Device and method for estimating antibiotic resistance by using antibiotic resistance estimation model

Abstract

Disclosed are a device and method for estimating antibiotic resistance by using an antibiotic resistance estimation model. The device for estimating antibiotic resistance by using an antibiotic resistance estimation model may comprise: a data reception unit for receiving antibiotic profile data including user information data of a user whose antibiotic resistance is to be estimated, vital sign data indicating the vital signs of the user, and the antibiotic usage history of the user; a data preprocessing unit for replacing missing values of the user information data, the vital sign data, and the antibiotic profile data and merging pieces of data in which the missing values are replaced; a feature extraction unit for extracting feature data from the merged data; and an antibiotic resistance estimation unit for acquiring, from the antibiotic resistance estimation model, result data including resistance estimation information for the target antibiotic of the user by inputting the feature data to the trained antibiotic resistance estimation model.

Description

Device and method for estimating antibiotic resistance using an antibiotic resistance estimation model

[0001] The following disclosure relates to an antibiotic resistance estimation device and method using an antibiotic resistance estimation model.

[0002] Antibiotic resistance refers to the resistance of pathogens to specific antibiotics. Microbial culture and susceptibility testing, genetic analysis-based testing, machine learning-based testing, or biomarker analysis can be used to predict antibiotic resistance. Predicting antibiotic resistance can shorten patient recovery times and prevent the spread of antibiotic resistance by using appropriate antibiotics. Furthermore, predicting antibiotic resistance and regulating antibiotic use can help prevent the overuse and misuse of antibiotics.

[0003] According to one embodiment, an antibiotic resistance estimation device using an antibiotic resistance estimation model may include a data receiving unit that receives user information data of a user who is a target of antibiotic resistance estimation, vital signs data indicating vital signs of the user, and antibiotic profile data including an antibiotic use history for the user, a data preprocessing unit that replaces missing values ​​of the user information data, vital signs data, and antibiotic profile data and merges data with replaced missing values, a feature extraction unit that extracts feature data from the merged data, and an antibiotic resistance estimation unit that obtains result data including resistance estimation information for a target antibiotic of the user from the antibiotic resistance estimation model by inputting the feature data into a learned antibiotic resistance estimation model.

[0004] The vital signs data may include at least average values ​​for the user's body temperature, the user's blood pressure and pulse, and the user's maximum respiration value.

[0005] User information data may include risk index data indicating a risk index based on the user's underlying disease.

[0006] The data preprocessing unit can impute missing values ​​based on multiple imputation by chained equations (MICE).

[0007] The antibiotic resistance estimation model may be based on at least one of a logistic regression (LR) model, an extreme gradient boosting (XGBoost) model, a light gradient boosting machine (LGBBM) model, a CatBoost model, a multi-layer perceptron (MLP) model, and a multi-task learning (MTL) model.

[0008] The feature extraction unit can extract feature data based on SHAP (Shapley additive explanations) values.

[0009] The feature data may include at least resistance data indicating whether or not the patient is resistant to existing antibiotics, administration frequency data indicating the number of times the antibiotic is administered, risk index data, and data on the length of hospitalization of the user.

[0010] A learning device for learning an antibiotic resistance estimation model according to one embodiment may include a data receiving unit that receives user information learning data of a user who is a target of antibiotic resistance estimation, vital signs learning data indicating vital signs of the user, and antibiotic profile learning data indicating antibiotic use history of the user, a data preprocessing unit that replaces missing values ​​in the user information learning data, vital signs learning data, and antibiotic profile learning data and merges learning data in which missing values ​​have been replaced, a feature extraction unit that extracts feature data from the merged learning data, and an antibiotic resistance estimation model learning unit that trains an antibiotic resistance estimation model by using result data obtained from an antibiotic resistance estimation model that inputs feature data, the user information learning data, the vital signs learning data, and the antibiotic profile learning data.

[0011] The vital signs learning data may include at least average values ​​for the user's body temperature, the user's blood pressure, and the user's maximum respiration value, and the antibiotic profile learning data may include prescription data for a target antibiotic and culture test result data indicating whether the target antibiotic is resistant.

[0012] The user information learning data may include risk index data indicating a risk index based on the user's underlying disease.

[0013] The antibiotic resistance estimation model learning unit can train the antibiotic resistance estimation model using a multi-task learning method based on at least one of hard-parameter sharing and soft-parameter sharing.

[0014] According to one embodiment, a method for estimating antibiotic resistance using an antibiotic resistance estimation model may include an operation of receiving user information data of a user who is a target of antibiotic resistance estimation, vital signs data indicating vital signs of the user, and antibiotic profile data including an antibiotic use history for the user, an operation of replacing missing values ​​of the user information data, vital signs data, and antibiotic profile data and merging data with replaced missing values, an operation of extracting feature data from the merged data, and an operation of obtaining result data including resistance estimation information for a target antibiotic of the user from the antibiotic resistance estimation model by inputting the feature data into a learned antibiotic resistance estimation model.

[0015] The user information data may include risk index data indicating a risk index based on the user's underlying disease, and the feature data may include at least resistance data indicating resistance to existing antibiotics, administration frequency data indicating the number of times antibiotics are administered, risk index data, and data on the user's hospitalization period.

[0016] According to one embodiment, a learning method for learning an antibiotic resistance estimation model may include an operation of receiving user information learning data of a user who is a target of antibiotic resistance estimation, vital signs learning data indicating vital signs of the user, and antibiotic profile learning data indicating an antibiotic use history of the user, an operation of replacing missing values ​​of the user information learning data, vital signs learning data, and antibiotic profile learning data and merging learning data in which missing values ​​have been replaced, an operation of extracting feature data from the merged learning data, and an operation of training the antibiotic resistance estimation model using result data obtained from an antibiotic resistance estimation model that has input the feature data, the user information learning data, the vital signs learning data, and the antibiotic profile learning data.

[0017] The training operation may include training an antibiotic resistance estimation model using a multi-task learning method based on at least one of hard-parameter sharing and soft-parameter sharing.

[0018] Figure 1 is a block diagram illustrating an antibiotic resistance estimation device according to one embodiment.

[0019] FIG. 2 is a block diagram illustrating a learning device for learning an antibiotic resistance estimation model according to one embodiment.

[0020] FIG. 3 is a diagram for explaining biosignal data according to one embodiment.

[0021] Figure 4 is a diagram illustrating a process for learning an antibiotic resistance estimation model according to one embodiment.

[0022] Figure 5 is a diagram illustrating a process for learning an antibiotic resistance estimation model according to one embodiment.

[0023] FIG. 6 is a flowchart illustrating the operations of an antibiotic resistance estimation method using an antibiotic resistance estimation model according to one embodiment.

[0024] Figure 7 is a flowchart illustrating the operations of a learning method for learning an antibiotic resistance estimation model according to one embodiment.

[0025] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Therefore, the actual implementation is not limited to the specific embodiments disclosed, and the scope of this specification includes modifications, equivalents, or alternatives within the technical concepts described in the embodiments.

[0026] Although terms such as "first" or "second" may be used to describe various components, these terms should be interpreted solely to distinguish one component from another. For example, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.

[0027] When it is said that a component is "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but there may also be other components in between.

[0028] Singular expressions include plural expressions unless the context clearly dictates otherwise. In this document, phrases such as "A or B", "at least one of A and B", "at least one of A or B", "A, B, or C", "at least one of A, B, and C", and "at least one of A, B, or C" can each include any one of the items listed together in that phrase, or all possible combinations thereof. In this specification, it should be understood that the terms "comprises" or "has" and the like are intended to specify the presence of a described feature, number, step, operation, component, part, or combination thereof, but do not exclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0029] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art. Terms defined in commonly used dictionaries should be interpreted to have a meaning consistent with their meaning in the context of the relevant technology, and will not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0030] The term "~part" as used in this document refers to a software or hardware component such as an FPGA or ASIC, and the "~part" performs certain roles. However, the "~part" is not limited to software or hardware. The "~part" may be configured to reside on an addressable storage medium and may be configured to execute one or more processors. For example, the "~part" may include components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functionality provided within the components and "~parts" may be combined into a smaller number of components and "~parts" or further separated into additional components and "~parts." Furthermore, the components and "~parts" may be implemented to execute one or more CPUs within a device or a secure multimedia card. Additionally, '~bu' may include one or more processors.

[0031] Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the attached drawings, identical components are assigned the same reference numerals regardless of the drawing numbers, and redundant descriptions thereof will be omitted.

[0032]

[0033] Figure 1 is a block diagram illustrating an antibiotic resistance estimation device according to one embodiment.

[0034] The antibiotic resistance estimation device (100) is a device for estimating whether a user is resistant to a target antibiotic. Referring to FIG. 1, the antibiotic resistance estimation device (100) may include a data receiving unit (110), a data preprocessing unit (120), a feature extraction unit (130), and an antibiotic resistance estimation unit (140). The operations of the data receiving unit (110), the data preprocessing unit (120), the feature extraction unit (130), and the antibiotic resistance estimation unit (140) of the antibiotic resistance estimation device (100) may be performed based on a processor and a memory.

[0035] A processor can execute instructions stored in memory. The processor may include a central processing unit (CPU), a graphics processing unit (GPU), a neural network processing unit (NPU), a media processing unit (MPU), a data processing unit (DPU), a vision processing unit (VPU), a video processor, an image processor, a display processor, a microprocessor, a processor core, a multi-core processor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or any combination thereof.

[0036] The memory can store instructions that can be executed by the processor. The memory can store instructions that can be executed by the processor. The instructions executable by the processor, when executed by the processor, can cause the processor to perform an image processing method or a method for training a transformation model. The memory can be integrated with the processor. For example, read-only memory (RAM) or flash memory can be arranged in an integrated circuit microprocessor, etc. The memory can also include a separate device, such as an external disk drive, a storage array, or other storage device usable by a database system. The memory and the processor can be operatively coupled or can communicate with each other via an I / O port, a network connection, etc., so that the processor can read a file stored in the memory. The memory can be a computer-readable storage medium that stores instructions, and the instructions stored in the memory, when executed by the processor, can prompt at least one processor to perform an image processing method or a method for training an image processing model.

[0037] Computer-readable storage media include read-only memory (ROM), programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, nonvolatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, BLU-RAY or optical disk memory, hard disk drive (HDD), solid state drive (SSD), card memory (e.g., multimedia card, secure digital (SD) card, or extreme digital (XD) card), magnetic tape, It may include floppy disks, magneto-optical data storage devices, optical data storage devices, hard disks, solid state disks, and other devices.

[0038] The data receiving unit (110) may receive user information data, vital signs data indicating the user's vital signs, and antibiotic profile data. The user information data may include user information of the user who is the subject of antibiotic resistance estimation. For example, the user information data may include personal information such as the user's name, age, and address. According to one embodiment, the user information data may include risk index data indicating a risk index based on the user's underlying disease. The underlying disease may indicate a chronic or long-term illness that the user already has. Since information based on the user's medical history is considered personal information, the user information learning data may include a risk index based on the underlying disease.

[0039] Vital signs data represents data obtained by measuring vital signs. For example, vital signs data may include a user's body temperature, blood pressure, pulse, respiration, or average values ​​for each.

[0040] Antibiotic profile data is data that describes a user's antibiotic use history. For example, antibiotic profile data may include whether antibiotics were prescribed and / or antibiotic use history for drugs such as Piperacillin, Tazobactam, or Meropenem.

[0041] The data preprocessing unit (120) can replace missing values ​​and merge data with replaced missing values. In one embodiment, the data preprocessing unit (120) can replace missing values ​​based on MICE (multiple imputation by chained equations). MICE is a method to replace missing values ​​more accurately by generating multiple imputations through multiple iterative imputation processes rather than filling in missing values ​​with a single value. For example, if a missing value exists in a user's body temperature or blood pressure among vital signs data, the missing value can be replaced based on generating multiple imputations (or replacement values) for the user's body temperature or blood pressure.

[0042] The feature extraction unit (130) can extract feature data from the merged data. For example, the feature extraction unit (130) can extract feature data from the merged data based on Shapley additive explanations (SHAP)-value. SHAP-value is a method of extracting feature data from input data based on the influence of each feature on the output value of the model. For example, if the patient's antibiotic use history and underlying disease among the merged data influence the output value of the antibiotic resistance estimation model, the feature extraction unit (130) can extract antibiotic profile data and user information data as features.

[0043] The antibiotic resistance estimation unit (140) can obtain result data including resistance estimation information for the user's target antibiotic from the antibiotic resistance estimation model by inputting feature data into the learned antibiotic resistance estimation model. The antibiotic resistance estimation model according to one embodiment may be based on at least one of a logistic regression (LR) model, an extreme gradient boosting (XGBoost) model, a light gradient boosting machine (LGBBM) model, a CatBoost model, a multi-layer perceptron (MLP) model, and a multi-task learning (MTL) model. For example, the antibiotic resistance estimation model may be a multi-learning-based MTL model. The antibiotic resistance estimation unit (140) can obtain result data including antibiotic resistance estimation information for the target antibiotic (e.g., Piperacillin) from the MTL model by inputting feature data into the MTL model. The multi-learning-based MTL model will be described in more detail with reference to FIGS. 4 and 5.

[0044]

[0045] FIG. 2 is a block diagram illustrating a learning device for learning an antibiotic resistance estimation model according to one embodiment.

[0046] The learning device (200) is a device that learns an antibiotic resistance estimation model for antibiotic resistance estimation. Referring to FIG. 2, the learning device (200) may include a data receiving unit (210), a data preprocessing unit (220), a feature extraction unit (230), and an antibiotic resistance estimation model learning unit (240). The operations of the data receiving unit (210), the data preprocessing unit (220), the feature extraction unit (230), and the antibiotic resistance estimation model learning unit (240) may be performed based on a processor and a memory.

[0047] A processor can execute instructions stored in memory. The processor may include a central processing unit (CPU), a graphics processing unit (GPU), a neural network processing unit (NPU), a media processing unit (MPU), a data processing unit (DPU), a vision processing unit (VPU), a video processor, an image processor, a display processor, a microprocessor, a processor core, a multi-core processor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or any combination thereof.

[0048] The memory can store instructions that can be executed by the processor. The memory can store instructions that can be executed by the processor. The instructions executable by the processor, when executed by the processor, can cause the processor to perform an image processing method or a method for training a transformation model. The memory can be integrated with the processor. For example, read-only memory (RAM) or flash memory can be arranged in an integrated circuit microprocessor, etc. The memory can also include a separate device, such as an external disk drive, a storage array, or other storage device usable by a database system. The memory and the processor can be operatively coupled or can communicate with each other via an I / O port, a network connection, etc., so that the processor can read a file stored in the memory. The memory can be a computer-readable storage medium that stores instructions, and the instructions stored in the memory, when executed by the processor, can prompt at least one processor to perform an image processing method or a method for training an image processing model.

[0049] Computer-readable storage media include read-only memory (ROM), programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, nonvolatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, BLU-RAY or optical disk memory, hard disk drive (HDD), solid state drive (SSD), card memory (e.g., multimedia card, secure digital (SD) card, or extreme digital (XD) card), magnetic tape, It may include floppy disks, magneto-optical data storage devices, optical data storage devices, hard disks, solid state disks, and other devices.

[0050] The data receiving unit (210) can receive user information learning data, vital signs learning data, and antibiotic profile learning data. The user information learning data represents user information of a user who is a target of antibiotic resistance estimation used to train an antibiotic resistance estimation model. For example, the user information learning data may include personal information such as the user's name, age, address, etc. In one embodiment, the user information learning data may include risk index data representing a risk index based on the user's underlying disease. The underlying disease may represent a chronic or long-term disease that the user already has. Since information based on the user's medical history is considered personal information, the risk index based on the underlying disease may be included in the user information learning data.

[0051] Vital sign learning data is data obtained by measuring vital signs to train an antibiotic resistance estimation model. The vital sign learning data may include at least the average values ​​for the user's body temperature, blood pressure, and maximum respiration.

[0052] Antibiotic profile training data is data representing the antibiotic use history used to train an antibiotic resistance estimation model. For example, antibiotic profile training data may include antibiotic prescriptions and / or antibiotic use history for antibiotics such as Piperacillin, Tazobactam, or Meropenem, as well as culture test results indicating antibiotic resistance.

[0053] The data preprocessing unit (220) can replace missing values ​​and merge training data with replaced missing values. The data preprocessing unit (220) can replace missing values ​​of user information training data, vital signs training data, and antibiotic profile training data and merge training data with replaced missing values. For example, the data preprocessing unit (220) can replace missing values ​​based on MICE (multiple imputation by chained equations; MICE) and merge training data with replaced missing values. Since replacing missing values ​​based on MICE (multiple imputation by chained equations; MICE) has been described in FIG. 1, a duplicate description will be omitted.

[0054] The feature extraction unit (230) can extract feature data from the merged training data. For example, the feature extraction unit (230) can extract feature data from the merged training data based on Shapley additive explanations (SHAP) values. The extracted feature data can include at least resistance data indicating resistance to existing antibiotics, administration frequency data indicating the number of antibiotic administrations, risk index data, and data on the user's hospitalization period. Since extracting features based on SHAP values ​​has been described in FIG. 1, a redundant description will be omitted.

[0055] The antibiotic resistance estimation model learning unit (240) can train the antibiotic resistance estimation model. The antibiotic resistance estimation model learning unit (240) can train the antibiotic resistance estimation model using result data, user information learning data, vital signs learning data, and antibiotic profile learning data obtained from the antibiotic resistance estimation model that inputs feature data. For example, the antibiotic resistance estimation model learning unit (240) can train the antibiotic resistance estimation model using a multi-task learning method based on at least one of hard parameter sharing and soft parameter sharing. The multi-task learning method will be described in more detail with reference to FIGS. 4 and 5.

[0056]

[0057] FIG. 3 is a diagram for explaining biosignal data according to one embodiment.

[0058] The vital signs data may include at least average values ​​for the user's (e.g., patient's) body temperature, the user's blood pressure and pulse, and the user's maximum respiration rate. Referring to FIG. 3, the vital signs data may include an average value for the patient's daily maximum body temperature.

[0059] According to one embodiment, a patient's body temperature may be measured more than once on the same day, or may be measured continuously over multiple days. For example, table (310) includes multiple body temperature information for patient A. Blocks (311), (312), and (313) represent body temperature information for patient A on April 7, April 8, and April 9. Block (311) includes multiple body temperature information for patient A measured on April 7.

[0060] Table (320) shows the daily maximum body temperature for patient A. Since the body temperature for patient A measured on April 8 and April 9 is 1, block (312) and block (313) of table (310) can correspond to block (322) and block (323) of table (320), respectively. Block (321) shows the maximum value among multiple body temperatures for patient A measured on April 7.

[0061] Table (330) shows the average value of the daily maximum body temperature for patient A. Since the respective daily maximum body temperatures of patient A in blocks (321), (322), and (323) of table (320) are 38.5 degrees, 36.5 degrees, and 36.5 degrees, the average value of the daily maximum body temperature for patient A from April 7 to 9 in table (330) is 37.17 degrees. The average value of the daily maximum body temperature for patient A determined through the above process can be included in the vital signs data.

[0062] The above process is not limited to just the patient's body temperature, but can equally be applied to the patient's blood pressure and pulse, as well as the user's maximum respiration value.

[0063]

[0064] Figure 4 is a diagram illustrating a process for learning an antibiotic resistance estimation model according to one embodiment.

[0065] Learning of an antibiotic resistance estimation model can be performed by a learning device (e.g., learning device (200) of FIG. 2). The antibiotic resistance estimation model performed by the learning device can be a model based on at least one of a logistic regression (LR) model, an extreme gradient boosting (XGBoost) model, a light gradient boosting machine (LGBBM) model, a CatBoost model, a multi-layer perceptron (MLP) model, and a multi-task learning (MTL) model.

[0066] Referring to FIG. 4, a learning device can learn an antibiotic resistance estimation model (e.g., an MTL model) using a multi-task learning method based on hard parameter sharing. The multi-task learning model is a method that can improve the generalization performance of a machine learning model by learning multiple related tasks simultaneously and utilizing the correlation between each task. Hyperparameters used in the multi-task learning model can be optimized through grid search. The multi-task learning model based on hard parameter sharing can include a shared layer (410) and a task-specific layer (420) (or a task-specific layer). The shared layer (410) is a layer that extracts feature data that can be commonly used for learning in multiple tasks. The task-specific layer (420) is a layer that is individually learned according to the purpose of each task based on the feature data extracted from the shared layer, and outputs a value (or probability value) according to the purpose of each task.

[0067] Learning data (401) can be input into a shared layer (410). The shared layer (410) can include, but is not limited to, a first layer (411), a second layer (412), and a third layer (413). Learning data (401) is input into the first layer (411) of the shared layer (410), and the output value of the first layer (411) can be input as an input value of the second layer (412). The output value of the second layer (412) can be input as an input value of the third layer (413), and the output value of the third layer (413) can be transferred to a task-specific layer (420). The weights of each layer included in the shared layer (410) can be updated by backpropagation. The weights of each layer included in the shared layer (410) can be updated based on the total loss. Since the weight update of the shared layers (410) uses the total loss based on the output values ​​of all tasks, it is possible to train an antibiotic resistance estimation model using common features useful for multiple tasks.

[0068] The task-specific layer (420) may include, but is not limited to, a first task layer (421), a second task layer (422), and a third task layer (423). The output value of the third layer (413) may be transmitted to each of the first task layer (421), the second task layer (422), and the third task layer (423). The loss values ​​of the task-specific layers (420) may be used for learning of each task-specific layer (420). Unlike learning in the shared layers (410), detailed features of each task can be learned by using each task-specific loss only in each task layer rather than using the entire loss.

[0069] The learning device can train an antibiotic resistance estimation model that estimates resistance to antibiotic A, resistance to antibiotic B, and resistance to antibiotic C through the above process.

[0070]

[0071] Figure 5 is a diagram illustrating a process for learning an antibiotic resistance estimation model according to one embodiment.

[0072] Training of an antibiotic resistance estimation model can be performed by a learning device (e.g., the learning device (200) of FIG. 2 ). The antibiotic resistance estimation model is not limited to a single model as described in FIG. 4, and may be an MTL model based on multi-task learning. A general description of the multi-task learning method is omitted below as it would be redundant.

[0073] Referring to FIG. 5, a learning device can learn an antibiotic resistance estimation model (e.g., an MTL model) using a multi-task learning method based on soft weight sharing. Unlike the multi-task learning method based on hard weight sharing of FIG. 4, the multi-task learning method based on soft parameter (weight) sharing may not include a shared layer (e.g., the shared layer (410) of FIG. 4). The multi-task learning method based on soft parameter sharing is a multi-task learning method that has independent machine learning models for each task and mutually adjusts the respective weights of some layers of the independent machine learning models.

[0074] According to one embodiment, learning data (511) is input to a first layer (512), the output of the first layer (512) is input to a second layer (513), the output of the second layer (513) is input to a third layer (514), the output of the third layer (514) is input to a first working layer (515), and the first working layer (515) can output a prediction value for resistance to antibiotic A.

[0075] According to one embodiment, learning data (511) is input to a first layer (522), the output of the first layer (522) is input to a second layer (523), the output of the second layer (523) is input to a third layer (524), the output of the third layer (524) is input to a second working layer (525), and the second working layer (525) can output an estimated value for the resistance of antibiotic B.

[0076] In the above process, the weight of the first layer (512) used to estimate antibiotic A resistance may be adjusted to the weight of the first layer (522) used to estimate antibiotic B resistance, the weight of the second layer (513) used to estimate antibiotic A resistance may be adjusted to the weight of the second layer (523) used to estimate antibiotic B resistance, and the weight of the third layer (514) used to estimate antibiotic A resistance may be adjusted to the weight of the third layer (524) used to estimate antibiotic B resistance. The adjustment of the weights for each layer may be adjusted using normalization based on the Euclidean distance between the weights (e.g., L2 normalization) or cosine similarity based on the directional similarity between two weights.

[0077]

[0078] FIG. 6 is a flowchart illustrating the operations of an antibiotic resistance estimation method using an antibiotic resistance estimation model according to one embodiment.

[0079] The operations of the antibiotic resistance estimation method can be performed by an antibiotic resistance estimation device (e.g., the antibiotic resistance estimation device (100) of FIG. 1).

[0080] In operation (610), the antibiotic resistance estimation device may receive user information data, vital signs data, and antibiotic profile data. The user information data may include personal information of the user who is the subject of antibiotic resistance testing. For example, the user information data may include personal information such as the user's name, age, and address. Additionally, the user information data may include risk index data indicating a risk index based on the user's underlying disease.

[0081] Vital signs data represents data obtained by measuring a user's vital signs. For example, vital signs data may include the user's body temperature, blood pressure, pulse, respiration, or average values ​​for each. Antibiotic profile data represents data indicating a user's antibiotic use history. For example, antibiotic profile data may indicate antibiotic use history for Piperacillin, Tazobactam, or Meropenem.

[0082] In operation (620), the antibiotic resistance estimation device can replace missing values ​​and merge data with replaced missing values.

[0083] The antibiotic resistance estimation device can impute missing values ​​in user information data, vital signs data, and antibiotic profile data. For example, the antibiotic resistance estimation device can impute missing values ​​based on multiple imputation by chained equations (MICE). The antibiotic resistance estimation device can merge data with missing values ​​imputed through the above process.

[0084] In operation (630), the antibiotic resistance estimation device may extract feature data from the merged data. For example, the antibiotic resistance estimation device may extract feature data based on Shapley additive explanations (SHAP) values ​​from the merged data. The feature data extracted by the antibiotic resistance estimation device may include at least resistance data, administration frequency data, risk index data, and the user's hospitalization period data. The resistance data is data indicating whether or not there is resistance to existing antibiotics. The administration frequency data is data indicating the number of administrations of antibiotics used by the user.

[0085] In operation (640), the antibiotic resistance estimation device can obtain result data from the antibiotic resistance estimation model. The antibiotic resistance estimation device can obtain result data including resistance estimation information for the user's target antibiotic from the antibiotic resistance estimation model by inputting feature data into the learned antibiotic resistance estimation model. For example, the antibiotic resistance estimation device can obtain result data regarding the user's antibiotic resistance estimation from an antibiotic estimation model based on at least one of a logistic regression (LR) model, an extreme gradient boosting (XGBoost) model, a light gradient boosting machine (LGBBM) model, a CatBoost model, a multi-layer perceptron (MLP) model, and a multi-task learning (MTL) model.

[0086]

[0087] Figure 7 is a flowchart illustrating the operations of a learning method for learning an antibiotic resistance estimation model according to one embodiment.

[0088] The operations of a learning method for learning an antibiotic resistance estimation model can be performed by a learning device (e.g., the learning device (200) of FIG. 2).

[0089] In operation (710), the learning device may receive user information learning data, vital signs learning data, and antibiotic profile learning data. The user information learning data represents user information of a user who is a subject of antibiotic resistance estimation used to train an antibiotic resistance estimation model. The user information learning data may include risk index data representing a risk index based on the user's underlying disease.

[0090] Vital sign learning data is data obtained by measuring vital signs to train an antibiotic resistance estimation model. The vital sign learning data may include at least the average values ​​for the user's body temperature, blood pressure, and maximum respiration.

[0091] Antibiotic profile training data represents antibiotic use history, used to train an antibiotic resistance estimation model. Antibiotic profile training data may include prescription data for target antibiotics and culture test results indicating resistance to the target antibiotic.

[0092] In operation (720), the learning device can impute missing values ​​and merge learning data with the missing values ​​replaced. The learning device can impute missing values ​​in user information learning data, vital signs learning data, and antibiotic profile learning data and merge the learning data with the missing values ​​replaced. For example, the learning device can impute missing values ​​based on multiple imputation by chained equations (MICE) and merge the learning data with the missing values ​​replaced.

[0093] In operation (730), the learning device may extract feature data from the merged learning data. For example, the learning device may extract feature data from the merged learning data based on Shapley additive explanations (SHAP) values. The extracted feature data may include at least resistance data indicating resistance to existing antibiotics, administration frequency data indicating the number of antibiotic administrations, risk index data, and data on the user's hospitalization period.

[0094] In operation (740), the learning device can train an antibiotic resistance estimation model. The learning device can train the antibiotic resistance estimation model using result data, user information learning data, vital signs learning data, and antibiotic profile learning data obtained from an antibiotic resistance estimation model that inputs feature data. For example, the learning device can train the antibiotic resistance estimation model using a multi-task learning method based on at least one of hard parameter sharing and soft parameter sharing. The multi-task learning method has been described in detail in FIGS. 3 to 5, and thus, a redundant description thereof will be omitted.

[0095] The embodiments described above may be implemented using hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using a general-purpose computer or a special-purpose computer, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and software applications running on the operating system. Furthermore, the processing device may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.

[0096] Software may include computer programs, codes, instructions, or a combination of one or more of these, which may configure a processing device to perform a desired operation or may, independently or collectively, command the processing device. The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium, or device for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on a computer-readable recording medium.

[0097] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may store program commands, data files, data structures, etc., alone or in combination, and the program commands recorded on the medium may be those specially designed and configured for the embodiment or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of program commands include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc.

[0098] The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

[0099] Although the embodiments described above have been described with limited drawings, those skilled in the art will appreciate that various technical modifications and variations can be applied based on the described embodiments. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0100] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims described below.

Claims

1. In an antibiotic resistance estimation device using an antibiotic resistance estimation model, A data receiving unit that receives user information data of a user who is a target of antibiotic resistance estimation, vital signs data indicating vital signs of the user, and antibiotic profile data including antibiotic use history of the user; A data preprocessing unit that replaces missing values ​​of the user information data, the vital signs data, and the antibiotic profile data and merges the data with the missing values ​​replaced; A feature extraction unit for extracting feature data from the above merged data; and An antibiotic resistance estimation unit that obtains result data including resistance estimation information for the user's target antibiotic from the antibiotic resistance estimation model by inputting the feature data into the learned antibiotic resistance estimation model. Including, Antibiotic resistance estimation device.

2. In paragraph 1, The above vital signs data are, At least including average values ​​for the user's body temperature, the user's blood pressure and pulse, and the user's maximum respiration value, Antibiotic resistance estimation device.

3. In paragraph 1, The above user information data is, Contains risk index data indicating a risk index based on the user's underlying disease, Antibiotic resistance estimation device.

4. In paragraph 1, The data preprocessing unit is Imputing the missing values ​​based on MICE (multiple imputation by chained equations; MICE), Antibiotic resistance estimation device.

5. In paragraph 1, The above antibiotic resistance estimation model is, Based on at least one of a logistic regression model (LR), an extreme gradient boosting (XGBoost) model, a light gradient boosting machine (LGBBM) model, a CatBoost model, a multi-layer perceptron (MLP) model, and a multi task learning (MTL) model, Antibiotic resistance estimation device.

6. In paragraph 1, The above feature extraction unit, Extracting the feature data based on SHAP (Shapley additive explanations)-value, Antibiotic resistance estimation device.

7. In paragraph 3, The above feature data is, At least, it includes resistance data indicating whether there is resistance to existing antibiotics, administration frequency data indicating the number of times the antibiotic is administered, the risk index data, and data on the period of hospitalization of the user. Antibiotic resistance estimation device.

8. In a learning device that learns an antibiotic resistance estimation model, A data receiving unit that receives user information learning data of a user who is a target of antibiotic resistance estimation, vital signs learning data representing vital signs of the user, and antibiotic profile learning data representing antibiotic use history of the user; A data preprocessing unit that replaces missing values ​​of the user information learning data, the vital signs learning data, and the antibiotic profile learning data, and merges the learning data with the missing values ​​replaced; A feature extraction unit for extracting feature data from the above merged learning data; and An antibiotic resistance estimation model learning unit that learns the antibiotic resistance estimation model by using the result data obtained from the antibiotic resistance estimation model that inputs the above feature data, the user information learning data, the vital signs learning data, and the antibiotic profile learning data. Including, Learning device.

9. In paragraph 8, The above vital signs learning data is, At least the average values ​​for the user's body temperature, the user's blood pressure, and the user's maximum respiration value, respectively, are included. The above antibiotic profile learning data is, Containing prescription data of a target antibiotic and culture test result data indicating whether or not the target antibiotic is resistant. Learning device.

10. In paragraph 8, The above user information learning data is, Contains risk index data indicating a risk index based on the user's underlying disease, Learning device.

11. In paragraph 8, The data preprocessing unit is Imputing the missing values ​​based on MICE (multiple imputation by chained equations; MICE), Learning device.

12. In paragraph 8, The above feature extraction unit, Extracting the feature data based on SHAP (Shapley additive explanations)-value, Learning device.

13. In paragraph 10, The above feature data is, At least, it includes resistance data indicating whether there is resistance to existing antibiotics, administration frequency data indicating the number of times the antibiotic is administered, the risk index data, and data on the period of hospitalization of the user. Learning device.

14. In paragraph 8, The above antibiotic resistance estimation model is, Based on at least one of a logistic regression model (LR), an extreme gradient boosting (XGBoost) model, a light gradient boosting machine (LGBBM) model, a CatBoost model, a multi-layer perceptron (MLP) model, and a multi task learning (MTL) model, Learning device.

15. In paragraph 8, The above antibiotic resistance estimation model learning unit is, Training the antibiotic resistance estimation model using a multi-task learning method based on at least one of hard-parameter sharing and soft-parameter sharing. Learning device.

16. In a method for estimating antibiotic resistance using an antibiotic resistance estimation model, An action of receiving user information data of a user who is a target of antibiotic resistance estimation, vital signs data representing vital signs of the user, and antibiotic profile data including antibiotic use history for the user; An operation of replacing missing values ​​of the user information data, the vital signs data and the antibiotic profile data and merging the data with the replaced missing values; An operation of extracting feature data from the above merged data; and An operation of obtaining result data including resistance estimation information for the user's target antibiotic from the antibiotic resistance estimation model by inputting the feature data into the learned antibiotic resistance estimation model. Including, Methods for estimating antibiotic resistance.

17. In paragraph 16, The above user information data is, It includes risk index data indicating a risk index based on the user's underlying disease, The above feature data is, At least, it includes resistance data indicating whether there is resistance to existing antibiotics, administration frequency data indicating the number of times the antibiotic is administered, the risk index data, and data on the period of hospitalization of the user. Methods for estimating antibiotic resistance.

18. In a learning method for learning an antibiotic resistance estimation model, An operation of receiving user information learning data of a user who is a target of antibiotic resistance estimation, vital signs learning data representing vital signs of the user, and antibiotic profile learning data representing antibiotic use history of the user; An operation of replacing missing values ​​of the user information learning data, the vital signs learning data, and the antibiotic profile learning data, and merging the learning data with the replaced missing values; An operation of extracting feature data from the above merged learning data; and An operation of training the antibiotic resistance estimation model using the result data obtained from the antibiotic resistance estimation model that input the above feature data, the user information learning data, the vital signs learning data, and the antibiotic profile learning data. Including, How to learn.

19. In paragraph 18, The above learning behavior is, An operation of training the antibiotic resistance estimation model using a multi-task learning method based on at least one of hard-parameter sharing and soft-parameter sharing. Including, How to learn.

20. A computer program stored on a computer-readable recording medium to execute the method of claim 16 in combination with hardware.

Images