System and method for early diagnosis and prognosis prediction of sepsis

Abstract

A system for early diagnosis and prognosis prediction of sepsis, according to an embodiment of the present invention, includes: a memory for storing general blood test data of a patient; and a processor, wherein the processor may be configured to pre-process the general blood test data of the patient stored in the memory as data to be input to an artificial intelligence model, input the pre-processed data to the artificial intelligence model trained for early diagnosis and prognosis prediction of sepsis, and obtain a sepsis prediction result from the artificial intelligence model.

Description

System and method for early diagnosis and prognosis prediction of sepsis

[0001] The present invention relates to a system and method for early diagnosis and prognosis prediction of sepsis.

[0002] The present invention is a result of the research project titled [Development and Validation of the Usefulness of a Sepsis Clinical Decision Support System (CDSS) Based on Liquid Biopsy Data] of the Research Project Name: Regional Innovation Cluster Promotion (R&D) (Ministry of Trade, Industry and Energy, Project Unique Number 141518873, Project Number P0025355).

[0003] Sepsis is a disease in which bacterial infection causes an inflammatory response throughout the body, damaging major organs. The mortality rate for patients with severe sepsis is approximately 25%, but if septic shock occurs due to lack of early treatment, the mortality rate increases to 50%. If sepsis is not diagnosed and treated early, the mortality rate increases by 7-10% per hour.

[0004] Moreover, sepsis is a manageable disease with a high mortality rate, for which early and intensive treatment can prevent organ failure and reduce the severity of the disease to improve treatment outcomes.

[0005] The likelihood of patient death is reduced by 32-40% when bundle therapy is performed within 1 hour as recommended by the Surviving Sepsis Campaign (SSC). However, the reality is that about half of sepsis patients do not receive bundle therapy even after 6 hours due to reasons such as doctors' late recognition of sepsis or difficulty in administering antibiotics within 1 hour.

[0006] To improve the performance of sepsis bundle therapy, a system for early detection of sepsis is important, and rapid response to surveillance and treatment of hospital-acquired sepsis is essential.

[0007] Meanwhile, although hundreds of sepsis biomarkers (e.g., C-reactive protein and procalcitonin, Cytokines, Lipopolysaccharide binding protein, Circulating leukocytes, D-dimer, Presepsin, etc.) exist, their clinical implementation is limited by moderate accuracy, long processing time, and high cost.

[0008] Therefore, the need for an early diagnosis system for sepsis with high accuracy and low cost is increasing.

[0009] The purpose of the present invention is to provide a system and method for early diagnosis and prognosis prediction of sepsis by learning through artificial intelligence using general blood test data.

[0010] The present invention extracts items that may affect the early diagnosis of sepsis and prediction of prognosis from the results of general blood tests of normal people and patients with sepsis, trains an artificial intelligence model for early diagnosis and prognosis prediction of sepsis with a data set including patient information including the patient's vital signs (respiratory rate, pulse, body temperature, blood pressure) and anthropological information (gender, age) and various diagnostic results, and applies the general blood test information and vital signs of a patient suspected of having sepsis to the trained artificial intelligence model for early diagnosis and prognosis prediction of sepsis, thereby providing a system and method for early diagnosis and prognosis prediction of sepsis capable of predicting in real time risk signs that frequently occur in patients with sepsis.

[0011] A system for early diagnosis and prognosis prediction of sepsis according to one embodiment of the present invention comprises: a memory for storing general blood test data of a patient; and a processor, wherein the processor is configured to preprocess the general blood test data of the patient stored in the memory into data for inputting into an artificial intelligence model, input the preprocessed data into an artificial intelligence model trained for early diagnosis and prognosis prediction of sepsis, and obtain a sepsis prediction result from the artificial intelligence model.

[0012] The above artificial intelligence model can be trained by receiving learning data for each of a number of sepsis patients.

[0013] The above processor can preprocess the patient's medical data stored in the memory and input the preprocessed data into the artificial intelligence model.

[0014] The above processor can preprocess the patient's medical data by replacing values ​​that appear as input errors with values ​​within a set range.

[0015] The above processor can preprocess the natural language input manually from the patient's medical data by changing it into a set single expression.

[0016] The above processor can preprocess the natural language input manually from the patient's medical data by changing it into a set single expression.

[0017] The above processor can preprocess the patient's medical data by filtering out data containing special characters and replacing them according to a set pattern.

[0018] The processor may preprocess the patient's medical data by removing or combining at least some of the highly correlated variables.

[0019] The above processor can preprocess the patient's vital information stored in the memory and input the preprocessed data into the artificial intelligence model.

[0020] The above vital information may include at least one of body temperature, blood pressure, pulse, and respiration.

[0021] A method for early diagnosis and prognosis prediction of sepsis according to one embodiment of the present invention may include the steps of: generating an artificial intelligence model for early diagnosis and prognosis prediction of sepsis by learning learning data for each of a plurality of sepsis patients through artificial intelligence; storing general blood test data of the patient in a memory; preprocessing the general blood test data of the patient stored in the memory into data to be input into the learned artificial intelligence model; inputting the preprocessed data into the learned artificial intelligence model; and outputting a sepsis prediction result from the artificial intelligence model.

[0022] The method may include a step of preprocessing the patient's medical data stored in the memory; and a step of inputting the preprocessed data into the artificial intelligence model.

[0023] The above method may include a step of preprocessing by replacing values ​​appearing as input errors in the patient's medical data with values ​​within a set range.

[0024] The method may include a step of preprocessing the patient's medical data by converting the manually entered natural language into a set single expression.

[0025] The method may include a step of preprocessing the patient's vital information stored in the memory; and a step of inputting the preprocessed data into the artificial intelligence model.

[0026] Routine blood tests are inexpensive, widely used in nearly all clinical settings, and their results offer the potential for application across a wide range of conditions. Furthermore, big data approaches to routine blood tests have already proven their usefulness, improving the sensitivity and specificity of disease diagnosis and, consequently, the predictive value of various diseases.

[0027] Because routine blood tests are inexpensive, easy to perform, and available in all wards from emergency rooms to intensive care units, a clinical decision support system utilizing routine blood tests could improve cost-effectiveness in the current healthcare economy without adding additional costs to the healthcare system.

[0028] Therefore, according to the present invention, sepsis can be diagnosed early by learning with artificial intelligence using general blood test data. According to the present invention, early diagnosis of sepsis is possible, and by implementing a sepsis early diagnosis model that distinguishes between normal people and sepsis patients using the patient's gender, age, and CBC test results, it can provide high accuracy and reliability in sepsis diagnosis with a sensitivity of 0.971, a specificity of 0.935, a positive predictive value of 0.892, a negative predictive value of 0.983, a CV accuracy of 0.944±0.039, and an AUC score of 0.987.

[0029] In addition, according to the present invention, septic shock can be predicted with high accuracy, and pulmonary edema and systemic inflammatory response syndrome, which may accompany patients at risk of septic shock, can be distinguished based on the frequency of occurrence and the degree of severity, and the possibility of occurrence of each risk symptom can be predicted.

[0030] FIG. 1 is a diagram showing a sepsis early diagnosis and prognosis prediction model according to one embodiment of the present invention.

[0031] FIG. 2 is a diagram showing a sepsis early diagnosis and prognosis prediction model according to one embodiment of the present invention.

[0032] FIG. 3 is a diagram illustrating the concept of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0033] FIG. 4 is a diagram showing a sepsis early diagnosis and prognosis prediction model according to one embodiment of the present invention.

[0034] FIG. 5 is a block diagram showing a detailed configuration of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0035] FIG. 6 is a block diagram showing a preprocessing unit of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0036] Figure 7 is a block diagram showing the detailed configuration of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0037] Figure 8 is a diagram showing the performance of a sepsis early diagnosis model according to one embodiment of the present invention.

[0038] FIG. 9 is a diagram showing the performance of a sepsis early diagnosis model according to one embodiment of the present invention.

[0039] FIG. 10 is a diagram showing the performance of a risk symptom prediction model according to one embodiment of the present invention.

[0040] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings and the contents described in the attached drawings, but the present invention is not limited or restricted by the embodiments.

[0041] The terminology used herein is for the purpose of describing embodiments only and is not intended to limit the present invention. In this specification, the singular also includes the plural unless the context clearly dictates otherwise. As used herein, the terms "comprises" and / or "comprising" do not exclude the presence or addition of one or more other components, steps, operations, and / or elements.

[0042] The terms “embodiment,” “example,” “aspect,” “example,” and the like as used herein are not to be construed as implying that any aspect or design described is better or advantageous over other aspects or designs.

[0043] Also, the term 'or' means 'inclusive or' rather than 'exclusive or'. That is, unless stated otherwise or clear from context, the expression 'x utilizes a or b' means any one of the natural inclusive permutations.

[0044] Additionally, as used in this specification and claims, the singular forms “a” or “an” should generally be construed to mean “one or more” unless otherwise indicated or clear from the context to be in the singular form.

[0045] Additionally, while the terms "first," "second," etc., used in this specification and claims may be used to describe various components, these components should not be limited by these terms. These terms are used solely to distinguish one component from another.

[0046] Unless otherwise defined, all terms (including technical and scientific terms) used herein may be used in their common sense to those of ordinary skill in the art to which the present invention pertains. Furthermore, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

[0047] Meanwhile, when describing the present invention, if a detailed description of a related known function or configuration is judged to unnecessarily obscure the gist of the present invention, such detailed description will be omitted. Furthermore, the terminology used in this specification is intended to appropriately express embodiments of the present invention and may vary depending on the intent of the user or operator, or the practices of the field to which the present invention pertains. Therefore, the definitions of these terms should be based on the contents throughout this specification.

[0048] FIG. 1 is a diagram showing a sepsis early diagnosis and prognosis prediction model according to one embodiment of the present invention.

[0049] Referring to FIG. 1, the system for early diagnosis and prognosis prediction of sepsis according to one embodiment of the present invention extracts, in a first step, items that may affect early diagnosis and prognosis prediction of sepsis from the results of a complete blood count (CBC) of a plurality of normal people and sepsis patients, and obtains a data set including patient information including vital signs (respiratory rate, pulse, body temperature, blood pressure) of the patient, anthropological information (gender, age) of the patient, and other diagnostic results.

[0050] Next, as a second step, an artificial intelligence model for early diagnosis of sepsis can be created by training the acquired data set on an artificial intelligence model for early diagnosis and prognosis prediction of sepsis.

[0051] Next, in the third step, the routine blood test information and vital signs of patients suspected of having sepsis are applied to the learned sepsis early diagnosis and prognosis prediction AI model, thereby outputting the results of the early diagnosis prediction. In the fourth step, the outputted sepsis early diagnosis prediction results are explained and interpreted, enabling their use as data for Clinical Decision Support Systems (CDSS) for sepsis patients.

[0052] FIG. 2 is a diagram showing a sepsis early diagnosis and prognosis prediction model according to one embodiment of the present invention.

[0053] Referring to FIG. 2, a sepsis early diagnosis and prognosis prediction model according to an embodiment of the present invention may collect EMR (electronic medical record) data of a patient diagnosed with an acute infectious disease and its complications. For example, the EMR data may include registration data, laboratory data, vital signs, and underlying disease data (problem list). For example, the registration data may include the patient's age, gender, and length of hospitalization. The laboratory data may include CBC test data. The vital signs may include respiration, pulse, body temperature, and blood pressure data. In addition, the EMR data may further include other data such as nursing information survey, initial examination records, diagnosis, and medication history data.

[0054] In one embodiment, an artificial intelligence (AI) model for the CDSS can generate a sepsis early diagnosis and prognosis prediction model by learning from the collected EMR data. The AI ​​model may include classical machine learning models and deep learning-based models. For example, the AI ​​model may include a data module and a guideline engine. As a result of the AI ​​learning, screening and management recommendation data may be generated.

[0055] According to one embodiment, the processing by the artificial intelligence model may be processed by at least one processor. The artificial intelligence model may be generated through machine learning. For example, the artificial intelligence model may be performed in the electronic device itself, or may be performed through a separate server. The learning algorithm may include, for example, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but is not limited to the examples described above. The artificial intelligence model may include a plurality of artificial neural network layers. The artificial neural network may be one of a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), deep Q-networks, or a combination of two or more of the above, but is not limited to the examples described above. The above artificial intelligence model may additionally or alternatively include a software structure in addition to the hardware structure.

[0056] The above artificial intelligence model may be one of LSTM (Long Short-Term Memory Model, LSTM), TCN (Temporal Convolutional Network, TCN), Transformer algorithm, or a combination of two or more of the above for detecting anomalies in time series data, but is not limited to the examples described above.

[0057] FIG. 3 is a diagram illustrating the concept of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0058] Referring to FIG. 3, a system for early diagnosis and prognosis prediction of sepsis according to an embodiment of the present invention extracts items that may affect early diagnosis and prognosis prediction of sepsis from the results of complete blood counts (CBC) of a plurality of normal people and sepsis patients as described above, and inputs a data set including vital information (e.g., respiratory rate, pulse, body temperature, blood pressure) of the patients as learning data into artificial intelligence. For example, by inputting the learning data into an artificial intelligence model, an artificial intelligence model for early diagnosis and prognosis prediction of sepsis can be created. Then, by applying the complete blood count (CBC) information and vital information (or vital signs) of a patient suspected of sepsis to the artificial intelligence model for early diagnosis and prognosis prediction of sepsis, a sepsis diagnosis prediction result can be output. For example, the artificial intelligence model can be used to screen patients suspected of having sepsis, diagnose early onset of septic shock within 24 hours of a sepsis patient, and predict the possibility of occurrence of high-frequency risk signs in sepsis patients (e.g., cardiogenic shock, water-electrolyte and acid-base imbalance, respiratory failure, pulmonary edema, intracerebral hemorrhage, non-traumatic intracranial hemorrhage, subarachnoid hemorrhage, respiratory abnormalities, systemic inflammatory response, and asphyxiation).

[0059] FIG. 4 is a diagram showing a sepsis early diagnosis and prognosis prediction model according to one embodiment of the present invention.

[0060] Referring to FIG. 4, a sepsis early diagnosis and prognosis prediction model according to an embodiment can collect data including complete blood count (CBC) results and vital information (e.g., respiratory rate, pulse, body temperature, blood pressure) of a large number of normal people and sepsis patients in the real world, as described above. Meanwhile, the collected data is applied to machine learning and deep learning algorithms when applying structured data to a learning model as general medical data, and in this case, it may be difficult to obtain satisfactory accuracy. Therefore, according to an embodiment, the collected data may undergo a preprocessing process. A detailed embodiment of the preprocessing process will be described in detail in the description of FIG. 7.

[0061] According to one embodiment, items that may affect the early diagnosis and prognosis prediction of sepsis may be extracted as feature data from the preprocessed data, and the corresponding feature data may be selected. Then, an artificial intelligence model may be constructed by training the selected feature data through artificial intelligence. The construction of the artificial intelligence model may be implemented through classical machine learning modeling and deep learning-based modeling as described above. The constructed artificial intelligence model may be optimized through a model evaluation process and a feedback process, thereby generating a final model. For example, the model evaluation and feedback process may include a feature importance plot process for reviewing key model attributes. In addition, the model evaluation and feedback process may include a process for examining whether there is overfitting or underfitting. In addition, a diagnostic accuracy evaluation may be performed as the model evaluation and feedback process. For example, evaluation items for confirming the diagnostic accuracy may include at least one of sensitivity, specificity, positive predictive value, negative predictive value, and AUROC (area under the ROC curve).

[0062] In one embodiment, the AI ​​model can distinguish between normal individuals and sepsis patients by inputting new data, such as the patient's gender, age, and CBC test results. The new data may also undergo preprocessing and feature extraction processes. In other words, the trained AI model can output early diagnosis and prognosis prediction results for sepsis by inputting new patient data.

[0063] FIG. 5 is a block diagram showing a detailed configuration of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0064] Referring to FIG. 5, a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention may include a learning data generation unit (10), a patient data acquisition unit (20), a preprocessing unit (30), a learning unit (40), and a diagnosis unit (50).

[0065] According to one embodiment, the learning data generation unit (10) can generate various learning data for training artificial intelligence. For example, the learning data generation unit (10) can include a CBC item selection unit (11), a vital information acquisition unit (12), a patient data acquisition unit (13), and a data set configuration unit (14). The CBC item selection unit (11) can select CBC items necessary for early diagnosis of sepsis from CBC test results collected from a plurality of patient data. The vital information acquisition unit (12) can acquire vital information (e.g., respiration, pulse, body temperature, blood pressure) from a plurality of patient data. The patient data acquisition unit (13) can acquire data such as patient basic information, nursing information survey, initial examination records, diagnosis names, and medication history from a plurality of patient data. The data set configuration unit (14) can configure a data set in a set format based on the data acquired from the CBC item selection unit (11), the vital information acquisition unit (12), and the patient data acquisition unit (13).

[0066] According to one embodiment, the preprocessing unit (30) may receive a data set configured by the data set configuration unit (14) and perform a preprocessing operation. The preprocessing operation may include outlier processing, natural language processing, special character processing, missing value processing, etc., and a detailed description thereof will be provided later in the description of FIG. 6.

[0067] According to one embodiment, the learning unit (40) may receive a data set processed by the preprocessing unit (30) and train the AI ​​model using the AI ​​model. The learning unit (40) may include an AI model for early diagnosis of sepsis (41) and an AI model for prognosis prediction (42). For example, the learning unit (40) may generate an AI model for early diagnosis of sepsis (41) and an AI model for prognosis prediction (42) through the learning.

[0068] According to one embodiment, the patient data acquisition unit (20) can input and generate various data of a patient suspected of sepsis. For example, the patient data acquisition unit (20) can include a CBC item selection unit (21), a vital information acquisition unit (22), a patient data acquisition unit (23), and a data set configuration unit (24). The CBC item selection unit (21) can select CBC items necessary for early diagnosis of sepsis from the CBC test results for a patient suspected of sepsis. The vital information acquisition unit (22) can acquire vital information (e.g., respiration, pulse, body temperature, blood pressure) for a patient suspected of sepsis. The patient data acquisition unit (23) can acquire data such as basic patient information, nursing information survey, initial examination record, diagnosis name, and medication history for a patient suspected of sepsis. The above data set configuration unit (24) can configure a data set in a set format based on data acquired from the CBC item selection unit (21), vitality information acquisition unit (22), and patient data acquisition unit (23).

[0069] According to one embodiment, the diagnosis unit (50) can output a diagnosis result (e.g., a sepsis prediction result or a risk symptom prediction result) by inputting the acquired data for a patient suspected of sepsis acquired from the patient data acquisition unit (20) into the artificial intelligence model learned by the learning unit (40). For example, the diagnosis unit (50) can output a sepsis prediction result by inputting the acquired data for a patient suspected of sepsis acquired from the patient data acquisition unit (20) into the sepsis early diagnosis AI model (51) learned by the learning unit (40). In addition, the diagnosis unit (50) can output a risk symptom prediction result by inputting the acquired data for a patient suspected of sepsis acquired from the patient data acquisition unit (20) into the prognosis prediction AI model (52) learned by the learning unit (40).

[0070] FIG. 6 is a block diagram showing a preprocessing unit of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0071] Referring to FIG. 6, the preprocessing unit (30) may include an outlier processing unit (31), a natural language processing unit (32), a special character processing unit (33), and a missing value processing unit (34).

[0072] According to one embodiment, the outlier processing unit (31) may perform clipping to replace values ​​that appear as input errors in medical data with values ​​within a set range. For example, there may be outliers that occur due to decimal point input errors, such as when 36.36 is entered as 363.6 in body temperature or 36.5 is entered as 365. The outlier processing unit (31) may perform clipping to set a range of values ​​generally observed in patients for these outliers, replace values ​​below the range with the minimum value of the range, and replace values ​​exceeding the range with the maximum value of the range.

[0073] According to one embodiment, the natural language processing unit (32) may preprocess the natural language existing in the medical data as the patient's condition is manually entered. For example, the natural language included in the medical data may exist in an inconsistent form. For example, in the respiratory data set, 'Ambubagging', which refers to the action of supplying oxygen with a manual device to a patient who has difficulty breathing on his / her own, is manually written in various expressions such as 'ambubag', 'ambu', 'ambubaging', and 'Ambu'. The natural language processing unit (32) may filter the corresponding values, visualize them through a word cloud, and change them into consistent words. For example, the natural language processing unit (32) may perform a preprocessing process to change the various expressions into a single expression (e.g., ambubag).

[0074] In one embodiment, the special character processing unit (33) may filter and replace data containing special characters in medical data manually recorded by medical staff. For example, the special character processing unit (33) may filter non-numeric values ​​as natural language, and process values ​​containing numbers based on patterns to convert them into numeric values. For example, blood pressure data may be replaced according to the following pattern.

[0075] 1) Change the values ​​displayed as ranges to median (e.g., if 88 to 90, 89)

[0076] 2) When measurements and comments are combined, comments are filtered and saved separately.

[0077] 3) The changed value is changed to the value after the change (e.g., if 59 is changed to 133, 133)

[0078] 4) Values ​​containing '>' or '<' symbols are converted to their numerical values ​​(e.g., >100 is converted to 100)

[0079] In one embodiment, the missing value processing unit (34) can preprocess multiple missing values ​​in medical data due to differences in test items performed depending on the patient's condition. For example, test items for a general blood test can vary significantly depending on the patient's condition, which can cause multiple missing values ​​in the test results for the same patient. Furthermore, the CBC test exhibits multicollinearity, indicating a high correlation between test items. The causes of multicollinearity in the CBC test are as follows. First, the CBC test measures various types of cells and components in blood, and these items are interrelated. For example, hemoglobin and hematocrit are closely related to red blood cells, and red blood cell count and hematocrit are also closely related. Second, CBC test results vary depending on the patient's physiological or health condition, affecting various CBC items such as infection and inflammation, thereby increasing the correlation. Third, some of the CBC test result items have different units or scales of measurement. For example, the classification of white blood cells, such as lymphocytes, monocytes, neutrophils, basophils, and eosinophils, has a high correlation between the two variables by displaying the proportion (%) of the total white blood cells and the number (#) thereof as separate items. One of the ways to resolve the large number of missing values ​​and multicollinearity of these CBC test items is feature extraction, and the missing value processing unit (34) can perform preprocessing by removing or combining some of the highly correlated variables. For example, variables with a high correlation of 0.9 or higher can be removed through correlation analysis between the attributes of the CBC test result data set.

[0080] Figure 7 is a block diagram showing the detailed configuration of a sepsis early diagnosis and prognosis prediction system according to one embodiment of the present invention.

[0081] Referring to FIG. 7, a sepsis early diagnosis and prognosis prediction system according to one embodiment may include a communication unit (110), an output unit (120), a memory (130), a user input unit (140), and a processor (150).

[0082] According to one embodiment, the user input unit (140) may receive learning data generated by the learning data generation unit (10) of FIG. 5. In addition, the user input unit (140) may receive patient data acquired by the patient data acquisition unit (20) of FIG. 5. Data input through the user input unit (140) may be stored in the memory (130). The processor (150) may receive data input through the user input unit (140) and train an artificial intelligence model, or output a result through the trained artificial intelligence model. For example, the processor (150) may process operations performed in the learning unit (40) or the diagnosis unit (50) of FIG. 5. A diagnosis result (e.g., a sepsis prediction result or a risk symptom prediction result) processed by the processor (150) may be output through the output unit (120) (e.g., a display).

[0083] According to one embodiment, the learning by the artificial intelligence may be performed by the processor (150) as described above, or may be performed by an external server. For example, the communication unit (110) may transmit learning data input through the user input unit (140) to an external server. The external server may receive the learning data received through the communication unit (110) and train it using an artificial intelligence model. In addition, the processor (150) may receive patient data and transmit it to the external server through the communication unit (110). The external server may input the patient data transmitted from the processor (150) into the trained artificial intelligence model to output a diagnosis result, and transmit the output diagnosis result to the sepsis early diagnosis and prognosis prediction system of FIG. 5. The system may output the diagnosis result received from the external server through the output unit (120) (e.g., a display).

[0084] FIG. 8 and FIG. 9 are diagrams showing the performance of a sepsis early diagnosis model according to one embodiment of the present invention.

[0085] Referring to FIGS. 8 and 9 , according to one embodiment, a sepsis early diagnosis model can be implemented that distinguishes between normal individuals and sepsis patients using patient gender, age, and CBC test results. As illustrated in FIGS. 8 and 9 , the model according to the present invention exhibits high accuracy and reliability, with a sensitivity of 0.971, a specificity of 0.935, a positive predictive value of 0.892, a negative predictive value of 0.983, a CV accuracy of 0.944±0.039, and an AUC score of 0.987.

[0086] FIG. 10 is a diagram showing the performance of a risk symptom prediction model according to one embodiment of the present invention.

[0087] Referring to Figure 10, according to one embodiment, nine risk symptoms, such as pulmonary edema and systemic inflammatory response syndrome, that may accompany patients at risk of septic shock can be identified based on their frequency and severity, and the likelihood of each risk symptom can be predicted. As illustrated in Figure 10, it can be confirmed that most symptoms can be identified with high accuracy and reliability.

[0088] The devices described above may be implemented as hardware components, software components, and / or a combination of hardware components and software components. For example, the devices and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.

[0089] Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure a processing device to perform a desired operation or, independently or collectively, command the processing device. The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave, for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0090] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., alone or in combination. The program commands recorded on the medium may be those specially designed and configured for the embodiment or may be those known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of the program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

[0091] Although the embodiments described above have been described by way of limited examples and drawings, those skilled in the art will appreciate that various modifications and variations can be made based on the above teachings. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0092] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims described below.

Claims

1. In the early diagnosis and prognosis prediction system for sepsis, Memory for storing the patient's general blood test data; and comprising a processor, said processor comprising: Preprocessing the patient's general blood test data stored in the above memory into data for inputting into an artificial intelligence model, The above preprocessed data is input into the learned artificial intelligence model for early diagnosis and prognosis prediction of sepsis. A system for early diagnosis and prognosis prediction of sepsis, set to obtain sepsis prediction results from the above artificial intelligence model.

2. In paragraph 1, the artificial intelligence model, A system for early diagnosis and prognosis prediction of sepsis, which learns by inputting learning data for each of a number of sepsis patients.

3. In the first paragraph, the processor, A system for early diagnosis and prognosis prediction of sepsis, which preprocesses the medical data of the patient stored in the memory and inputs the preprocessed data into the artificial intelligence model.

4. In the third paragraph, the processor, A system for early diagnosis and prognosis prediction of sepsis, which preprocesses values ​​that appear as input errors in the medical data of the above patient by replacing them with values ​​within a set range.

5. In the third paragraph, the processor, A system for early diagnosis and prognosis prediction of sepsis by preprocessing manually entered natural language from the medical data of the above patient into a set single expression.

6. In the third paragraph, the processor, A system for early diagnosis and prognosis prediction of sepsis by preprocessing manually entered natural language from the medical data of the above patient into a set single expression.

7. In the 6th paragraph, the processor, A system for early diagnosis and prognosis prediction of sepsis, which preprocesses data containing special characters in the medical data of the above patient by filtering the data and replacing it according to a set pattern.

8. In the third paragraph, the processor, A system for early diagnosis and prognosis prediction of sepsis, which preprocesses the medical data of the patient by removing or combining at least some of the highly correlated variables.

9. In paragraph 1, the processor, A system for early diagnosis and prognosis prediction of sepsis, which preprocesses the vital information of the patient stored in the memory and inputs the preprocessed data into the artificial intelligence model.

10. In paragraph 1, the vitality information is: A system for early diagnosis and prognosis prediction of sepsis, comprising at least one of body temperature, blood pressure, pulse, and respiration.

11. In the method of early diagnosis and prognosis prediction of sepsis, A step of creating an artificial intelligence model for early diagnosis and prognosis prediction of sepsis by learning learning data for each of a number of sepsis patients through artificial intelligence; A step of storing the patient's general blood test data into memory; A step of preprocessing the patient's general blood test data stored in the memory into data for inputting into the learned artificial intelligence model; A step of inputting the preprocessed data into the learned artificial intelligence model; and A method for early diagnosis and prognosis prediction of sepsis, comprising a step of outputting a sepsis prediction result from the artificial intelligence model.

12. In the 11th paragraph, the method, A step of preprocessing the medical data of the patient stored in the above memory; and A method for early diagnosis and prognosis prediction of sepsis, comprising a step of inputting the above preprocessed data into the above artificial intelligence model.

13. In paragraph 12, the method, A method for early diagnosis and prognosis prediction of sepsis, comprising a step of preprocessing by replacing values ​​appearing as input errors in the medical data of the patient with values ​​within a set range.

14. In paragraph 12, the method, A method for early diagnosis and prognosis prediction of sepsis, comprising a step of preprocessing by changing manually entered natural language from the medical data of the patient into a set single expression.

15. In paragraph 11, the method, A step of preprocessing the vital information of the patient stored in the memory; and A method for early diagnosis and prognosis prediction of sepsis, comprising a step of inputting the above preprocessed data into the above artificial intelligence model.

Images