Method and apparatus for predicting patient status

Abstract

According to one embodiment of the present disclosure, a method for predicting a patient's condition performed by a computing device is disclosed. The method may include at least one processor included in the computing device receiving input data for a plurality of measurement items; determining a substitute value for a measurement item among the plurality of measurement items for which a measurement value is absent; and inputting the measurement value included in the input data and the substitute value into a prediction model to produce a prediction result.

Description

Technology Field

[0001] The present invention relates to a method for predicting a patient's condition, and more specifically, to a method for predicting a patient's condition based on an artificial neural network model and missing information. Background Technology

[0003] In the medical field, early prediction of a patient's clinical deterioration is a critical task for improving patient safety and treatment outcomes. In particular, if it is possible to predict in advance when a patient admitted to a general ward rapidly deteriorates to the point of unplanned transfer to the intensive care unit, cardiac arrest, or death, medical staff can intervene in a timely manner to prevent these adverse events or minimize their impact.

[0004] However, existing methods for predicting patient status relied solely on the measured values ​​themselves, such as vital signs or blood test results, and had limitations in effectively handling missing data that frequently occurs in clinical settings. In clinical practice, not all tests are performed on every patient; additional tests are conducted only when medical staff assess the patient's condition and determine that they are necessary. Therefore, while the mere fact that a specific test has been performed contains important information regarding the patient's condition, existing prediction methods have failed to utilize this information.

[0005] Therefore, there is an increasing need for new prediction methods and devices that can improve the accuracy of patient condition prediction.

[0006] Korean Patent No. 10-2772326 discloses a user-customized disease prediction system and method using an artificial intelligence model. The problem to be solved

[0008] The present disclosure is conceived in response to the aforementioned background technology and aims to provide a method for predicting a patient's condition. means of solving the problem

[0010] According to one embodiment of the present disclosure for realizing the aforementioned objectives, a method for predicting a patient's condition performed by a computing device is disclosed. The method may include the step of at least one processor included in the computing device receiving input data for a plurality of measurement items; the step of determining a substitute value for a measurement item among the plurality of measurement items for which a measurement value is absent; and the step of inputting the measurement value included in the input data and the substitute value into a prediction model to produce a prediction result.

[0011] In one embodiment, the substitute value may be the clinical reference normal value or the median of the normal range of the corresponding measurement item.

[0012] In one embodiment, the prediction model can learn information regarding whether the measurement value is missing through the distribution pattern of the replacement value.

[0013] In one embodiment, the step of determining the replacement value may be performed by applying the previous time point measurement value if the previous time point measurement value of the corresponding measurement item exists, and by applying the clinical reference normal value or the median of the normal range if the previous time point measurement value of the corresponding measurement item does not exist.

[0014] In one embodiment, the plurality of measurement items includes essential measurement items and optional measurement items, and the step of determining the alternative value may be performed for the optional measurement items.

[0015] In one embodiment, the essential measurement item includes at least one of systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, body temperature, or age, and the optional measurement item may include at least one of oxygen saturation, Glasgow Coma Scale, or blood test item.

[0016] In one embodiment, the input data is received at a plurality of points in time, and the step of determining the replacement value and the step of calculating can be performed for each of the plurality of points in time.

[0017] In one embodiment, the prediction model may include a neural network that learns the temporal dependency of time series data.

[0018] In one embodiment, the neural network may include a recurrent neural network or an attention-based model.

[0019] In one embodiment, the predicted result is the probability of an adverse event occurring within a predetermined time, and the adverse event may include at least one of unplanned transfer to an intensive care unit, cardiac arrest, or death.

[0020] In one embodiment, the method may further include the step of generating missing pattern information indicating whether a measurement value of each measurement item exists; and the step of using the missing pattern information as an additional input to the prediction model.

[0021] A computer program stored in a computer-readable storage medium according to one embodiment of the present disclosure for realizing the aforementioned objectives, wherein when the computer program is executed by one or more processors, the one or more processors are configured to perform operations for predicting a patient's condition, and the operations may include: receiving input data for a plurality of measurement items; determining a substitute value for a measurement item among the plurality of measurement items for which a measurement value is absent; and inputting the measurement value included in the input data and the substitute value into a prediction model to produce a prediction result.

[0022] A computing device according to one embodiment of the present disclosure for realizing the aforementioned objectives may include at least one processor and memory. The at least one processor may receive input data for a plurality of measurement items, determine a replacement value for a measurement item among the plurality of measurement items for which a measurement value is absent, and input the measurement value included in the input data and the replacement value into a prediction model to produce a prediction result. Effects of the invention

[0024] The present disclosure may provide a method for predicting a patient's condition. Brief explanation of the drawing

[0026] FIG. 1 is a drawing illustrating a system including a server, a user terminal, and a communication network according to one embodiment of the present disclosure. FIG. 2 is a block diagram of a server according to one embodiment of the present disclosure. FIG. 3 is a block diagram of a user terminal according to one embodiment of the present disclosure. FIG. 4 is a flowchart showing the overall flow of a patient condition prediction method according to one embodiment of the present disclosure. FIG. 5 is a diagram illustrating the structure of an artificial neural network according to one embodiment of the present disclosure in an exemplary manner. Figure 6 is a diagram showing the results of comparing the frequency of missing values ​​for each measurement item in the event group (patient group in which clinical deterioration occurred) and the non-event group (patient group in which clinical deterioration did not occur). FIG. 7 is a drawing showing alternative values ​​for each measurement item according to one embodiment of the present disclosure. FIG. 8 is a diagram showing the structure of a prediction model according to one embodiment of the present disclosure. FIG. 9 is a drawing showing a predicted effect according to one embodiment of the present disclosure. FIG. 10 is a block diagram of a computing device according to one embodiment of the present disclosure. Specific details for implementing the invention

[0027] Various embodiments are now described with reference to the drawings. In this specification, various descriptions are provided to provide an understanding of the present disclosure. However, it is evident that these embodiments can be practiced without such specific descriptions.

[0028] As used herein, terms such as “component,” “module,” “system,” etc. refer to computer-related entities, hardware, firmware, software, combinations of software and hardware, or executions of software. For example, a component may be, but is not limited to, a procedure executed on a processor, a processor, an object, an execution thread, a program, and / or a computer. For example, both an application executed on a computing device and the computing device itself may be a component. One or more components may reside within a processor and / or an execution thread. A component may be localized within a single computer. A component may be distributed among two or more computers. Additionally, these components may be executed from various computer-readable media having various data structures stored therein. Components may communicate through local and / or remote processes, for example, according to signals having one or more data packets (e.g., data from a component interacting with another component in a local system or distributed system, and / or data transmitted through signals to other systems and networks such as the Internet).

[0029] Furthermore, the term "or" is intended to mean an implicit "or" rather than an exclusive "or." That is, unless otherwise specified or evident from the context, "X uses A or B" is intended to mean one of the natural implicit substitutions. In other words, if X uses A; if X uses B; or if X uses both A and B, "X uses A or B" may apply to any of these cases. Additionally, the term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the enumerated related items.

[0030] Additionally, the terms “comprising” and / or “comprising” should be understood to mean that such features and / or components are present. However, the terms “comprising” and / or “comprising” should be understood not to exclude the presence or addition of one or more other features, components and / or groups thereof. Furthermore, unless otherwise specified or clearly evident from the context to indicate a singular form, the singular in this specification and claims should generally be interpreted to mean “one or more.”

[0031] Expressions such as "first," "second," or "first," "second" used in this document are used to distinguish one object from another when referring to multiple objects of the same kind, unless otherwise indicated in the context, and do not limit the order or importance of said objects. For example, multiple artificial neural network models according to the present disclosure may be distinguished from one another by being expressed as "first artificial neural network model," "second artificial neural network model," and so on.

[0032] The term “at least one of A or B” as used in this document should be interpreted to mean “a case including only A,” “a case including only B,” or “a combination of A and B.”

[0033] As used in this document, the term “part” may refer to software or hardware components such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application Specific Integrated Circuits). However, “part” is not limited to hardware and software. “Part” may be configured to be stored on an addressable storage medium or configured to execute one or more processors. In one embodiment, “part” may include components such as software components, object-oriented software components, class components, and task components, as well as processors, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables.

[0034] The expression "based on" as used in this document is used to describe one or more factors influencing an act or action of a decision or judgment described in the phrase or sentence containing such expression, and this expression does not exclude additional factors influencing said act or action of a decision or judgment.

[0035] As used in this document, the expression that a certain component (e.g., a first component) is "connected" or "connected" to another component (e.g., a second component) may mean not only that the said certain component is directly connected or connected to the said other component, but also that it is connected or connected through a new other component (e.g., a third component).

[0036] As used in this document, the expression "configured to" may have meanings such as "set to," "capable of," "modified to," "made to," or "able to." This expression is not limited to the meaning of "specifically designed in hardware." For example, a processor configured to perform a specific action may refer to a generic-purpose processor capable of performing that specific action by executing software, or a special-purpose computer structured through programming to perform that specific action.

[0037] Hereinafter, various embodiments of the present disclosure will be described with reference to the attached drawings. In the attached drawings and the description thereof, the same reference numerals may be assigned to identical or substantially equivalent components. Furthermore, in the description of the various embodiments below, the description of identical or corresponding components may be omitted, but this does not mean that such components are not included in the embodiments.

[0038] FIG. 1 is a diagram illustrating a system including a server (100), a user terminal (200), and a communication network (300) according to one embodiment of the present disclosure. The server (100) and the user terminal (200) can give or receive information to each other through the communication network (300).

[0039] The server (100) may be an electronic device of a service provider. The service provider may be the operator of a service that provides the patient condition prediction method disclosed in this specification. The server (100) is an electronic device that transmits information or provides services to a user terminal (200) connected via wired or wireless connection, and may be, for example, an application server, a proxy server, a cloud server, etc.

[0040] The user terminal (200) may be a terminal of a user who uses a service regarding a patient condition prediction method. The user terminal (200) may be, for example, at least one of a smartphone, a tablet computer, a PC (Personal Computer), a mobile phone, a PDA (Personal Digital Assistant), an audio player, and a wearable device.

[0041] When describing the configuration or operation of a device in the disclosure of this specification, the term "device" may be used to refer to the device being described, and the term "external device" may be used to refer to a device existing externally from the perspective of the device being described. For example, if the server (100) is described as the "device," the user terminal (200) may be referred to as the "external device" from the perspective of the server (100). Additionally, for example, if the user terminal (200) is described as the "device," the server (100) may be referred to as the "external device" from the perspective of the user terminal (200). That is, the server (100) and the user terminal (200) may each be referred to as the "device" and "external device," or as the "external device" and "device," respectively, depending on the perspective of the operating entity.

[0042] The communication network (300) may include both wired and wireless communication networks. The communication network (300) may operate to exchange data between the server (100) and the user terminal (200). The wired communication network may include, for example, a communication network based on a method such as USB (Universal Serial Bus), HDMI (High Definition Multimedia Interface), RS-232 (Recommended Standard-232), or POTS (Plain Old Telephone Service). A wireless communication network may include, for example, a communication network based on methods such as eMBB (enhanced Mobile Broadband), URLLC (Ultra Reliable Low-Latency Communications), MMTC (Massive Machine Type Communications), LTE (Long-Term Evolution), LTE-A (LTE Advance), NR (New Radio), UMTS (Universal Mobile Telecommunications System), GSM (Global System for Mobile communications), CDMA (Code Division Multiple Access), WCDMA (Wideband CDMA), WiBro (Wireless Broadband), WiFi (Wireless Fidelity), Bluetooth, NFC (Near Field Communication), GPS (Global Positioning System), or GNSS (Global Navigation Satellite System). The communication network (300) of this specification is not limited to the examples described above and may include, without limitation, various types of communication networks that enable data exchange between multiple entities or devices.

[0043] FIG. 2 is a block diagram of a server (100) according to one embodiment of the present disclosure. The server (100) may include one or more processors (110), communication interfaces (120), or memory (130) as components. In some embodiments, at least one of these components of the server (100) may be omitted, or other components may be added to the server (100). In some embodiments, additionally or alternatively, some components may be implemented as an integrated unit or as a single or multiple entity. At least some of the components inside or outside the server (100) may be connected to each other via a bus, GPIO (General Purpose Input / Output), SPI (Serial Peripheral Interface), or MIPI (Mobile Industry Processor Interface) to give or receive data or signals.

[0044] In this specification, one or more processors (110) may be referred to as processors (110). The term processor (110) may mean a set of one or more processors unless the context clearly indicates otherwise. A processor (110) may control at least one component of a server (100) connected to the processor (110) by running software (e.g., instructions, programs, etc.). Additionally, the processor (110) may perform various operations such as computation, processing, data generation, or processing. Additionally, the processor (110) may load data, etc. from memory (130) or store it in memory (130).

[0045] The communication interface (120) can perform wireless or wired communication between the server (100) and another device (e.g., user terminal (200) or another server). For example, the communication interface (120) can perform wireless communication according to methods such as eMBB, URLLC, MMTC, LTE, LTE-A, NR, UMTS, GSM, CDMA, WCDMA, WiBro, WiFi, Bluetooth, NFC, GPS, or GNSS. Additionally, for example, the communication interface (120) can perform wired communication according to methods such as USB (Universal Serial Bus), HDMI (High Definition Multimedia Interface), RS-232 (Recommended Standard-232), or POTS (Plain Old Telephone Service).

[0046] Memory (130) can store various data. Data stored in memory (130) may include software (e.g., instructions, programs, etc.) as data acquired, processed, or used by at least one component of the server (100). Memory (130) may include volatile or non-volatile memory. The term memory (130) may mean a set of one or more memories unless the context clearly indicates otherwise. The expressions "set of instructions stored in memory (130)" or "program stored in memory (130)" mentioned herein may be used to refer to an operating system, an application for controlling the resources of the server (100), or middleware that provides various functions to the application so that the application can utilize the resources of the server (100). In one embodiment, when the processor (110) performs a specific operation, the memory (130) can store instructions that are performed by the processor (110) and correspond to the specific operation.

[0047] In one embodiment, the server (100) may further include an input unit (140). The input unit (140) may be a component that transmits data received from the outside to at least one component included in the server (100). For example, the input unit (140) may include at least one of a mouse, a keyboard, or a touchpad.

[0048] In one embodiment, the server (100) may further include an output unit (150). The output unit (150) may display (output) information processed by the server (100) or transmit (send) it externally. For example, the output unit (150) may visually display information processed by the server (100). The output unit (150) may display UI (User Interface) information or GUI (Graphic User Interface) information, etc. In this case, the output unit (150) may include at least one of a Liquid Crystal Display (LCD), a Thin Film Transistor-Liquid Crystal Display (TFT-LCD), an Organic Light-Emitting Diode (OLED), a Flexible Display, a 3D Display, and an E-ink Display. Additionally, for example, the output unit (150) may audibly display information processed by the server (100). The output unit (150) can display audio data following any audio file format (e.g., MP3, FLAC, WAV, etc.) through an audio device. In this case, the output unit (150) may include at least one of a speaker, a headset, or headphones. Additionally, for example, the output unit (150) may transmit information processed by the server (100) to an external output device. The output unit (150) may transmit or send information processed by the server (100) to an external output device using a communication interface (120). The output unit (150) may also transmit or send information processed by the server (100) to an external output device using a separate output communication interface.

[0049] FIG. 3 is a block diagram of a user terminal (200) according to one embodiment of the present disclosure. The user terminal (200) may include one or more processors (210), a communication interface (220), a memory (230), an input unit (240), or an output unit (250) as components. In some embodiments, at least one of these components of the user terminal (200) may be omitted, or other components may be added to the user terminal (200). In some embodiments, additionally or alternatively, some components may be implemented by being integrated, or implemented as a single or multiple entities. At least some of the components inside or outside the user terminal (200) may be connected to each other via a bus, GPIO (General Purpose Input / Output), SPI (Serial Peripheral Interface), or MIPI (Mobile Industry Processor Interface), etc., to give or receive data or signals.

[0050] The processor (210) can control at least one component of a user terminal (200) connected to the processor (210) by running software (e.g., instructions, programs, etc.). Additionally, the processor (210) can perform various operations such as computation, processing, data generation, or processing. Furthermore, the processor (210) can load data, etc. from memory (230) or store it in memory (230).

[0051] The communication interface (220) can perform wireless or wired communication between a user terminal (200) and another device (e.g., a server (100) or another user terminal). For example, the communication interface (220) can perform wireless communication according to methods such as eMBB, URLLC, MMTC, LTE, LTE-A, NR, UMTS, GSM, CDMA, WCDMA, WiBro, WiFi, Bluetooth, NFC, GPS, or GNSS. Additionally, for example, the communication interface (220) can perform wired communication according to methods such as USB, HDMI, RS-232, or POTS.

[0052] Memory (230) can store various data. Data stored in memory (230) may include software (e.g., instructions, programs, etc.) as data acquired, processed, or used by at least one component of the user terminal (200). Memory (230) may include volatile or non-volatile memory. Unless the context clearly indicates otherwise, the term memory (230) may mean a set of one or more memories. The expressions "set of instructions stored in memory (230)" or "program stored in memory (230)" mentioned in this specification may be used to refer to an operating system, an application for controlling the resources of the user terminal (200), or middleware that provides various functions to the application so that the application can utilize the resources of the user terminal (200). In one embodiment, when the processor (210) performs a specific operation, the memory (230) can store instructions that are performed by the processor (210) and correspond to the specific operation.

[0053] In one embodiment, the user terminal (200) may further include an input unit (240). The input unit (240) may be a component that transmits data received from an external source to at least one component included in the user terminal (200). For example, the input unit (240) may include at least one of a mouse, a keyboard, or a touchpad.

[0054] In one embodiment, the user terminal (200) may further include an output unit (250). The output unit (250) may display (output) information processed by the user terminal (200) or transmit (send) it externally. For example, the output unit (250) may visually display information processed by the user terminal (200). The output unit (250) may display UI (User Interface) information or GUI (Graphic User Interface) information, etc. In this case, the output unit (250) may include at least one of a Liquid Crystal Display (LCD), a Thin Film Transistor-Liquid Crystal Display (TFT-LCD), an Organic Light-Emitting Diode (OLED), a Flexible Display, a 3D Display, or an E-ink Display. Additionally, for example, the output unit (250) may audibly display information processed by the user terminal (200). The output unit (250) can display audio data following any audio file format (e.g., MP3, FLAC, WAV, etc.) through an audio device. In this case, the output unit (250) may include at least one of a speaker, a headset, or headphones. Additionally, for example, the output unit (250) may transmit information processed at the user terminal (200) to an external output device. The output unit (250) may transmit or send information processed at the user terminal (200) to an external output device using a communication interface (220). The output unit (250) may also transmit or send information processed at the user terminal (200) to an external output device using a separate output communication interface.

[0055] In the following description, it is assumed that the subject of each step according to the present disclosure is the processor (110) of the server, but it is obvious that the processor (210) of the user terminal can perform the same function.

[0056] FIG. 4 is a flowchart showing the overall flow of a patient condition prediction method according to one embodiment of the present disclosure.

[0057] Referring to FIG. 4, a patient condition prediction method according to one embodiment of the present disclosure may include the step of receiving input data for a plurality of measurement items (S410), the step of determining a substitute value for a measurement item for which a measurement value is absent (S420), and the step of inputting the measurement value and the substitute value included in the input data into a prediction model to produce a prediction result (S430).

[0058] Each step of the present disclosure may be performed by at least one processor included in a computing device. The computing device may be a server, a cloud computing system, or a dedicated computing device installed within a medical institution, but is not limited thereto.

[0059] The processor (110) can receive input data for multiple measurement items (S410).

[0060] In this disclosure, "measurement item" refers to a type of quantitative information obtained from a patient and may include vital signs, blood test results, and patient basic information. In this disclosure, "measurement item" may be used interchangeably with "test item."

[0061] In the present disclosure, "input data" refers to data received from an external source for a plurality of measurement items, and a measurement value for each measurement item may exist or be absent. That is, the input data may include both measurement items for which a measurement value exists and measurement items for which a measurement value is absent.

[0062] In this disclosure, "measurement value" refers to a value actually obtained for a specific measurement item. For example, if the measurement item is "heart rate," the measurement value may be obtained in a form such as "80 bpm." In this disclosure, "measurement value" may be used interchangeably with "test result value."

[0063] In this disclosure, "substitute value" means a value determined to be used as input to a prediction model for a measurement item for which a measurement value is absent. The substitute value may be one of a measurement value from a previous time point, a clinical reference normal value, or the median of the normal range.

[0064] In one embodiment, a plurality of measurement items may be classified into mandatory measurement items and optional measurement items.

[0065] In this disclosure, "essential measurements" refers to items typically obtained during the patient monitoring process, meaning items for which measurements exist for most patients. Essential measurements may include, for example, systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate, respiratory rate, body temperature, and the patient's age. These items are generally obtained for almost all patients as part of basic vital sign measurements when a patient is hospitalized or visits an outpatient clinic.

[0066] In the present disclosure, "optional measurement items" refers to test items that are performed selectively at the discretion of medical staff, and which may or may not have a measurement value depending on the patient's condition. Optional measurement items may include, for example, oxygen saturation (SpO2), the Glasgow Coma Scale (GCS), and blood test items.

[0067] The Glasgow Coma Scale (GCS) is a scale used to assess a patient's level of consciousness, which evaluates three items—eye response, verbal response, and motor response—and is expressed as a score between 3 and 15 points.

[0068] Blood test items include, for example, total bilirubin, lactate, creatinine, platelets, pH, sodium, potassium, hematocrit, white blood cells, and bicarbonate (HCO3). - It may include, but is not limited to, C-reactive protein (CRP), etc.

[0069] In the present disclosure, measurements may be obtained at multiple points in time. For example, vital signs may be measured every hour during a patient's hospitalization, and blood tests may be performed at the discretion of medical staff. That is, measurements may be sampled at predetermined time intervals (e.g., 1 hour) and configured as time-series data.

[0070] The processor (110) can determine a replacement value for a measurement item for which a measurement value is missing (S420).

[0071] In a clinical setting, measurement values ​​for all measurement items may not always be present. If a specific test is not performed on a patient, the measurement value for that measurement item will not exist. Accordingly, the processor (110) can determine whether a measurement value is present or absent for each measurement item.

[0072] Figure 6 is a diagram showing the results of comparing the frequency of missing values ​​for each measurement item in the event group (patient group in which clinical deterioration occurred) and the non-event group (patient group in which clinical deterioration did not occur).

[0073] Referring to Figure 6, the results of comparing the frequency of missing values ​​for each measurement item in the event group (patient group with clinical deterioration, n=217) and the non-event group (patient group without clinical deterioration, n=5822) are shown. In the table, each number represents the number of cases where the corresponding measurement item was missing, and the value in parentheses represents the percentage of missing values ​​for the entire group.

[0074] Specifically, for lactate, 4,850 cases (83.30%) were missing in the non-event group, whereas only 16 cases (7.37%) were missing in the event group, confirming that the measurement rate in the event group (92.63%) was significantly higher than that in the non-event group (16.70%). Similarly, for pH, 4,814 cases (82.69%) were missing in the non-event group and only 15 cases (6.91%) were missing in the event group, and bicarbonate (HCO₃) In the case of ), 4,814 cases (82.69%) were missing in the non-event group, and only 16 cases (7.37%) were missing in the event group.

[0075] On the other hand, for items such as sodium, potassium, creatinine, hematocrit, white blood cell count, and platelets, there were 0 missing cases (0.00%) in the event group, indicating that the corresponding test was performed on all patients in the event group. In the case of total bilirubin, the missing rate was 0.00% in the event group, whereas 216 cases (3.71%) were missing in the non-event group.

[0076] In the case of C-reactive protein, 2,391 cases (41.07%) were missing in the non-event group and only 2 cases (0.92%) were missing in the event group, so the measurement rate in the event group (99.08%) was higher than the measurement rate in the non-event group (58.93%).

[0077] These results reflect a clinical decision-making pattern in which medical staff perform additional blood tests when they determine there is a possibility that the patient's condition may worsen, and skip unnecessary tests when they determine that the patient's condition is stable. Therefore, the presence or absence (or missing value) of a measurement for a specific measurement item itself contains information regarding the patient's condition, and in this disclosure, this aspect of data is referred to as "informative presence."

[0078] In the present disclosure, the processor (110) can determine a replacement value for a measurement item for which a measurement value is missing.

[0079] In one embodiment, the processor (110) may use the average or median of the measurement values ​​of the item as a substitute value for the measurement item for which a measurement value is missing. For example, if the average value of lactate in the training dataset is calculated to be 1.5 mmol / L, 1.5 mmol / L may be set as a substitute value for patients for whom a lactate measurement value is missing.

[0080] In another embodiment, the processor (110) may use the clinical reference normal value or the median of the normal range of the item as a substitute value for the item for a measurement item for which a measurement value is absent.

[0081] FIG. 7 is a drawing showing alternative values ​​for each measurement item according to one embodiment of the present disclosure.

[0082] Referring to FIG. 7, a substitute value for each measurement item is defined. In the present disclosure, the substitute value may be determined as follows depending on the characteristics of the corresponding measurement item.

[0083] First, if the relevant measurement item is defined as a single normal value, said single normal value may be applied as a surrogate value. For example, since the Glasgow Coma Scale (GCS) has a single value of 15 in a normal state, 15 is applied as a surrogate value when the GCS measurement is absent.

[0084] Similarly, in the case of oxygen saturation (SpO2), since it has a single value of 100% in the steady state, 100% is applied as a substitute value.

[0085] Second, if the relevant measurement item is defined as a normal range, the median of the normal range may be applied as a substitute value. For example, since the normal range for white blood cell count is 3.5 to 10.5 × 10³ / μL, if the white blood cell measurement value is absent, the median of the normal range, 7.0 × 10³ / μL, is applied as a substitute value.

[0086] Similarly, since the normal range for sodium is 135–145 mmol / L, the median value of 140 mmol / L is applied as a substitute value; since the normal range for potassium is 3.6–5.2 mmol / L, the median value of 4.4 mmol / L is applied as a substitute value; and since the normal range for pH is 7.35–7.45, the median value of 7.4 is applied as a substitute value.

[0087] These clinical reference normal values ​​are established based on representative values ​​within the general normal range observed in a population of healthy adults and are applied as surrogates when a measurement for a specific item is absent and previous measurements are also unavailable. The method of applying clinical reference normal values ​​as surrogates reflects the clinical assumption that "if the test had not been performed, the item would presumed to be within the normal range," which aligns with the clinical practice of medical professionals omitting unnecessary tests when they determine that the patient's condition is stable.

[0088] In another embodiment, the processor (110) can determine the replacement value in a hierarchical replacement method according to priority.

[0089] Specifically, for a measurement item for which a measurement value is missing, the processor (110) may first check whether a previous measurement value for the same measurement item of the patient exists. If a previous measurement value exists, the processor (110) applies the measurement value from the previous time closest in time as a substitute value for the current time. For example, if a lactate measurement value is missing at the current time but there is a record of lactate being measured at 2.5 mmol / L at 6 hours ago, the processor may set the lactate substitute value for the current time to 2.5 mmol / L. This method has the advantage of being able to reflect the individual physiological baseline of the patient.

[0090] On the other hand, if there is no previous measurement value of the relevant measurement item for the patient, the processor (110) may apply a predefined clinical reference normal value as a substitute value. In this disclosure, "clinical reference normal value" is a term referring to a representative value within the general normal range of the relevant measurement item observed in a population of healthy adults. For example, 75 beats per minute for heart rate, 36.5 degrees Celsius for body temperature, and 1.0 mmol / L for lactate may be applied.

[0091] The processor (110) of the present disclosure can perform more personalized predictions by setting replacement values ​​for missing values ​​in such a hierarchical replacement manner, thereby compensating for data incompleteness caused by missing values, while prioritizing the use of individual patient history information when it exists.

[0092] Referring again to FIG. 4, the processor (110) can input the measurement value included in the input data and the replacement value into a prediction model to produce a prediction result (S430).

[0093] In the present disclosure, the prediction model may be an artificial neural network model.

[0094] FIG. 5 is a diagram illustrating the structure of an artificial neural network according to one embodiment of the present disclosure in an exemplary manner.

[0095] Throughout this specification, terms such as neural network, artificial neural network, network function, and neural network may be used interchangeably. An artificial neural network may be composed of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is composed of at least one node. The nodes (or neurons) constituting the neural networks may be interconnected by one or more links.

[0096] In a neural network, one or more nodes connected via links can form relative input and output node relationships. The concepts of input and output nodes are relative; any node in an output node relationship with respect to one node may be in an input node relationship with respect to another node, and vice versa. As described above, the input node versus output node relationship can be generated based on links. One or more output nodes may be connected to a single input node via links, and vice versa.

[0097] In a relationship between an input node and an output node connected through a single link, the value of the output node's data can be determined based on the data input to the input node. Here, the link interconnecting the input node and the output node may have a weight. The weight can be variable and can be varied by the user or an algorithm to enable the neural network to perform the desired function. For example, if one or more input nodes are interconnected to a single output node by respective links, the output node's value can be determined based on the values ​​input to the input nodes connected to the output node and the weights set on the links corresponding to each input node.

[0098] As described above, a neural network consists of one or more nodes interconnected through one or more links, forming input-output node relationships within the network. The characteristics of a neural network can be determined by the number of nodes and links within the network, the relationships between the nodes and links, and the weight values ​​assigned to each link. For example, if two neural networks exist with the same number of nodes and links but different weight values ​​for the links, the two neural networks may be recognized as different from each other.

[0099] A neural network can be composed of a set of one or more nodes. A subset of nodes constituting a neural network can form a layer. Some of the nodes constituting a neural network can form a layer based on their distances from an initial input node. For example, a set of nodes with a distance of n from an initial input node can form n layers. The distance from the initial input node can be defined by the minimum number of links that must be traversed to reach that node from the initial input node. However, this definition of a layer is arbitrary for illustrative purposes, and the degree of a layer within a neural network can be defined in a way different from that described above. For example, a layer of nodes may be defined by its distance from a final output node.

[0100] Initial input nodes may refer to one or more nodes within a neural network to which data is directly input without passing through links in their relationships with other nodes. Alternatively, in terms of link-based relationships between nodes within the neural network, they may refer to nodes that do not have other input nodes connected by links. Similarly, final output nodes may refer to one or more nodes within a neural network that do not have output nodes in their relationships with other nodes. Furthermore, hidden nodes may refer to nodes constituting the neural network that are neither initial input nodes nor final output nodes.

[0101] A neural network according to one embodiment of the present disclosure may have the number of nodes in the input layer equal to the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases and then increases again as it progresses from the input layer to the hidden layer. Additionally, a neural network according to another embodiment of the present disclosure may have the number of nodes in the input layer less than the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases as it progresses from the input layer to the hidden layer. Additionally, a neural network according to yet another embodiment of the present disclosure may have the number of nodes in the input layer greater than the number of nodes in the output layer, and may be a neural network in which the number of nodes increases as it progresses from the input layer to the hidden layer. A neural network according to yet another embodiment of the present disclosure may be a neural network in which the above-described neural networks are combined.

[0102] A deep neural network (DNN) may refer to a neural network that includes multiple hidden layers in addition to input and output layers. Using a deep neural network allows for the identification of latent structures in data. That is, it is possible to identify the latent structures of photos, text, videos, voice, and music (e.g., what objects are present in a photo, what the content and emotions of the text are, what the content and emotions of the voice are, etc.). Deep neural networks may include convolutional neural networks (CNN), recurrent neural networks (RNN), autoencoders, Generative Adversarial Networks (GAN), restricted Boltzmann machines (RBM), deep belief networks (DBN), Q networks, U networks, Siamese networks, Generative Adversarial Networks (GAN), etc. The description of deep neural networks described above is merely illustrative and the present disclosure is not limited thereto.

[0103] In one embodiment of the present disclosure, the network function may include an autoencoder. The autoencoder may be a type of artificial neural network for outputting output data similar to the input data. The autoencoder may include at least one hidden layer, and an odd number of hidden layers may be placed between the input and output layers. The number of nodes in each layer may be reduced from the number of nodes in the input layer to an intermediate layer called a bottleneck layer (encoding), and then expanded symmetrically from the bottleneck layer to the output layer (symmetrically with respect to the input layer). The autoencoder may perform non-linear dimensionality reduction. The number of input and output layers may correspond to the dimension after preprocessing of the input data. In the autoencoder structure, the number of nodes in the hidden layer included in the encoder may have a structure in which the number of nodes decreases as it moves away from the input layer. If the number of nodes in the bottleneck layer (the layer with the fewest nodes located between the encoder and decoder) is too small, a sufficient amount of information may not be transmitted, so it may be maintained at a certain number or more (e.g., more than half of the input layer).

[0104] An artificial neural network model can be trained using at least one of supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning. Training of an artificial neural network model may be a process of applying knowledge to the artificial neural network model to perform a specific action.

[0105] Artificial neural network models can be trained to minimize the error in their output. The training process involves repeatedly inputting training data into the model, calculating the error between the model's output and the target for the training data, and updating the weights of each node in the model by backpropagating the error from the output layer to the input layer in a direction that reduces the error. In the case of supervised learning, training data is used where the correct answer is labeled for each data point (i.e., labeled training data), whereas in the case of unsupervised learning, the correct answer may not be labeled for each training data point. For instance, in the case of supervised learning for data classification, the training data may consist of data where each training point is labeled with a category. Labeled training data is input into the artificial neural network model, and the error can be calculated by comparing the model's output (category) with the labels of the training data. As another example, in the case of unsupervised learning for data classification, the error can be calculated by comparing the input training data with the output of the artificial neural network model. The calculated error is backpropagated in the artificial neural network model (i.e., from the output layer to the input layer), and through backpropagation, the connection weights of each node in each layer of the artificial neural network model can be updated. The amount of change in the connection weights of each updated node can be determined by the learning rate. The computation of the artificial neural network model on the input data and the backpropagation of the error can constitute a learning cycle (epoch). The learning rate can be applied differently depending on the number of iterations of the artificial neural network model's learning cycle. For example, a high learning rate can be used in the early stages of training to quickly achieve a certain level of performance and increase efficiency, while a low learning rate can be used in the later stages to improve accuracy.

[0106] In the training of artificial neural network models, the training data is generally a subset of real-world data (i.e., the data intended to be processed using the trained model). Consequently, a training cycle may exist where errors decrease on the training data but increase on real-world data. Overfitting is a phenomenon where the model learns excessively on the training data, leading to increased errors on real-world data. For example, an artificial neural network model trained on yellow cats may fail to recognize cats other than yellow ones as cats; this can be considered a form of overfitting. Overfitting can act as a cause for increased errors in machine learning algorithms. Various optimization methods can be used to prevent such overfitting. To prevent overfitting, methods such as increasing the training data, regularization, dropout (which disables some nodes in the network during training), and the use of batch normalization layers can be applied.

[0107] The prediction model according to the present disclosure may be a neural network model that learns the temporal dependency of time-series data. Patient clinical data is acquired continuously over time, and not only the measurement value at a specific point in time but also the trend of change in the measurement value over time can provide important information for predicting the patient's condition. For example, a trend in which the heart rate gradually increases from 80 beats per minute to 110 beats per minute may have a different clinical significance than when the same heart rate of 110 beats per minute appears suddenly. A neural network model that learns the temporal dependency of time-series data can capture these patterns of change over time to improve prediction accuracy.

[0108] In one embodiment, the prediction model may include a Recurrent Neural Network (RNN) or an attention-based model.

[0109] Recurrent neural networks are neural network structures suitable for the sequential processing of time-series data, and can model temporal dependencies by processing the hidden state of a previous time point together with the input of the current time point. In one embodiment, the prediction model may include a Long Short-Term Memory (LSTM) or a Bidirectional LSTM (BiLSTM). LSTMs can effectively solve the long-term dependency problem through cell states and gate mechanisms (forget gate, input gate, output gate), thereby enabling the learning of the impact of changes in measurements and missing information from past time points on the prediction of the current time point. BiLSTM is a structure that combines a forward LSTM and a backward LSTM, and has the advantage of being able to consider both the context before and after a specific time point by processing the time series bidirectionally.

[0110] Attention-based models include a mechanism that assigns higher weights to points in a time series that are important for prediction. For example, high attention weights may be assigned to points where a patient's condition changes rapidly or where missing information changes (such as when additional tests are ordered). In one embodiment, the prediction model may include a Transformer architecture or a self-attention mechanism, which allows each point in the time series to directly calculate the relationship with all other points. Transformer-based models have the advantage of enabling parallel processing, resulting in faster learning and inference speeds, and can enhance the interpretability of the model by analyzing attention weights.

[0111] The structure of a prediction model according to one embodiment of the present disclosure will be described below with reference to FIG. 8.

[0112] FIG. 8 is a diagram showing the structure of a prediction model according to one embodiment of the present disclosure. Referring to FIG. 8, the prediction model may be composed of two paths, a time series data processing path and a static data processing path, and the outputs of the two paths are combined to produce a final prediction score.

[0113] A time series data processing path is a path that processes dynamic data acquired at multiple time points. The input to the time series data processing path may be time series data containing vital signs and blood test results (Lab) from multiple time points. Each time point of the time series data may include an input vector containing acquired measurements and set replacement values. For example, if 24 hours of data sampled at 1-hour intervals is input, the time series data consists of input vectors for each of the 24 time points.

[0114] The time series data processing path may include a Bidirectional Long Short-Term Memory Network (BiLSTM) layer. The BiLSTM layer processes the time series data in the forward and backward directions to generate a forward hidden state (h) and a backward hidden state (h~) for each time point. The forward hidden state encodes time series information prior to the corresponding time point, and the backward hidden state encodes time series information after the corresponding time point. Through this, the BiLSTM layer can generate a representation for each time point that considers both the context before and after that time point.

[0115] The time series data processing path may further include an attention layer applied to the output of the BiLSTM layer. The attention layer performs a mechanism that assigns higher weights to points of importance for prediction among multiple time points. For example, high attention weights may be assigned at points where the patient's condition changes rapidly or at points where missing information changes (points where additional tests are ordered). The attention layer generates a context vector representing the entire time series by weighted summing the multiple time-point hidden states, which are the outputs of the BiLSTM layer, based on the attention weights.

[0116] The static data processing path is a path that processes static data that does not change over time. The inputs to the static data processing path may include descriptive data with time-invariant characteristics, such as patient age, and last-time data. The static data processing path may include a batch normalization layer and a dense layer. The batch normalization layer normalizes the input data to improve the stability of the learning process, while the dense layer performs non-linear transformations on the normalized data to extract features.

[0117] The output (context vector) of the time series data processing path and the output of the static data processing path can be concatenated and input into an additional dense layer. The additional dense layer can perform a non-linear transformation on the concatenated features to generate a final representation. The final representation is input into a softmax layer to calculate the probability of an adverse event occurring. The output of the softmax layer has a value between 0 and 1, and this value can be multiplied by 100 to be converted into a prediction score between 0 and 100.

[0118] Through the structure of the prediction model described above, the processor (110) can perform a comprehensive prediction that considers both dynamic changes over time (trends in vital signs, changes in blood test results) and static characteristics of the patient (age, etc.). In particular, through the application of BiLSTM and attention mechanisms, the prediction model can learn important points and patterns for prediction within time series data, and can improve prediction accuracy by effectively utilizing the clinical significance contained in missing information.

[0119] The prediction model of the present disclosure has a structure as described above with reference to FIG. 8, and can learn information regarding whether a measurement value is missing through the distribution pattern of a replacement value and use this to predict the patient's condition.

[0120] Specifically, when a clinical reference normal value or the median of the normal range is applied as a substitute value for a measurement item for which a measurement value is missing, a systematic difference occurs in the distribution pattern of the input vector between the event group and the non-event group. Referring to Figure 6 above, in the non-event group, 83.30% of lactate, 82.69% of pH, and 82.69% of bicarbonate are missing and replaced with normal values, whereas in the event group, only 7.37%, 6.91%, and 7.37%, respectively, are missing and most of the actual values ​​are input.

[0121] These differences in missing values ​​cause the following characteristic differences in the distribution patterns of the input vectors.

[0122] First, the input vector of the non-event group has multiple measurement items with improvised values, namely clinical reference normal values ​​or the median of the normal range. For example, specific values ​​within the normal range appear repeatedly, such as lactate at 0.7 mmol / L, pH at 7.4, and bicarbonate at 24 mmol / L.

[0123] Second, the input vectors of the event group consist mostly of measured items with actual values; these actual values ​​reflect the patient's actual clinical condition and may fall outside the normal range or vary from patient to patient. For example, values ​​outside the normal range, such as lactate at 2.1 mmol / L and pH at 7.31, appear, or various distributions are observed among patients.

[0124] The predictive model learns the differences in the distribution patterns of these input vectors during the training process. Specifically, input patterns where multiple measurement items simultaneously have normal surrogate values ​​are learned to correlate with the information that "the corresponding tests were not performed, which suggests that the patient's condition is stable."

[0125] Conversely, input patterns where most measurements have varying actual values ​​are learned to correlate with the information that "extensive testing was performed, which suggests that medical staff suspect a deterioration in the patient's condition."

[0126] This distribution pattern-based learning method can utilize the clinical significance inherent in informative presence for prediction through the distributional characteristics of the replacement values ​​themselves, even without explicitly encoding missing data in the form of binary vectors or providing them as separate inputs.

[0127] The prediction result calculated by the processor (110) of the present disclosure may be, for example, the probability that an adverse reaction will occur within a predetermined time (e.g., 6 hours).

[0128] In the present disclosure, "adverse event" refers to an event indicating clinical deterioration of a patient and may include at least one of unplanned ICU transfer, cardiac arrest, or death. "Unplanned ICU transfer" means the transfer of a patient admitted to a general ward to an intensive care unit without prior notice, excluding planned transfers where the patient is transferred directly from an emergency room or operating room to an intensive care unit. "Cardiac arrest" means a case where the patient's cardiac function has ceased and cardiopulmonary resuscitation (CPR) is performed. "Death" means the death of a patient during hospitalization.

[0129] The prediction result can be expressed as a score between 0 and 100 or a probability value between 0 and 1, where a higher score indicates a higher probability of an adverse event occurring within a set time. For example, if the prediction score calculated by the processor (110) for a specific patient is 75 points, this means that the patient has a relatively high risk of experiencing an adverse event, such as unplanned transfer to an intensive care unit, cardiac arrest, or death, within the next 6 hours. On the other hand, if the prediction score is 15 points, it suggests that the patient's clinical condition is relatively stable.

[0130] In one embodiment of the present disclosure, the processor (110) can produce a prediction result based additionally on missing pattern information.

[0131] In one embodiment of the present disclosure, the processor (110) can generate missing pattern information indicating whether a measurement value of each measurement item exists.

[0132] In this disclosure, "missing pattern information" refers to information indicating whether a measurement value exists for each of a plurality of measurement items. Missing pattern information may be expressed in various forms. Several embodiments of missing pattern information are described below.

[0133] In one embodiment, missing pattern information may be represented in the form of a vector in which the presence or absence of a measurement value for each measurement item is encoded as a binary value (0 or 1). That is, it may be encoded as "1" if the measurement value exists and as "0" if it does not.

[0134] For example, assume that the optional measurement items consist of five items: oxygen saturation, Glasgow Coma Scale, total bilirubin, lactate, and creatinine. If, at a specific time, only oxygen saturation and creatinine are measured for the patient and the remaining items are not measured, the missing pattern vector can be represented as [1, 0, 0, 0, 1]. Here, the first element "1" indicates the presence of an oxygen saturation measurement, the second element "0" indicates the absence of a Glasgow Coma Scale measurement, and the last element "1" indicates the presence of a creatinine measurement. This binary encoding method converts missing pattern information into a numerical form that can be processed by a computer, allowing it to be used as input for a prediction model.

[0135] In another embodiment, missing pattern information may be expressed as a missing rate over a predetermined period. For example, the missing rate may be defined as the ratio of the number of times a measurement item is actually missing to the total number of times a measurement item should have been measured within a predetermined observation period. For example, if measurements are sampled at 1-hour intervals, there are a total of 24 times over 24 hours, and among these, lactate measurements are present at only 6 times and absent at the remaining 18 times, the missing rate for lactate may be calculated as 18 / 24 = 0.75. This missing rate has a continuous value between 0 and 1, where a value closer to 0 indicates that the measurement item was measured frequently, and a value closer to 1 indicates that the measurement item was rarely measured.

[0136] In another embodiment, missing pattern information may be expressed as the time elapsed since a specific measurement item was last measured. For example, if a lactate test was last performed 12 hours ago, the missing pattern information for that item may be expressed as "12 hours". This time-based missing pattern information reflects the recency of the measurement value, allowing the model to learn the clinical characteristic that older measurements may reflect the current patient condition less.

[0137] In another embodiment, missing pattern information may be expressed as a categorical variable combining the presence or absence of multiple measurement items. For example, the case where pH, bicarbonate (HCO3-), and lactic acid—items related to blood gas tests—are all present can be defined as the first pattern; the case where only pH and bicarbonate are present and lactic acid is absent can be defined as the second pattern; and the case where all three items are absent can be defined as the third pattern. Such categorical missing patterns may reflect the clinical practice of medical professionals prescribing specific tests together.

[0138] Specifically, when a blood gas test is performed, pH, bicarbonate, and lactate tend to be measured simultaneously; therefore, the presence patterns of these items may appear as limited combinations such as [1, 1, 1], [1, 1, 0], or [0, 0, 0]. The processor may assign a unique categorical identifier to each of these combinations, assigning "Pattern_A" to the first pattern (all present), "Pattern_B" to the second pattern (only pH and bicarbonate present), and "Pattern_C" to the third pattern (all absent). Subsequently, the processor can utilize the categorical missing pattern as input to a prediction model by converting it into a numerical vector using one-hot encoding or embedding techniques.

[0139] In another embodiment, missing pattern information may be expressed as the total number of times a specific measurement item was measured within a predetermined observation period. For example, a patient whose blood pressure was measured six times in the last 24 hours and a patient whose blood pressure was measured only twice are likely to be in different clinical situations, and this difference in measurement frequency may indirectly reflect the instability of the patient's condition or the level of interest of medical staff.

[0140] Specifically, the processor can generate a measurement frequency vector by counting the number of times each measurement item was measured during the last 24 hours. For example, if the optional measurement items consist of three items—oxygen saturation, lactate, and creatinine—and oxygen saturation was measured 12 times, lactate 2 times, and creatinine 1 time during the last 24 hours, the measurement frequency vector can be expressed as [12, 2, 1]. This measurement frequency information can be converted into a value between 0 and 1 through normalization, for example, if the maximum expected number of measurements is set to 24 times, the normalized measurement frequency vector can be expressed as [0.5, 0.083, 0.042].

[0141] As described above, the present disclosure discloses various methods for generating missing pattern information, which function as methods for numerically representing the informational existence described above. Specifically, various forms of missing pattern information, such as missing pattern information in the form of binary vectors, a missing rate over a predetermined period, elapsed time since the last measurement, categorical missing patterns, and measurement frequency vectors, can be utilized as inputs to a prediction model.

[0142] In addition, the various forms of missing pattern information described above may be used individually or in combination. For example, missing pattern information in the form of a binary vector and elapsed time information since the last measurement may be included together in the input vector.

[0143] This missing pattern information numerically encodes information regarding "which tests were performed" or "which tests were not performed," thereby enabling a predictive model to utilize information embedded in the clinical decision-making of medical staff for learning. By receiving this missing pattern information as input along with measurements and replacement values, the predictive model according to the present disclosure can perform a prediction that comprehensively considers not only the numerical anomalies of the measurements but also the severity of the patient's condition reflected in the test execution patterns.

[0144] In an embodiment that additionally utilizes missing pattern information, the processor (110) may input an input vector combining missing pattern information, along with acquired measurements and set replacement values, into a prediction model to produce a prediction result.

[0145] Specifically, the input vector may be composed of a concatenation of three components. The first component is an acquired measurement value, which includes the values ​​of items for which actual measurements exist among the essential measurement items and optional measurement items. For example, the first component may include essential measurement items such as systolic blood pressure 120 mmHg, diastolic blood pressure 75 mmHg, heart rate 72 beats per minute, respiratory rate 16 breaths per minute, and body temperature 36.5 degrees Celsius, and optional measurement items such as oxygen saturation 98% and creatinine 0.9 mg / dL which were actually measured.

[0146] The second component is a set substitute value and includes a substitute value set for a measurement item where the measurement value is determined to be absent. As described above, if a previous time point measurement value exists, the corresponding previous time point measurement value (LOCF, Last Observation Carried Forward) is applied as the substitute value, and if a previous time point measurement value does not exist, a predefined clinical reference normal value may be applied. For example, if a lactate measurement value is absent and a previous time point measurement value also does not exist, 0.7 mmol / L, which is the clinical reference normal value for lactate, may be included in the second component.

[0147] The third component is missing pattern information, which includes information indicating whether a measurement value of each measurement item exists. The missing pattern information can be expressed, for example, in the form of a binary vector, where items with a measurement value exist are encoded as "1" and items without a measurement value are encoded as "0". For example, if among 11 optional measurement items only oxygen saturation, creatinine, hematocrit, leukocytes, and platelets are measured and the remaining items (lactic acid, pH, bicarbonate, total bilirubin, sodium, potassium) are not measured, the missing pattern vector can be expressed as [1, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0].

[0148] An input vector combining these three components can be input into a prediction model in the following form. For example, in the case of patient A belonging to the event group, most tests have been performed, so the input vector can be constructed as [90, 55, 110, 28, 38.2, 72, 92, 3.2, 7.31, 2.1, 24.5, 1.2, 140, 4.5, 32, 12.5, 180, 2.8, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]. Here, the numerical values ​​in the first part are the acquired measurements and replacement values, and the binary values ​​in the second part are missing pattern vectors. On the other hand, for patient B, who belongs to the non-event group, only some tests were performed, so the input vector may be composed as [120, 75, 72, 16, 36.5, 45, 98, 0.7, 7.4, 0.8, 24, 0.6, 140, 4.2, 42, 6.5, 250, 1.5, 1, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0]. The missing pattern vector for patient B contains multiple "0s," indicating that tests for lactate, pH, bicarbonate, etc., were not performed.

[0149] This method of constructing input vectors provides the following technical benefits. First, by including missing pattern information as a separate feature, the prediction model can utilize information regarding "which tests were performed" for training. As mentioned above, since the very fact of whether a test was performed contains information about the patient's condition, the inclusion of missing pattern information can contribute to improving prediction performance. Second, because items with set surrogates and items with actual values ​​can be distinguished through missing pattern vectors, the prediction model can process actual values ​​and surrogates differently. Third, since the dimensionality of the input vector is fixed, consistent input can be provided to the prediction model regardless of the presence of missing values.

[0150] The processor (110) can input measurement values, replacement values, and missing pattern information obtained at multiple time points (e.g., t1, t2, t3, ..., tn) into the prediction model in the form of a time series. For example, if 24 hours of data sampled at 1-hour intervals is input, the prediction model can learn pattern changes over time by sequentially processing the input vectors for each of the 24 time points. Through this, the prediction model can perform predictions that reflect not only snapshot information of a single time point but also dynamic changes in the patient's condition.

[0151] FIG. 9 is a diagram illustrating a prediction effect according to one embodiment of the present disclosure. The graph in FIG. 9 shows a Receiver Operating Characteristic (ROC) curve, where the x-axis of the ROC curve represents 1-specificity (False Positive Rate) and the y-axis represents sensitivity (True Positive Rate, Sensitivity).

[0152] Referring to Figure 9, the predictive performance of a model using only vital signs as input (VitalSign Only) and a model that additionally inputs the status of missing blood tests in the form of a separate vector to the vital signs (VitalSign + Lab Pattern) is compared. Specifically, the AUC (Area Under Curve) of the model using only vital signs was 0.831 (95% confidence interval: 0.828-0.834), whereas the AUC of the model that added the status of missing data in the form of a separate vector was 0.849 (95% confidence interval: 0.846-0.851). These results indicate that including missing pattern information as input to the prediction model contributes to improving prediction performance.

[0153] That is, information regarding "what tests were performed" provides meaningful information for predicting the patient's condition, and the prediction model according to the present disclosure can achieve improved prediction accuracy compared to existing methods that use only measurement values ​​by utilizing this missing pattern information.

[0154] FIG. 10 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0155] FIG. 10 illustrates a brief and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented.

[0156] Although the present disclosure has generally been described in relation to computer-executable instructions that can be executed on one or more computers, those skilled in the art will know that the present disclosure may be combined with other program modules and / or implemented as a combination of hardware and software.

[0157] Generally, a program module includes routines, programs, components, data structures, etc., that perform a specific task or implement a specific abstract data type. Furthermore, those skilled in the art will be well aware that the method of the present disclosure may be implemented in other computer system configurations, including single-processor or multi-processor computer systems, minicomputers, mainframe computers, as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, etc. (each of which may be connected to and operated with one or more associated devices).

[0158] The embodiments described in this disclosure may also be implemented in a distributed computing environment in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

[0159] Computers typically include various computer-readable media. Any medium accessible by a computer may be a computer-readable medium. Computer-readable media include volatile and non-volatile media, transitory and non-transitory media, and removable and non-removable media. By example, but not by limitation, computer-readable media may include computer-readable storage media and computer-readable transmission media. Computer-readable storage media include volatile and non-volatile media, transitory and non-transitory media, and removable and non-removable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, DVD (digital video disk) or other optical disk storage devices, magnetic cassettes, magnetic tapes, magnetic disk storage devices or other magnetic storage devices, or any other media that can be accessed by a computer and used to store desired information.

[0160] Computer-readable transmission media typically include all information transmission media that implement computer-readable instructions, data structures, program modules, or other data, etc., on a modulated data signal, such as other transport mechanisms. The term modulated data signal means a signal in which one or more of the characteristics of the signal are set or modified to encode information within the signal. By example, not limiting, computer-readable transmission media include wired media such as wired networks or direct-wired connections, and wireless media such as acoustic, RF, infrared, and other wireless media. Any combination of the media described above is also considered to be included within the scope of computer-readable transmission media.

[0161] An exemplary environment (1100) for implementing various aspects of the present disclosure, including a computer (1102), is shown, wherein the computer (1102) includes a processing unit (1104), a system memory (1106), and a system bus (1108). The system bus (1108) connects system components, including the system memory (1106) (but not limited thereto), to the processing unit (1104). The processing unit (1104) may be any processor among various commercial processors. Dual processors and other multiprocessor architectures may also be used as the processing unit (1104).

[0162] The system bus (1108) may be any of several types of bus structures that can be additionally interconnected to a local bus using any of the memory bus, peripheral bus, and various commercial bus architectures. System memory (1106) includes read-only memory (ROM) (1110) and random access memory (RAM) (1112). The basic input / output system (BIOS) is stored in non-volatile memory (1110), such as ROM, EPROM, EEPROM, etc., and this BIOS includes basic routines that help transfer information between components within the computer (1102) at times such as during startup. The RAM (1112) may also include high-speed RAM, such as static RAM, for caching data.

[0163] The computer (1102) also includes an internal hard disk drive (HDD) (1114) (e.g., EIDE, SATA)—this internal hard disk drive (1114) may also be configured for external use within a suitable chassis (not shown)—a magnetic floppy disk drive (FDD) (1116) (e.g., for reading from or writing to a removable diskette (1118)), and an optical disk drive (1120) (e.g., for reading from a CD-ROM disk (1122) or reading from or writing to other high-capacity optical media such as a DVD). The hard disk drive (1114), the magnetic disk drive (1116), and the optical disk drive (1120) may each be connected to the system bus (1108) by a hard disk drive interface (1124), a magnetic disk drive interface (1126), and an optical drive interface (1128). The interface (1124) for implementing an external drive includes at least one or both of USB (Universal Serial Bus) and IEEE 1394 interface technologies.

[0164] These drives and associated computer-readable media provide non-volatile storage of data, data structures, computer-executable instructions, etc. In the case of a computer (1102), the drives and media correspond to storing any data in a suitable digital format. Although the description of computer-readable media above refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those skilled in the art will know that other types of computer-readable media, such as zip drives, magnetic cassettes, flash memory cards, cartridges, etc., may also be used in exemplary operating environments and that any of these media may contain computer-executable instructions for performing the methods of the present disclosure.

[0165] A number of program modules, including an operating system (1130), one or more application programs (1132), other program modules (1134), and program data (1136), may be stored in the drive and RAM (1112). All or part of the operating system, application, module and / or data may also be cached in RAM (1112). It will be well known that the present disclosure may be implemented in various commercially available operating systems or combinations of operating systems.

[0166] The user can input commands and information into the computer (1102) through one or more wired / wireless input devices, such as a pointing device like a keyboard (1138) and a mouse (1140). Other input devices (not shown) may include a microphone, an IR remote control, a joystick, a game pad, a stylus pen, a touch screen, etc. These and other input devices are often connected to the processing unit (1104) via an input device interface (1142) connected to the system bus (1108), but may also be connected via other interfaces such as a parallel port, an IEEE 1394 serial port, a game port, a USB port, an IR interface, etc.

[0167] A monitor (1144) or other type of display device is also connected to the system bus (1108) via an interface such as a video adapter (1146). In addition to the monitor (1144), the computer generally includes other peripheral output devices (not shown), such as speakers, a printer, and so on.

[0168] The computer (1102) may operate in a networked environment using a logical connection to one or more remote computers, such as remote computer(s) (1148), via wired and / or wireless communication. The remote computer(s) (1148) may be a workstation, a computing device computer, a router, a personal computer, a portable computer, a microprocessor-based entertainment device, a peer device, or other conventional network node, and generally include many or all of the components described for the computer (1102), but for brevity, only the memory storage device (1150) is illustrated. The illustrated logical connection includes a wired / wireless connection to a local area network (LAN) (1152) and / or a larger network, e.g., a wide area network (WAN) (1154). Such LAN and WAN networking environments are common in offices and companies and facilitate enterprise-wide computer networks, such as intranets, all of which can be connected to a global computer network, e.g., the Internet.

[0169] When used in a LAN networking environment, the computer (1102) is connected to a local network (1152) via a wired and / or wireless communication network interface or adapter (1156). The adapter (1156) may facilitate wired or wireless communication to the LAN (1152), and the LAN (1152) may also include a wireless access point installed therein to communicate with the wireless adapter (1156). When used in a WAN networking environment, the computer (1102) may include a modem (1158), be connected to a communication computing device on the WAN (1154), or have other means to establish communication through the WAN (1154), such as through the Internet. The modem (1158), which may be an internal or external and a wired or wireless device, is connected to the system bus (1108) via a serial port interface (1142). In a networked environment, program modules or parts thereof described for a computer (1102) may be stored in a remote memory / storage device (1150). It will be well known that the illustrated network connection is exemplary and that other means of establishing a communication link between computers may be used.

[0170] The computer (1102) operates to communicate with any wireless device or object that is deployed and operated via wireless communication, for example, a printer, scanner, desktop and / or portable computer, PDA (portable data assistant), communication satellite, any equipment or place associated with a wireless detectable tag, and a telephone. This includes at least Wi-Fi and Bluetooth wireless technologies. Accordingly, the communication may be a predefined structure as in a conventional network, or simply ad hoc communication between at least two devices.

[0171] Wi-Fi (Wireless Fidelity) enables connectivity to the Internet and other sources without wires. Wi-Fi is a wireless technology, similar to a cell phone, that allows devices, such as computers, to transmit and receive data indoors and outdoors—that is, anywhere within the coverage area of ​​a base station. Wi-Fi networks use a wireless technology called IEEE 802.11 (a, b, g, etc.) to provide secure, reliable, and high-speed wireless connections. Wi-Fi can be used to connect computers to each other, to the Internet, and to wired networks (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in unlicensed 2.4 and 5 GHz wireless bands, for example, at data rates of 11 Mbps (802.11a) or 54 Mbps (802.11b), or in products that include both bands (dual band).

[0172] Those skilled in the art of the present disclosure will understand that information and signals may be represented using any various different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced in the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0173] Those skilled in the art to which this disclosure is made will understand that the various exemplary logic blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented by electronic hardware, various forms of programs or design code (referred to herein as “software”), or a combination of all such. To clearly illustrate this interoperability between hardware and software, various exemplary components, blocks, modules, circuits, and steps have been generally described above in relation to their functions. Whether such functions are implemented as hardware or software depends on the design constraints imposed on the specific application and the overall system. Those skilled in the art to which this disclosure is made may implement the functions described in various ways for each specific application, but such implementation decisions should not be interpreted as being outside the scope of this disclosure.

[0174] The various embodiments presented herein may be implemented as methods, devices, or articles manufactured using standard programming and / or engineering techniques. The term “article manufactured” includes a computer program or media accessible from any computer-readable device. For example, computer-readable media include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical discs (e.g., CDs, DVDs, etc.), smart cards, and flash memory devices (e.g., EEPROMs, cards, sticks, key drives, etc.). Additionally, the various storage media presented herein include one or more devices and / or other machine-readable media for storing information.

[0175] It should be understood that the specific order or hierarchy of steps in the presented processes is an example of exemplary approaches. It should be understood that the specific order or hierarchy of steps in the processes may be rearranged within the scope of this disclosure based on design priorities. The appended method claims provide various step elements in a sample order, but do not imply being limited to the specific order or hierarchy presented.

[0176] Description of the presented embodiments is provided so that a person skilled in the art may use or practice the present disclosure. Various modifications to these embodiments will be apparent to a person skilled in the art. The general principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure is not limited to the embodiments presented herein, but should be interpreted in the broadest possible scope consistent with the principles and novel features presented herein.

Claims

Claim 1 A method for predicting a patient's condition performed by a computing device, wherein at least one processor included in the computing device receives input data for a plurality of measurement items; determines a substitute value for a measurement item among the plurality of measurement items for which a measurement value is absent; and inputs the measurement value included in the input data and the substitute value into a prediction model to produce a prediction result; wherein the step of determining the substitute value is performed by applying the previous time point measurement value if the previous time point measurement value of the corresponding measurement item exists, and applying the clinical reference normal value or the median of the normal range if the previous time point measurement value of the corresponding measurement item does not exist. Claim 2 A method for predicting a patient's condition according to claim 1, wherein the substitute value is the clinical reference normal value or the median of the normal range of the corresponding measurement item. Claim 3 A method for predicting a patient's condition according to claim 1, wherein the prediction model learns information regarding whether a measurement value is missing through the distribution pattern of the replacement value. Claim 4 delete Claim 5 A method for predicting a patient's condition according to claim 1, wherein the plurality of measurement items include essential measurement items and optional measurement items, and the step of determining the substitute value is performed on the optional measurement items. Claim 6 A method for predicting a patient's condition according to claim 5, wherein the essential measurement item includes at least one of systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, body temperature, or age, and the optional measurement item includes at least one of oxygen saturation, Glasgow Coma Scale, or blood test item. Claim 7 A method for predicting a patient's condition according to claim 1, wherein the input data is received at a plurality of points in time, and the step of determining the replacement value and the step of calculating are performed for each of the plurality of points in time. Claim 8 A method for predicting a patient's condition according to claim 7, wherein the prediction model includes a neural network that learns the temporal dependency of time series data. Claim 9 A method for predicting a patient's condition according to claim 8, wherein the neural network comprises a recurrent neural network or an attention-based model. Claim 10 A method for predicting a patient's condition according to claim 1, wherein the predicted result is the probability of an adverse reaction occurring within a predetermined time, and the adverse reaction includes at least one of unplanned transfer to an intensive care unit, cardiac arrest, or death. Claim 11 A method for predicting a patient's condition according to claim 1, further comprising: a step of generating missing pattern information indicating whether a measurement value of each measurement item exists; and a step of using the missing pattern information as an additional input to the prediction model. Claim 12 A computer program stored on a computer-readable storage medium, wherein, when the computer program is executed by one or more processors, the one or more processors are configured to perform operations for predicting a patient's condition, the operations include: receiving input data for a plurality of measurement items; determining a substitute value for a measurement item among the plurality of measurement items for which a measurement value is absent; and inputting the measurement value included in the input data and the substitute value into a prediction model to produce a prediction result; wherein the operation of determining the substitute value is performed by applying the previous measurement value if a previous measurement value of the corresponding measurement item exists, and applying a clinical reference normal value or the median of the normal range if a previous measurement value of the corresponding measurement item does not exist. Claim 13 A computing device comprising at least one processor and memory, wherein the at least one processor receives input data for a plurality of measurement items, determines a substitute value for a measurement item among the plurality of measurement items for which a measurement value is absent, and inputs the measurement value included in the input data and the substitute value into a prediction model to produce a prediction result, wherein the determination of the substitute value is performed by applying the previous measurement value if a previous measurement value of the corresponding measurement item exists, and applying the clinical reference normal value or the median of the normal range if a previous measurement value of the corresponding measurement item does not exist.

Images