Method for predicting state of object body based on moving image data and computing device

Abstract

Disclosed is an apparatus for predicting a state of a subject on the basis of moving image data. The apparatus includes a memory and at least one processor performing communication with the memory. The processor may acquire initial moving image data corresponding to initial intervals up to a preset time point after the drug is injected into the training subject, and input initial moving image data corresponding to a first interval among the initial intervals. A first prediction model for predicting first image data representing anatomical information or blood flow information about a training subject corresponding to a previous point in time of the first interval is trained, and initial moving image data corresponding to a second interval among the initial intervals is input. And training a second prediction model that predicts second image data representing disease-specific information on the training subject corresponding to a reference point in time after the initial interval. Therefore, the user convenience can be improved.

Description

Technical Field

[0001] This disclosure relates to medical-related technologies. More specifically, this disclosure relates to a method for predicting the state of a diagnostic subject based on dynamic image data, and a computing device for performing the method. Background Technology

[0002] Positron emission tomography (PET) is a state-of-the-art nuclear medicine imaging technique that delivers positron-emitting radioactive isotopes to a subject and captures the emitted radiation to obtain useful diagnostic information about metabolic changes and receptor distribution in the body. Recently, it has evolved from simply acquiring simple PET images to hybrid scanners that combine computed tomography (CT) or magnetic resonance imaging (MRI) techniques.

[0003] Therefore, in recent PET examinations, anatomical information from CT images is obtained by using a PET / CT scanner that combines a PET device and a CT device into one, so as to provide accurate location and depth information of lesions confirmed from PET images.

[0004] Typically, PET scans are used to diagnose patients based on PET images acquired after a radioactively labeled tracer has been injected into the body, once the specific and non-specific binding of the tracer has reached a stable state or after a specific time (e.g., 1 hour 30 minutes to 3 hours) has elapsed and the difference between them has reached its maximum.

[0005] However, PET image-based diagnosis is performed at a relatively late time point after tracer injection, so a method to improve this situation is needed. Summary of the Invention

[0006] Technical issues One of the technical problems to be solved by this disclosure is to provide a method for predicting the state of a diagnostic object based on dynamic image data.

[0007] The problems to be solved by this invention are not limited to the technical problems mentioned above. Those skilled in the art to which this invention pertains can clearly understand other technical problems not mentioned through the following description.

[0008] Technical solution A method for predicting the state of a subject based on dynamic image data, executed by at least one processor according to an embodiment of the present invention, includes the following steps: acquiring initial dynamic image data corresponding to an initial interval from when a drug is injected into a training subject to a predetermined time point; using the initial dynamic image data corresponding to a first interval in the initial interval as input to train a first prediction model that predicts first image data representing anatomical or blood flow information of the training subject corresponding to a previous time point in the first interval; and using the initial dynamic image data corresponding to a second interval in the initial interval as input to train a second prediction model that predicts second image data representing disease-specific information of the training subject corresponding to a reference time point after the initial interval.

[0009] The method may further include the following steps: acquiring initial dynamic image data corresponding to the first interval after the drug is injected into the diagnostic subject; and inputting the initial dynamic image data of the diagnostic subject corresponding to the first interval into the trained first prediction model to predict first image data representing anatomical or blood flow information of the diagnostic subject corresponding to the previous time point.

[0010] The method may further include the following steps: acquiring initial dynamic image data corresponding to the second interval after the drug is injected into the diagnostic subject; and inputting the initial dynamic image data of the diagnostic subject corresponding to the second interval into a trained second prediction model to predict second image data representing disease-specific information about the diagnostic subject corresponding to the reference time point.

[0011] When the drug, in reduced, pre-set amounts, is injected into the subject of diagnosis, the processor can acquire and process initial dynamic image data corresponding to the first interval and initial dynamic image data corresponding to the second interval, to acquire initial dynamic image data corresponding to the processed first interval and initial dynamic image data corresponding to the processed second interval, input the initial dynamic image data corresponding to the processed first interval into the first prediction model, and input the initial dynamic image data corresponding to the processed second interval into the second prediction model.

[0012] The amount of drug injected into the training subject can be a reduced amount than a reference amount. The first prediction model can be trained based on first labeled image data corresponding to the first image data, and the second prediction model can be trained based on second labeled image data corresponding to the second image data. The first and second labeled image data can be labeled image data processed according to the amount of drug injected into the training subject.

[0013] The method may further include: performing spatial normalization before inputting the initial dynamic image data of the training subjects and the diagnostic subjects corresponding to the first interval into the first prediction model, so as to achieve machine learning with less training subject data; and performing spatial normalization before inputting the initial dynamic image data of the training subjects and the diagnostic subjects corresponding to the second interval into the second prediction model.

[0014] The processor can set the acquisition time interval of the initial dynamic image data of the training object and the diagnostic object corresponding to the first interval in the initial interval, and can also set the acquisition time interval of the initial dynamic image data of the training object and the diagnostic object corresponding to the second interval in the initial interval.

[0015] The processor can identify the initial interval when the drug is injected into the diagnostic subject, and acquire initial dynamic image data corresponding to the first interval and the second interval based on the identified initial interval.

[0016] The acquired initial dynamic image data, the first image data, and the second image data may be positron emission tomography (PET) image data.

[0017] An object state prediction device based on dynamic image data according to an embodiment of the present invention may include: a memory; and at least one processor communicating with the memory.

[0018] The processor can acquire initial dynamic image data corresponding to an initial interval up to a predetermined time point after the drug is injected into the training subject, and use the initial dynamic image data corresponding to a first interval in the initial interval as input to train a first prediction model that predicts first image data representing anatomical or blood flow information of the training subject corresponding to a previous time point in the first interval, and use the initial dynamic image data corresponding to a second interval in the initial interval as input to train a second prediction model that predicts second image data representing disease-specific information of the training subject corresponding to a reference time point after the initial interval.

[0019] The processor can acquire initial dynamic image data corresponding to the first interval after the drug is injected into the diagnostic subject, and input the initial dynamic image data of the diagnostic subject corresponding to the first interval into the trained first prediction model to predict first image data representing anatomical information or blood flow information of the diagnostic subject corresponding to the previous time point.

[0020] The processor can acquire initial dynamic image data corresponding to the second interval after the drug is injected into the diagnostic subject, and input the initial dynamic image data of the diagnostic subject corresponding to the second interval into the trained second prediction model to predict second image data representing disease-specific information about the diagnostic subject corresponding to the reference time point.

[0021] When the drug, in reduced, pre-set amounts, is injected into the subject of diagnosis, the processor can acquire and process initial dynamic image data corresponding to the first interval and initial dynamic image data corresponding to the second interval, to acquire initial dynamic image data corresponding to the processed first interval and initial dynamic image data corresponding to the processed second interval, input the initial dynamic image data corresponding to the processed first interval into the first prediction model, and input the initial dynamic image data corresponding to the processed second interval into the second prediction model.

[0022] The amount of drug injected into the training subject can be a reduced amount than a reference amount. The first prediction model can be trained based on first labeled image data corresponding to the first image data, and the second prediction model can be trained based on second labeled image data corresponding to the second image data. The first and second labeled image data can be labeled image data processed according to the amount of drug injected into the training subject.

[0023] The processor can perform spatial normalization before inputting the initial dynamic image data of the training object and the diagnostic object corresponding to the first interval into the first prediction model, and can also perform spatial normalization before inputting the initial dynamic image data of the training object and the diagnostic object corresponding to the second interval into the second prediction model.

[0024] Technical effect As a characteristic of nuclear medicine molecular imaging, waiting for the distribution of the image tracking agent (tracer) in the body after drug injection inevitably requires an intake time of up to several hours. However, according to various embodiments of this disclosure, initial blood flow image data and delayed intake image data can be generated simultaneously using only the initial dynamic image data after drug injection, and diagnostic waiting time can be significantly reduced, thereby greatly improving user convenience.

[0025] The effects of the present invention are not limited to those mentioned above. Those skilled in the art to which this invention pertains can clearly understand other effects not mentioned through the following description. Attached Figure Description

[0026] Figure 1 This is an overall schematic diagram illustrating a method for predicting the state of an object based on dynamic image data according to this disclosure.

[0027] Figure 2 This is a block diagram illustrating the configuration of an object state prediction device based on dynamic image data according to the present disclosure.

[0028] Figure 3 This is a flowchart illustrating an object state prediction method based on dynamic image data according to the present disclosure.

[0029] Figure 4 This is a diagram illustrating the process of learning a first prediction model and a second prediction model based on initial dynamic image data according to this disclosure.

[0030] Figure 5 This is a flowchart illustrating a method for predicting first image data representing anatomical or blood flow information of a diagnostic subject and second image data representing disease-specific information, according to the present disclosure.

[0031] Figure 6 This is a flowchart illustrating the preprocessing procedure for execution space normalization according to this disclosure.

[0032] Figure 7This is a diagram illustrating the spatial normalization performed in the preprocessing step prior to training an artificial neural network model according to this disclosure.

[0033] Figure 8 This is a graph showing the change in radiation dose over time after injection of a drug according to this disclosure.

[0034] Figure 9a It is a graph used to compare delayed image data generated with reference line measurements according to this disclosure with labeled image data. Figure 9b It is a graph used to compare delayed image data generated with label image data according to this disclosure using low line count. Figure 9c It is a chart used to compare delayed image data generated with reference line count and delayed image data generated with low line count according to this disclosure.

[0035] Figure 10a The simulated PET data used in the training and testing steps for the anterior putamen (AP) and posterior putamen (PP) according to this disclosure are shown. Figure 10b The digital PET data used in the steps of testing AP and PP respectively, according to this disclosure, are shown.

[0036] Figure 11 This is a diagram illustrating a method for acquiring extremely initial image data according to the present disclosure.

[0037] Figure 12 This is a diagram illustrating a method for acquiring delayed image data according to the present disclosure. Detailed Implementation

[0038] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The advantages and features of the present invention, as well as methods of implementing them, will become apparent from the detailed embodiments described below in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but can be implemented in many different forms. These embodiments are provided only to complete the disclosure of the present invention and to fully inform those skilled in the art of the scope of the invention, which is defined only by the scope of the claims. Throughout this specification, the same reference numerals refer to the same constituent elements.

[0039] Unless otherwise defined, all terms used in this specification (including technical and scientific terms) are to be used in the sense that can be commonly understood by one of ordinary skill in the art to which this invention pertains. Furthermore, unless explicitly defined otherwise, terms as defined in commonly used dictionaries are not to be interpreted ideally or excessively.

[0040] The terminology used in this specification is for illustrative purposes and is not intended to limit the invention. In this specification, singular forms include plural forms unless specifically mentioned. The terms "comprises" and / or "comprising" as used in this specification do not exclude the presence or addition of more than one of the mentioned constituent elements.

[0041] In this specification, "computing device" includes all kinds of devices that perform computational processing. A "computing device" may include more than one computer. For example, a computer can be not only a desktop personal computer or a notebook computer, but also a smartphone, tablet computer, cellular phone, personal communication service phone (PCS phone), synchronous / asynchronous International Mobile Telecommunication-2000 (IMT-2000) mobile terminal, palm personal computer, personal digital assistant (PDA), etc. Furthermore, a computer can also be a medical device for acquiring or viewing medical images. Additionally, a computer can be a server computer connected to various client computers.

[0042] In this specification, "image data" refers to images acquired by a medical imaging device.

[0043] In this specification, "medical imaging device" refers to a machine used to acquire medical images. For example, "medical imaging device" may include a positron emission tomography (PET) imaging machine, a magnetic resonance imaging (MRI) machine, etc.

[0044] In this specification, "delayed image data" refers to images acquired after a reference time and used for diagnostic purposes in diagnosing patients.

[0045] In this specification, "reference time" refers to the time from the initial point in time when the drug (e.g., a contrast agent or tracer) is injected into the human body to the point in time when image data capable of diagnosing the patient's condition can be obtained (i.e., the reference time point).

[0046] In this specification, "medication" refers to a substance injected into the body during the acquisition of medical imaging data. For example, "medication" can be a contrast agent used in magnetic resonance imaging (MRI) or computed tomography (CT), a tracer used in positron emission tomography (PET), etc.

[0047] In this specification, "initial motion image data" refers to image data comprising multiple consecutive image frames. "Initial motion image data" is data acquired before a reference time point for acquiring delayed image data, and is acquired within an initial time range (e.g., a time range shortly after the administration of a contrast agent or tracer when inserting the image).

[0048] In this specification, dynamic image data can refer to image data from which initial dynamic image data and delayed image data have been removed.

[0049] Specifically, dynamic image data can refer to the playback interval images from the time when the blood flow effect begins to decrease to a predetermined reference time point after the time point when the drug is injected into the training subject body, which is included in multiple image data.

[0050] In this specification, anatomical information may be information related to the structure and function of the object, and may include image data containing blood flow information representing blood flow in the object.

[0051] In this specification, disease-specific information may refer to image data showing drug residues at a specific target location (e.g., an organ or tissue) that can be used to determine the condition of the subject.

[0052] Hereinafter, with reference to the accompanying drawings, a detailed description of a diagnostic image generation method and procedure based on initial dynamic image data according to an embodiment of the present invention is provided.

[0053] Figure 1 This is an overall schematic diagram illustrating a method for predicting the state of an object based on dynamic image data according to the present disclosure, which can be performed using an object state prediction device 100 (hereinafter "object state prediction device") based on dynamic image data.

[0054] The object state prediction device 100 can acquire initial dynamic image data 10. The initial dynamic image data 10 can correspond to an initial interval from the injection of the drug into the object via a tracer to a predetermined time point. The initial dynamic image data 10 may include multiple image frames, and each of the multiple image frames may be acquired at a period of 1 minute or 2 minutes, but this disclosure is not limited thereto.

[0055] Here, the preset time point can be a predetermined range based on the peak radiation level in the subject's bloodstream. Furthermore, the initial interval can be varied depending on the type of tracer and the matrix factors of the subject being diagnosed.

[0056] The object state prediction device 100 can input the acquired initial dynamic image data 10 into the first prediction model EM1 and the second prediction model EM2 (S11, S12).

[0057] In this case, the initial dynamic image data corresponding to the first interval in the initial dynamic image data 10 can be input into the first prediction model EM1, and the initial dynamic image data corresponding to the second interval in the initial dynamic image data 10 can be input into the second prediction model EM2. However, according to the embodiment, the object state prediction device 100 can input the same initial dynamic image data 10 into the first prediction model EM1 and the second prediction model EM2.

[0058] The first prediction model EM1 can predict TrueEarly image data 20 (S21) based on the input initial dynamic image data. The TrueEarly image data can be dynamic image data corresponding to a pre-set short period of time after drug injection or image data at a specific time point.

[0059] Furthermore, the second prediction model EM2 can predict delayed image data 30 (S21) based on the input initial dynamic image data. The delayed image data can be dynamic image data corresponding to a reference time point after the initial time period or image data at a specific time point.

[0060] Furthermore, the initial dynamic image data corresponding to the first interval, the initial dynamic image data corresponding to the second interval, the extreme initial image data, and the delayed image data can all be positron emission tomography (PET) image data, but this disclosure is not limited thereto.

[0061] Figure 2 This is a block diagram showing the configuration of the object state prediction device 100 according to the present disclosure.

[0062] Reference Figure 2The object state prediction device 100 may include a medical image capturing device or image data captured by the medical image capturing device, and may include a communication unit 110, an input unit 120, a display 130, a memory 150, and at least one processor 190. Figure 2 The constituent elements of the object state prediction device 100 shown are not essential for implementing the object state prediction device 100 according to the present disclosure. Therefore, the object state prediction device 100 described in this specification may have more constituent elements than the constituent elements listed above, or it may have fewer constituent elements.

[0063] The communication unit 110 in the constituent elements may include one or more constituent elements capable of communicating with various devices equipped with communication devices. For example, it may include at least one of the following: a satellite device, a wired communication device, a cellular wireless communication device, a wireless communication device based on IEEE 802.11 (e.g., it may also be called Wifi), a communication device based on short-range communication (e.g., it may be Bluetooth, Bluetooth Low Energy, UWB, Zifeng protocol, but not limited thereto), and a location information module.

[0064] The input unit 120 is used to input image information (or signals), audio information (or signals), data, or information input from the user, and may include, but is not limited to, at least one camera, a touch input device equipped on the touchscreen, and / or at least one microphone. Touch input, voice data, and / or image data collected by the input unit 120 can be analyzed and processed into user control commands. The input unit 120 may include a machine (inputter) for various types of input.

[0065] The camera processes image frames of still or moving images obtained by the image sensor in shooting mode. The processed image frames can be displayed on the display 130 (or the screen of the object state prediction device 100 of this disclosure) or stored in the memory 150.

[0066] The microphone processes external sound signals into electronic speech data. The processed speech data can be used in various ways depending on the functions (or applications) currently running on the device. Furthermore, the microphone can implement various noise removal algorithms to remove noise generated during the reception of external sound signals.

[0067] The output unit is used to generate outputs related to vision, hearing, or touch, and may include at least one of a display 130, at least one speaker, a haptic module, and an optical output device. The display 130 may be stacked with or integrally formed with a touch input device to realize a touch screen. Such a touch screen can perform output and / or input functions. The output device may include a machine (outputter) for various outputs.

[0068] The memory 150 may store at least one instruction that causes the object state prediction device 100 to perform various functions. The memory 150 may store data for content presentation (e.g., music files, still images, moving images, etc.). The memory 150 may store at least one application program, data for the operation of the object state prediction device 100, and commands that drive the object state prediction device 100 to perform operations according to various embodiments of the present disclosure. At least a portion of such an application may be downloaded wirelessly from an external server. For example, the object state prediction device 100 may download the application and store it in the memory 150. The object state prediction device 100 may perform operations according to various embodiments of the present disclosure by running the application. Alternatively, the object state prediction device 100 may temporarily download data from a server (e.g., for performing operations according to various embodiments of the present disclosure) and store the downloaded data in the memory 150.

[0069] This memory 150 can be a storage medium corresponding to at least one of the following types: flash memory type, hard disk type, solid state disk type, silicon disk drive type, multimedia card micro type, card-type memory (e.g., SD or XD memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, magnetic disk, and optical disk.

[0070] Those skilled in the art will understand that memory 150 may also refer to, for example, cache memory associated with processor 190 and / or cache memory and / or registers included in processor 190. Furthermore, memory 150 may be a database separate from object state prediction device 100 but connected to object state prediction device 100 in a wired or wireless manner, or it may be implemented as a database system.

[0071] Processor 190 may include more than one processor and may include at least one core. Processor 190 can execute instructions stored in memory 150. Processor 190 may be implemented by a memory storing data of a program for an algorithm or a reproduction algorithm for controlling the operation of constituent elements within object state prediction device 100, and at least one processor (not shown) performing the above operations using the data stored in the memory. In this case, the memory and processor may be implemented as separate chips. Alternatively, the memory and processor may be implemented as a single chip.

[0072] In this embodiment, the object state prediction device 100 can provide various UIs in the form of web services based on a platform, such as websites or web applications, but is not limited to these. Furthermore, the platform can be provided in the form of PC applications, mobile applications, etc., but this embodiment is not limited to these. In this case, various user terminals can utilize the various UIs provided by the object state prediction device 100 based on the platform.

[0073] The predefined operating rules or features of the artificial intelligence model are created through learning. Here, "created through learning" means that the basic artificial intelligence model learns from multiple learning data using a learning algorithm, thereby creating predefined operating rules or the artificial intelligence model set to perform desired characteristics (or purposes). This learning can be performed by the machine itself performing the artificial intelligence according to this disclosure, or it can be performed by a separate server and / or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples mentioned above.

[0074] Artificial intelligence models can be constructed using multiple neural network layers. Each neural network layer has multiple weight values, and neural network operations are performed through operations between the results of the previous layer and these weight values. The weights across multiple neural network layers can be optimized by the learning results of the artificial intelligence model. For example, multiple weights can be updated to reduce or minimize the loss or cost values ​​acquired by the artificial intelligence model during the learning process. Artificial neural networks can include deep neural networks (DNNs), such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), restricted Boltzmann machines (RBMs), deep belief networks (DBNs), bidirectional recurrent deep neural networks (BRDNNs), and deep Q-networks, but are not limited to the examples mentioned above.

[0075] According to exemplary embodiments of this disclosure, a processor can implement artificial intelligence. Artificial intelligence refers to machine learning methods based on artificial neural networks, which mimic human biological neurons to enable machine learning. Methodologies of artificial intelligence can be categorized as follows: supervised learning, which provides input and output data along with training data according to the learning method, and determines the answer (output data) to a question (input data); unsupervised learning, which provides only input data without output data, and does not determine the answer (output data) to a question (input data); and reinforcement learning, which, in the current state, provides a reward from the external environment whenever an action is taken, and learns to maximize this reward. Furthermore, methodologies of artificial intelligence can also be categorized based on the architecture of the learning model structure. The architectures of widely used deep learning technologies can be categorized as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformers, and Generative Adversarial Networks (GANs), etc.

[0076] This device and system may include an artificial intelligence model. The artificial intelligence model can be a single model or implemented as multiple models. The artificial intelligence model may be constructed using a neural network (or artificial neural network) and may include statistical learning algorithms that mimic biological neural networks in machine learning and cognitive science. A neural network can refer to an entire model in which artificial neurons (nodes) forming a network through synaptic connections learn to change the strength of synaptic connections, thereby gaining problem-solving capabilities. The neurons of a neural network may include combinations of weights or biases. A neural network may include one or more layers constructed using more than one neuron or node. Exemplarily, the device may include an input layer, a hidden layer, and an output layer. The neural network constituting the device can learn to change the weights of its neurons, thereby inferring the predicted outcome (output) from any input.

[0077] Processors can generate neural networks, or train (or learn) neural networks, or perform operations based on received input data and generate information signals based on the execution results, or retrain neural networks. Neural network models can include, but are not limited to, various types of models such as Convolutional Neural Networks (CNNs) (e.g., GoogleNet, AlexNet, VGG networks), Region with Convolutional Neural Networks (R-CNNs), Region Proposal Networks (RPNs), Recurrent Neural Networks (RNNs), Stacking-based Deep Neural Networks (S-DNNs), State-Space Dynamic Neural Networks (S-SDNNs), Deconvolution Networks, Deep Belief Networks (DBNs), Restricted Boltzmann Machines (RBMs), Fully Convolutional Networks, Long Short-Term Memory (LSTM) networks, and Classification Networks. Processors can include one or more processors for performing operations based on the neural network model. For example, a neural network can include a Deep Neural Network.

[0078] Neural networks can include Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), perceptrons, multilayer perceptrons, feedforward networks (FFs), Radial Basis Networks (RBFs), Deep Feedforward networks (DFFs), Long Short-Term Memory (LSTMs), Gated Recurrent Units (GRUs), Autoencoders (AEs), Variational Autoencoders (VAEs), Denoising Autoencoders (DAEs), Sparse Autoencoders (SAEs), Markov Chains (MCs), Hopfield Networks (HNs), and Boltzmann Machines (BMs).The network includes, but is not limited to, Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), Deep Convolutional Networks (DCNs), Deconvolutional Networks (DNs), Deep Convolutional Inverse Graphics Networks (DCIGNs), Generative Adversarial Networks (GANs), Liquid State Machines (LSMs), Extreme Learning Machines (ELMs), Echo State Networks (ESNs), Deep Residual Networks (DRNs), Differentiable Neural Computers (DNCs), Neural Turing Machines (NTMs), Capsule Networks (CNs), Kohonen Networks (KNs), and Attention Networks (ANs). Those skilled in the art will understand that it can include any neural network.

[0079] and Figure 2 The performance of the constituent elements shown can be adjusted by adding or deleting at least one constituent element. Furthermore, those skilled in the art will readily understand that the relative positions of the constituent elements can be changed depending on the performance or structure of the system.

[0080] in addition, Figure 2 The components shown refer to software and / or hardware components such as Field Programmable Gate Arrays (FPGAs) and Application Specific Integrated Circuits (ASICs).

[0081] Figure 3 This is a flowchart illustrating the object state prediction method based on dynamic image data according to the present disclosure (including steps S300, S310 to S330). Figure 4 This is a diagram illustrating the process of learning a first prediction model EM1 and a second prediction model EM2 based on initial dynamic image data according to this disclosure. Figure 3 During the process, necessary parts will be referred to together. Figure 4 .

[0082] In step S310, the processor 190 can acquire initial dynamic image data corresponding to the initial interval from when the drug is injected into the training subject to a preset time point.

[0083] Here, the drug can be a tracer. Tracers can be of various types depending on the target tissue or organ. For example, different types of tracers can be used to confirm the location or size of a tumor, examine brain function, or diagnose cardiac blood flow and metabolic activity.

[0084] In addition, the preset time point can be a time point within a preset range at a location where the radiation dose in a part that is not the target tissue or organ reaches its peak, and according to the embodiment, it can also be used as the default value.

[0085] In step S320, the processor 190 can use the initial dynamic image data corresponding to the first interval in the initial interval as input to train a first prediction model EM1 that predicts the first image data. The first image data represents anatomical information (or blood flow information) about the training object corresponding to a previous time point in the first interval. Here, the first image data can be very initial image data.

[0086] Reference Figure 4 The first prediction model EM1 may include a generative model EM1G ​​and a discriminative model EM1D.

[0087] When the processor 190 inputs the initial dynamic image data 710a corresponding to the first interval into the generation model EM1G ​​of the first prediction model EM1, it can predict (generate) the initial image data 710b of the intermediate pole.

[0088] The processor 190 can input the generated intermediate pole initial image data 710b and the label image data 710c (first label) of the pole initial image data into the discriminant model EM1D to verify the comparison between the intermediate pole initial image data 710b generated by the generative model EM1G ​​and the first label 710c. The processor 190 can repeat the generation and discrimination process until the prediction model EM1 meets the preset conditions (e.g., similarity of 95% or higher).

[0089] The first prediction model EM1 can be generated based on a GAN model, and a conditional GAN ​​model can be applied with additional use of conditional information (e.g., labels), but this disclosure is not limited thereto.

[0090] In step S330, the processor 190 can use the initial dynamic image data corresponding to the second interval in the initial interval as input to train a second prediction model EM2 that predicts the second image data representing disease-specific information about the training object corresponding to a reference time point after the initial interval. Here, the second image data can be delayed image data.

[0091] Reference Figure 4 The processor 190 can predict intermediate delayed image data 720b when the initial dynamic image data 720a corresponding to the second interval is input into the generator model EM2G of the second prediction model EM2.

[0092] Processor 190 inputs the generated intermediate delayed image data 720b and the labeled image data 720c (second label) of the delayed image data into the discriminative model EM2D to verify the comparison between the intermediate delayed image data 720b generated by the generative model EM2G and the second label 730c. Processor 190 can repeat the generation and discrimination process until the prediction model EM2 meets a preset condition (e.g., a similarity of 95% or higher).

[0093] The second prediction model EM2 can be generated based on a GAN model, and a conditional GAN ​​model can be applied with additional conditional information (e.g., labels), but this disclosure is not limited thereto.

[0094] In this embodiment, the amount of drug injected into the training subject may be a reduced amount compared to a reference amount. For example, the reduced amount may be half of the reference amount, but this disclosure is not limited thereto.

[0095] The first prediction model EM1 can be trained based on first labeled image data corresponding to the initial image data, and the second prediction model EM2 can be trained based on second labeled image data corresponding to the delayed image data. In this case, the first and second labeled image data can be labeled image data processed according to the drug injection amount of the training subject. When the injection amount of drug into the training subject is half of the reference amount, the processor 190 needs to use the first and second labeled image data as labeled image data equivalent to half of the reference amount.

[0096] In one embodiment, the processor 190 may also directly obtain first-label image data and second-label image data regarding cases where the drug injection dose is less than the reference dose through clinical trials, and use them for training.

[0097] However, even when injecting a small amount of drug compared to the reference amount, it may be difficult to ensure that permission from the state authorities is required. Therefore, when generating first and second label image data with a drug injection amount of the reference amount, the processor 190 can shorten the frame generation time to generate first and second label image data corresponding to the case where the drug injection amount is less than the reference amount.

[0098] For example, when generating 5 image data in 2-minute intervals, the processor 190 can, after generating image data in 1-minute intervals, acquire frames every 2 minutes (without using the image data originally generated every 2 minutes) and use the acquired image data as first label image data and second label image data. That is, the processor 190 can generate first label image data and second label image data corresponding to an injection dose equivalent to half the reference dose. Specifically, the processor 190 can utilize list-mode data to generate images showing a reduction in drug injection dose, or it can apply an approximation method to generate image frames in an interleaving manner in 1-minute intervals instead of the original 2-minute intervals.

[0099] According to the implementation example, the order between steps S320 and S330 can be changed for execution.

[0100] Figure 5 This is a flowchart illustrating a method S400 for predicting first image data representing anatomical information of a diagnostic subject and second image data representing disease-specific information according to the present disclosure.

[0101] In step S410, the processor 190 can acquire initial dynamic image data corresponding to the first interval after the drug is injected into the diagnostic subject.

[0102] In step S420, the processor 190 can input the initial dynamic image data of the diagnostic object corresponding to the first interval into the trained first prediction model EM1 to predict the first image data representing the anatomical information (or blood flow information) of the diagnostic object corresponding to the previous time point.

[0103] Here, the first image data can be very initial image data, and the first interval can be 10 to 30 minutes after drug injection, but the embodiment is not limited to this. The previous time point is the time point corresponding to the very initial image data, which can be within a predetermined range after drug injection (e.g., 1 minute), but the embodiment is not limited to this.

[0104] In step S430, the processor 190 can acquire initial dynamic image data corresponding to the second interval after the drug is injected into the diagnostic subject.

[0105] The second interval can be 10 to 20 minutes after drug injection, but this disclosure is not limited to this.

[0106] In step S440, the processor 190 can input the initial dynamic image data of the diagnostic object corresponding to the second interval into the trained second prediction model EM2 to predict the second image data representing the disease-specific information of the diagnostic object corresponding to the reference time point.

[0107] In one embodiment, the processor 190 can inject a reduced amount of drug into the diagnostic subject and, based on this, acquire and process initial dynamic image data corresponding to the first interval and initial dynamic image data corresponding to the second interval.

[0108] Here, the drug can be injected in a amount that is less than a predetermined amount than the reference dose, in order to reduce the amount of radiation injected.

[0109] The processor 190 can process initial dynamic image data corresponding to the first interval and initial dynamic image data corresponding to the second interval before inputting them into the prediction models EM1 and EM2. This is because the amount of refracted radiation is less than the reference amount, so it is processed into initial dynamic image data corresponding to the case of injecting the reference amount.

[0110] The processor 190 can process the movement path of the drug within the human body into a deeper and more defined form in the initial dynamic image data corresponding to the first interval, and can process each pixel value to the value of the injection reference dose using a smoothing filter (e.g., a Gaussian filter). That is, the processor 190 can process each pixel value using smoothing techniques (e.g., a spatial smoothing method).

[0111] In an embodiment, when the processor 190 inputs initial dynamic image data corresponding to a first interval with a small amount of drug and initial dynamic image data corresponding to a second interval to the processing model, the processor 190 can also use the processing model to output initial dynamic image data corresponding to a first interval with a reference amount of drug and initial dynamic image data corresponding to a second interval.

[0112] After acquiring the processed initial dynamic image data corresponding to the first interval and the processed initial dynamic image data corresponding to the second interval, the processor 190 can input the processed initial dynamic image data corresponding to the first interval into the first prediction model EM1 to predict the first image data representing the corresponding anatomical information, and can input the processed initial dynamic image data corresponding to the second interval into the second prediction model EM2 to predict the second image data representing the corresponding disease-specific information.

[0113] Figure 6 This is a flowchart illustrating the preprocessing procedure (including steps S500, S510 to S520) for execution space normalization according to this disclosure. Figure 7 This is a diagram illustrating the spatial normalization performed in the preprocessing step prior to training an artificial neural network model according to this disclosure.

[0114] In step S510, the processor 190 may perform spatial normalization before inputting the initial dynamic image data of the training objects and diagnostic objects corresponding to the first interval into the first prediction model EM1, so as to enable machine learning with less training object data.

[0115] In step S520, the processor 190 may perform spatial normalization before inputting the initial dynamic image data of the training object and the diagnostic object corresponding to the second interval into the second prediction model EM2.

[0116] Reference Figure 7 Before inputting the initial dynamic image data corresponding to the first interval of the training object and the diagnostic object into the first prediction model EM1, or before inputting the initial dynamic image data corresponding to the second interval of the training object and the diagnostic object into the second prediction model EM2, the processor 190 performs spatial normalization on the initial dynamic image data, thereby performing spatial normalization on the first image 1210, the second image 1220, and the third image 1230 based on the reference part.

[0117] For example, the processor 190 can spatially arrange the first image 1210 using horizontal lines D1a and vertical lines D2a with a reference center as a reference, spatially arrange the second image 1220 using horizontal lines D1b and vertical lines D2a with a reference center as a reference, and spatially arrange the third image 1230 using horizontal lines D1a and vertical lines D2b with a reference center as a reference. This improves training efficiency and enables effective diagnosis.

[0118] The processor 190 can set the acquisition time interval of the initial dynamic image data of the training object and the diagnostic object corresponding to the first interval and / or the second interval in the initial interval, in order to take into account the device operability and the predictability of the diagnostic object.

[0119] The processor 190 can identify an initial interval when the drug is injected into the diagnostic subject, and acquire initial dynamic image data corresponding to the first interval and the second interval based on the identified initial interval.

[0120] The processor 190 can set the time for acquiring initial dynamic image data corresponding to the first interval and the second interval according to the characteristics of each person, but this disclosure is not limited thereto.

[0121] Figure 8 This is a graph showing the change in radiation dose over time after injection of a drug according to this disclosure.

[0122] Reference Figure 8 When using FP-CIT as a drug (i.e., a tracer), radiation levels are high throughout the brain during the initial time frame of drug administration due to blood flow effects, which may make diagnosis difficult. However, as time passes, the drug is excreted in urine or metabolized by the liver, plasma concentrations decrease, and a reduction in blood drug levels can be confirmed. Furthermore, to maintain a balance in the reduction of blood drug levels, the drug flows from brain tissue into the bloodstream. The processor 190 can acquire initial dynamic imaging data of the initial range of radiation level changes within brain tissue.

[0123] Subsequently, FP-CIT binds strongly to dopamine, making it effective in identifying Parkinson's disease. Therefore, when FP-CIT binds to the dopamine neurotransmitter at the target site, the blood flow influence in other regions 610–630 is reduced, thus maintaining high concentrations at target sites 640 and 650 even with decreased concentrations. The processor 190 can determine the initial interval of the initial dynamic image data based on the point where the linear concentration in other regions (e.g., reference regions 610–630) reaches its peak.

[0124] The processor 190 can acquire delayed image data at a point in time after drug administration when the blood flow effect decreases after a reference time. The delayed image data can be existing image data used by medical personnel to diagnose the size, shape, etc. of the target site.

[0125] In embodiments where the drug used to capture image data is a tracer that binds to a specific target region, the initial interval for acquiring initial dynamic image data can be determined based on various benchmarks. For example, the initial interval can be set as the time range following the point at which the effect of blood flow begins to decrease in a reference region. The reference region is typically a specific region of the brain where there is little or no tracer binding and no difference between diseases. The reference region varies depending on the type and characteristics of the tracer. For example, in the case of FP-CIT, the reference region is a region with fewer dopamine neurotransmitters, such as the cerebellum or occipital cortex (OC). Furthermore, as another example, the initial interval can be set around the time point where the linear ratio difference between the target region and the reference region exceeds a specific value or where the ratio difference becomes maximum.

[0126] In cases where the drug used to capture image data is a tracer whose amount of binding to the target region continues to increase over time (e.g., fluorodeoxyglucose, hereinafter, FDG PET), the radiation levels in different regions of the brain increase without a peak. Therefore, the processor 190 acquires initial dynamic image data in an initial interval where the difference in radiation levels is not significant, and acquires delayed image data at a reference time point where the ratio of radiation level differences between regions increases.

[0127] Figure 9a It is a graph used to compare delayed image data generated with reference line measurements according to this disclosure with labeled image data. Figure 9b It is a graph used to compare delayed image data generated with label image data according to this disclosure using low line count. Figure 9c It is a chart used to compare delayed image data generated with reference line count and delayed image data generated with low line count according to this disclosure.

[0128] Reference Figure 9a If the DAT intake between the delayed image data generated with reference line quantity and the labeled image data is compared with respect to the anterior putamen (AP, the front part of the base nucleus), a correlation of 0.95 can be calculated, and with respect to the posterior putamen (PP, the rear part of the base nucleus), a correlation of 0.96 can be calculated.

[0129] Reference Figure 9b If we compare the DAT ingestion between the delayed image data generated with a low line count (half the reference line count) and the labeled image data relative to AP, we can calculate a correlation of 0.93, and relative to the posterior putamen (PP: the posterior part of the base nucleus), we can calculate a correlation of 0.95.

[0130] Reference Figure 9c If we compare the DAT ingestion between the delayed image data generated with the reference line count and the delayed image data generated with the low line count relative to the AP, we can calculate a correlation of 0.99, and relative to the PP, we can calculate a correlation of 1.00 between the delayed image data generated with the reference line count and the delayed image data generated with the low line count.

[0131] Reference Figures 9a to 9c As a characteristic of nuclear medicine molecular imaging, images are obtained under conditions of minimal decay due to half-life. Therefore, low-line images with reduced isotope injection doses can also predict delayed-uptake images (delayed images) of the same quality.

[0132] Figure 10aThe simulated PET data used in the training and testing steps for the anterior putamen (AP) and posterior putamen (PP) according to this disclosure are shown. Figure 10b The digital PET data used in the steps of testing AP and PP respectively, according to this disclosure, are shown.

[0133] Reference Figure 10a For simulated PET data, the test set results confirm a correlation of 0.95 with AP and 0.96 with PP.

[0134] Reference Figure 10b For digital PET data, for the independent test set, the correlation with AP is confirmed to be 0.93, and the correlation with PP is confirmed to be 0.97.

[0135] Figure 11 This is a diagram illustrating a method for acquiring extremely initial image data according to the present disclosure.

[0136] Within the initial interval, the correlation between the dynamic image data 10 minutes after drug injection and the initial image data was confirmed to be 0.94 (10¹⁰), the correlation between the dynamic image data 20 minutes after drug injection and the initial image data was confirmed to be 0.77, and the correlation between the dynamic image data 30 minutes after drug injection and the initial image data was confirmed to be 0.53. Thus, the highest correlation was confirmed for the dynamic image data 10 minutes after drug injection, which is closest in time to the initial image data.

[0137] The processor 190 can predict the initial image data 1040 based on dynamic image data from 10 to 30 minutes after drug injection. That is, the initial image data 1040 can be effectively predicted using the initial dynamic image data (e.g., the first interval of the initial interval).

[0138] Figure 12 This is a diagram illustrating a method for acquiring delayed image data according to the present disclosure.

[0139] Within the initial interval, the correlation between the dynamic image data 10 minutes after drug injection and the delayed image data was confirmed to be 0.80 (1110), and the correlation between the dynamic image data 20 minutes after drug injection and the delayed image data was confirmed to be 0.94. Thus, the highest correlation was found in the dynamic image data 20 minutes after drug injection, which is closest in time to the delayed image data.

[0140] The processor 190 can predict delayed image data 1130 based on dynamic image data from 10 to 20 minutes after drug injection. That is, delayed image data 1130 can be effectively predicted using initial dynamic image data (e.g., a second interval of the initial interval).

[0141] Furthermore, the image quality of moving image data can be set according to the required number of frames. That is, if the same amount of drug is used, the maximum amount of radiation released to the outside is limited, thereby limiting the amount of signal used to generate image frames. Therefore, the computing device can set the image quality of each image frame differently based on the number of image frames, while using the same amount of signal.

[0142] For example, if the number of image frames is increased, the time required to generate each image frame becomes shorter, thus reducing the amount of signal available to generate one image frame, and the computing device generates low-quality image frames.

[0143] As one embodiment, the computing device can learn by matching the changes of each pixel in dynamic image data over time with each pixel in initial image data and delayed image data.

[0144] That is, in a combination of image data for a specific patient (i.e., a combination of initial dynamic image data, dynamic image data and delayed image data), the computing device can predict the initial dynamic image data and delayed image data based on the dynamic image data for each point (i.e., pixel) of body tissue (e.g., brain tissue), match anatomical information in the initial dynamic image data and disease-specific information in the delayed image data to construct a dataset for each combination of image data, and learn about the datasets of multiple patients according to each point (i.e., pixel) to construct a diagnostic image prediction model.

[0145] In this embodiment, the diagnostic image prediction model is constructed using a deep neural network (DNN). That is, the diagnostic image prediction model applies deep learning algorithms to learn dynamic image data about more than one patient.

[0146] The processor can generate the image by normalizing the brightness of the dynamic image data based on the maximum or average value.

[0147] In an embodiment, when the drug used to capture image data is a tracer whose amount of binding to the target region continues to increase over time (e.g., fluorodeoxyglucose, hereinafter, FDG PET), due to the characteristic of the PET tracer being taken up in proportion to the amount of drug administered, the brightness ratio of the reference area to the target area will not differ depending on the amount of drug administered. Therefore, the computing device generates image data by normalizing based on the maximum or average brightness of the dynamic image data.

[0148] Furthermore, while the operation is described using PET in this disclosure, the operation can be applied not only to PET but also to MR and the entire medical imaging process.

[0149] However, in the case of MR, disease-specific information can be derived from initial dynamic image data, and anatomical information can be derived from delayed image data.

[0150] When the drug used to capture image data is a tracer that binds to a specific target region (in the case of using FP-CIT as a tracer), in a linear tracer kinetic model that takes a reference region as input and the target region as output, the parameters of the tracer model are the same even if the input and output increase or decrease linearly at the same time. Therefore, the computing device can generate new delayed image data by normalizing the brightness of the initial dynamic image data for learning and the initial dynamic image data for diagnosis based on the maximum or average value (count normalization or intensity normalization).

[0151] That is, when the drug used to capture image data is a tracer that binds to a specific target region, the learning dynamic image data can be formed from the following image data: image data acquired from the time point when the difference in linear ratio between the target region and the reference region becomes above a specific value or the ratio difference becomes the largest, until the time point when the influence of blood flow is excluded.

[0152] Furthermore, learning dynamic image data can be formed from the following image data: image data acquired from the time point when the difference between the linear ratio of the target region and the reference region is above a certain value or when the linear change in the reference region caused by blood flow decreases, to the time point when the difference between the linear ratio of the target region and the reference region becomes the maximum value.

[0153] The radiation ratio can refer to the ratio of the radiation dose at the initial time point to the radiation dose at a specific measurement time point.

[0154] Therefore, the line ratio difference can refer to the difference between the line ratio of the target area and the line ratio of the reference area.

[0155] Specifically, the processor can acquire information about the line quantity of the target image and the reference image.

[0156] In addition, the processor can acquire the line quantity corresponding to the target image and the line quantity corresponding to the reference, calculate the ratio value of each line quantity, and calculate the difference of each line quantity ratio.

[0157] In addition, the processor can determine the time point when the reference linear change decreases and the time point when the linear ratio difference becomes the largest.

[0158] Typically, when medical personnel perform diagnoses based on delayed-image data, they administer a large dose of medication at the initial time point so that even after the radioactive isotopes have decreased due to physical half-life and external excretion such as urine / feces (i.e., physiological reduction), they can still perform diagnoses using delayed-image data taken with a sufficient dose of radiation. In this case, there is a problem of increased radiation exposure to the patient.

[0159] Therefore, using one embodiment of the present invention, when acquiring diagnostic dynamic image data, only a tracer of a level sufficient to acquire a sufficient amount of radiation is inserted, and then diagnostic dynamic image data is acquired. The diagnostic dynamic image data is then inserted into a prediction model to acquire initial dynamic image data and final diagnostic delayed image data. As a result, the amount of radioactive material injected into the patient can be reduced, thereby reducing the amount of radiation delivered to the patient's body.

[0160] The predetermined reference time can be determined based on the type of tracer.

[0161] In addition, the processor can generate second learning data, which is based on a second diagnostic image prediction model that uses the initial dynamic image data and the delayed image data to predict anatomical information and disease-specific information from the initial dynamic image data and the delayed image data.

[0162] In summary, the first diagnostic prediction model predicts initial dynamic image data and delayed image data from dynamic image data, and learns and predicts anatomical information and disease-specific information corresponding to the initial dynamic image data and delayed image data. The second diagnostic prediction model can perform learning based on the initial dynamic image data and delayed image data corresponding to the dynamic image data. Subsequently, if the initial dynamic image data or delayed image data for diagnosis is input, it can comprehensively predict anatomical information and disease-specific information.

[0163] In addition, the processor can determine the state of an object based on anatomical information and disease-specific information.

[0164] Specifically, the processor can use the disease-specific and anatomical information of the subject derived through the above methods to derive information about whether the subject suffers from Parkinson's disease or dementia.

[0165] Specifically, if the tracer remains in the dopamine neurotransmitter of the subject, it can be determined that the subject is in a state of Parkinson's disease.

[0166] Furthermore, the presence of tracers containing residual amyloid protein can indicate that the individual is in a state of dementia.

[0167] The processor can determine the diagnostic image prediction model with higher accuracy from the accuracy of the first diagnostic image prediction model and the accuracy of the second diagnostic image prediction model.

[0168] For example, when a diagnostician performs a diagnosis using initial dynamic image data, the accuracy of a first diagnostic prediction model for Parkinson's disease and the accuracy of a second diagnostic prediction model for Parkinson's disease can be calculated. In this case, the processor can select the diagnostic prediction model with high accuracy from the various diagnostic prediction models to determine the state of the object.

[0169] Under detailed examination conditions, such as appropriate dynamic image acquisition time points, the aforementioned diagnostic image generation method based on initial dynamic image data according to an embodiment of the present invention can be implemented as a program (or application program) and stored in a medium for integration and operation with a computer as hardware.

[0170] The diagnostic image generation method based on initial dynamic image data according to an embodiment of the present invention described above can be implemented as a program (or application) and stored in a medium for integration and operation with a computer as hardware.

[0171] The aforementioned program may include code encoded in computer languages ​​such as C, C++, JAVA, and machine language, which can be read by the computer's processor (CPU) through the computer's device interface. This code may include functional code, such as functions defining the functions required to run the method, and may include control code related to the execution steps required for the computer's processor to run the function according to predetermined steps. Furthermore, this code may also include additional information required for the computer's processor to run the function, or memory reference code regarding the location (address) of the medium in the computer's internal or external memory. Additionally, if the computer's processor needs to communicate with any other computer or server located remotely to run the function, the code may also include communication-related code regarding which other computer or server needs to be communicated with using the computer's communication module, how communication should be conducted, and what information or media needs to be sent / received during communication.

[0172] The storage medium refers to a medium that semi-permanently stores data and can be read by a machine, rather than a medium that stores data for a short period of time, such as registers, caches, or memory. Specifically, examples of the storage medium include, but are not limited to, ROM, RAM, CD-ROM, magnetic tape, floppy disks, and optical data storage devices. That is, the program can be stored on various recording media on various servers accessible to the computer or on various recording media on the user's computer. Furthermore, the medium can be distributed across computer systems connected via a network and store computer-readable code in a distributed manner.

[0173] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, those skilled in the art will understand that the present invention can be implemented in other specific forms without changing its technical concept or essential features. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive.

Claims

1. A method, the method being executed by at least one processor, the method comprising the following steps: Acquire initial dynamic image data corresponding to the initial interval up to a pre-set time point after the drug is injected into the training subject; Using initial dynamic image data corresponding to the first interval in the initial interval as input, a first prediction model is trained to predict first image data representing anatomical or blood flow information of the training object corresponding to previous time points in the first interval; and The initial dynamic image data corresponding to the second interval in the initial interval is used as input to train a second prediction model that predicts the second image data representing the disease-specific information of the training object corresponding to a reference time point after the initial interval.

2. The method according to claim 1, further comprising the following steps: Acquire initial dynamic image data corresponding to the first interval after the drug is injected into the diagnostic subject; and The initial dynamic image data of the diagnostic object corresponding to the first interval is input into the trained first prediction model to predict the first image data representing the anatomical or blood flow information of the diagnostic object corresponding to the previous time point.

3. The method according to claim 2, further comprising the following steps: Acquire initial dynamic image data corresponding to the second interval after the drug is injected into the diagnostic subject; and The initial dynamic image data of the diagnostic subject corresponding to the second interval is input into the trained second prediction model to predict second image data representing disease-specific information about the diagnostic subject corresponding to the reference time point.

4. The method according to claim 3, wherein, The processor is configured such that, When a predetermined amount of the drug is injected into the subject of the diagnosis, initial dynamic image data corresponding to the first interval and initial dynamic image data corresponding to the second interval are acquired and processed accordingly to obtain initial dynamic image data corresponding to the processed first interval and initial dynamic image data corresponding to the processed second interval. The initial dynamic image data corresponding to the processed first interval is input into the first prediction model, and the initial dynamic image data corresponding to the processed second interval is input into the second prediction model.

5. The method according to claim 3, wherein, The amount of drug injected into the training subject was reduced compared to the reference amount. The first prediction model is trained based on first-label image data corresponding to the first image data, and the second prediction model is trained based on second-label image data corresponding to the second image data. The first and second label image data are label image data processed based on the drug injection dosage of the training subject.

6. The method according to claim 3, further comprising the following steps: Spatial normalization is performed before inputting the initial dynamic image data of the training object and the diagnostic object corresponding to the first interval into the first prediction model; and Spatial normalization is performed before inputting the initial dynamic image data of the training object and the diagnostic object corresponding to the second interval into the second prediction model.

7. The method according to claim 3, wherein, The processor is configured such that, The acquisition time intervals for the initial dynamic image data of the training object and the diagnostic object are defined to correspond to the first interval in the initial interval. Set the acquisition time interval for the initial dynamic image data of the training object and the diagnostic object, which corresponds to the second interval in the initial interval.

8. The method according to claim 3, wherein, The processor is configured as follows: When the drug is injected into the body of the diagnostic subject, the initial interval is identified, and based on the identified initial interval, initial dynamic image data corresponding to the first interval and the second interval is acquired.

9. The method according to claim 1, wherein, The acquired initial dynamic image data, the first image data, and the second image data are positron emission tomography (PET) image data.

10. An apparatus for predicting the state of an object based on dynamic image data, comprising: Memory; as well as At least one processor communicates with the memory. The processor is configured as follows: Initial dynamic image data corresponding to an initial interval up to a predetermined time point after the drug is injected into the training subject is acquired. Initial dynamic image data corresponding to a first interval in the initial interval is used as input to train a first prediction model that predicts first image data representing anatomical or blood flow information of the training subject corresponding to a previous time point in the first interval. Initial dynamic image data corresponding to a second interval in the initial interval is used as input to train a second prediction model that predicts second image data representing disease-specific information of the training subject corresponding to a reference time point after the initial interval.

11. The apparatus according to claim 10, wherein, The processor is configured such that, Acquire initial dynamic image data corresponding to the first interval after the drug is injected into the subject of diagnosis. The initial dynamic image data of the diagnostic object corresponding to the first interval is input into the trained first prediction model to predict the first image data representing the anatomical or blood flow information of the diagnostic object corresponding to the previous time point.

12. The apparatus according to claim 11, wherein, The processor is configured such that, Acquire initial dynamic image data corresponding to the second interval after the drug is injected into the body of the diagnostic subject. The initial dynamic image data of the diagnostic subject corresponding to the second interval is input into the trained second prediction model to predict second image data representing disease-specific information about the diagnostic subject corresponding to the reference time point.

13. The apparatus according to claim 12, wherein, The processor is configured such that, When a predetermined amount of the drug is injected into the subject of the diagnosis, initial dynamic image data corresponding to the first interval and initial dynamic image data corresponding to the second interval are acquired and processed accordingly to obtain initial dynamic image data corresponding to the processed first interval and initial dynamic image data corresponding to the processed second interval. The initial dynamic image data corresponding to the processed first interval is input into the first prediction model, and the initial dynamic image data corresponding to the processed second interval is input into the second prediction model.

14. The apparatus according to claim 12, wherein, The amount of drug injected into the training subject was reduced compared to the reference amount. The first prediction model is trained based on the first labeled image data corresponding to the first image data. The second prediction model is trained based on the second labeled image data corresponding to the second image data. The first and second label image data are label image data processed based on the drug injection dosage of the training subject.

15. The apparatus according to claim 12, wherein, The processor is configured such that, Spatial normalization is performed before inputting the initial dynamic image data of the training object and the diagnostic object corresponding to the first interval into the first prediction model. Spatial normalization is performed before inputting the initial dynamic image data of the training object and the diagnostic object corresponding to the second interval into the second prediction model.

Images