | 학습 데이터 생성 방법, 컴퓨터 프로그램 및 장치 | |
| BR112021015798A2 (pt) | Determinação do parâmetro de modelo com o uso de um modelo preditivo | |
| KR102428725B1 (ko) | 영상 개선 방법 및 이를 수행하는 컴퓨터 프로그램 | |
| Chen et al. | Estimating tissue microstructure with undersampled diffusion data via graph convolutional neural networks | |
| WO2019124836A1 (ko) | 제1 의료 영상의 관심 영역을 제2 의료 영상 위에 맵핑하는 방법 및 이를 이용한 장치 | |
| WO2022216073A1 (ko) | 슬라이스 레졸루션 향상이 적용된 자기 공명 영상 처리 장치 및 그 방법 | |
| WO2023249372A1 (ko) | 딥러닝 모델의 학습을 위한 의료 데이터의 처리 방법, 프로그램 및 장치 | |
| CN112397195B (zh) | 用于生成体格检查模型的方法、装置、电子设备和介质 | |
| WO2023128074A1 (ko) | 스캔 파라미터를 이용하는 자기 공명 영상 처리 방법 및 장치 | |
| US11630177B2 (en) | Coil mixing error matrix and deep learning for prospective motion assessment | |
| KR102442591B1 (ko) | 라벨 생성 방법, 프로그램 및 장치 | |
| WO2023249402A1 (ko) | 딥러닝 모델의 학습을 위한 의료 데이터의 처리 방법, 프로그램 및 장치 | |
| WO2023177151A1 (ko) | 3차원 의료 데이터의 처리 방법, 프로그램 및 장치 | |
| WO2023249411A1 (ko) | 데이터 처리 방법, 컴퓨터 프로그램 및 장치 | |
| CN114972118B (zh) | 检查图像的降噪方法、装置、可读介质和电子设备 | |
| CN118629635A (zh) | 深静脉血栓溶栓疗效预测系统及方法 | |
| WO2024053828A1 (ko) | 인공 신경망 기반 해상도 향상 방법, 프로그램 및 장치 | |
| CN111599447B (zh) | 一种数据处理方法、装置、电子设备及存储介质 | |
| WO2024080488A1 (ko) | 혈관 검출 방법 및 이를 수행하는 컴퓨터 프로그램 | |
| WO2024010248A1 (ko) | 딥러닝 기반 의료 데이터의 품질 개선 방법, 프로그램 및 장치 | |
| De et al. | QROP: Quantum Learning‐Based Identification of Retinopathy of Prematurity | |
| KR102859458B1 (ko) | 딥 러닝에 모델에 기반한 자기 공명 영상을 획득하는 장치 및 그 제어 방법 | |
| WO2024080489A1 (ko) | 의료 이미지를 기반으로 하는 접촉 판단 방법 및 이를 수행하는 컴퓨터 프로그램 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 23812158 Country of ref document: EP Kind code of ref document: A1 |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 18721768 Country of ref document: US |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 23812158.6 Country of ref document: EP Ref document number: 2023812158 Country of ref document: EP |
|
| ENP | Entry into the national phase |
Ref document number: 2023812158 Country of ref document: EP Effective date: 20241028 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |