Learning system and method for diagnosing liver disease by using data from all levels among intestinal microbiota

Abstract

The present invention relates to a learning system and method, in which an artificial intelligence model analyzes intestinal microbiota extracted from a patient's stool and automatically diagnoses alcoholic liver disease on the basis of the results of the analysis. The learning system comprises: a data reception unit that receives biological classification data of a patient's intestinal microbiota; a data selection unit that selects, from among the received biological classification data of the intestinal microbiota, intestinal microbiota having a high correlation to disease diagnosis at each level by using data from all levels, so as to construct a data sample; a data correction unit that artificially generates intestinal microbiota data corresponding to the number of patients per disease and a disease of the patient, so as to correct the data sample; and a data learning unit that constructs learning data allowing the results of disease diagnosis to be inferred by using the data sample corrected by the data correction unit at each level, and learns the learning data.

Description

Learning system and method for diagnosing liver disease using data from the entire intestinal microbiota

[0001] The present invention relates to a learning system and method for diagnosing liver disease using an artificial intelligence model, and more particularly, to a system and method for constructing learning data that enables an artificial intelligence model to diagnose alcoholic liver disease based on a microbiome, and for an artificial intelligence model to learn from the constructed learning data.

[0002] Humans are a super-organism, a mixture of human cells and genomes, as well as the cells and genetic characteristics of countless microorganisms that coexist with us [Nature 2007;449:804-810]. The human microbiome is the collective of all microbial communities naturally present in the human body. The human microbiota exists in various body parts, including the skin, oral cavity, teeth, genitals, respiratory tract, and gastrointestinal tract, but the gastrointestinal tract harbors the largest number and greatest diversity of microorganisms [Nature 2012;486:215-221].

[0003] The gastrointestinal tract harbors a diverse community of microorganisms, numbering between 4,000 and 10,000 species, of which approximately 1,565 are known to be associated with liver disease. The concentration and diversity of microbial communities (especially bacteria) within the gastrointestinal tract vary considerably by location, with the stomach having the lowest concentration, while the large intestine has the highest concentration.

[0004] Understanding the fundamental structure, dynamics, diversity, and stability of the human microbiome is largely due to large-scale research led by global microbiome consortia. Representative examples include the Human Microbiome Project (HMP), supported by the US National Institutes of Health (NIH), the Metagenomics of the human intestinal tract (Meta-HIT), supported by the European Commission, and the International Human Microbiome Consortium (IHMC), which aims to share data resources through human microbiome research.

[0005] Referring to research results from various institutions, it is known that various microorganisms within the human body influence all functions, including metabolic regulation, digestive function, and various diseases. They also influence genetic mutations due to environmental changes and the process of genetic transmission to the next generation. In particular, various metabolic and immune disorders related to allergies, rhinitis, atopy, obesity, enteritis, heart disease, and liver disease have been reported to be related to the microbiome.

[0006] What we can learn from these reports from various institutions is that the microbiome influences or is correlated with specific diseases. Therefore, if we can identify the microbiome that is related to or influences specific diseases, it will be able to contribute to disease diagnosis and prevention.

[0007] As conventional techniques for diagnosing liver disease, Patent Publication No. 10-2021-0061742 and Patent Publication No. 10-2019-0020581 are known.

[0008] Patent Publication No. 10-2021-0061742 discloses a biomarker composition for predicting the prognosis of chronic liver disease, comprising PODN (podocan), C7 and EPHA3 proteins or a gene encoding them as active ingredients.

[0009] In addition, Patent Publication No. 10-2019-0020581 discloses a composition for predicting or diagnosing liver disease, which comprises an agent for measuring the expression of a nucleic acid sequence encoding growth differentiation factor 15 (GDF15) protein or the activity of the GDF15 protein.

[0010] However, according to the prior art disclosed in Patent Publication No. 10-2021-0061742 and Patent Publication No. 10-2019-0020581, there is a problem that pain and side effects occur due to the use of invasive methods such as removing a portion of a patient's liver tissue to collect a sample for predicting or diagnosing liver disease.

[0011] Therefore, when building a disease diagnosis system using an artificial neural network, if the artificial neural network model is pre-trained using the microbiome (hereinafter referred to as “microorganisms”) that affect the disease, a system that can diagnose the disease more safely and conveniently can be built.

[0012] [Patent Document]

[0013] Patent Publication No. 10-2021-0061742 (Published on May 28, 2021, Title: Biomarker Composition for Predicting the Prognosis of Chronic Liver Disease)

[0014] Patent Publication No. 10-2019-0020581 (Published on March 4, 2019, Title: Composition for Predicting or Diagnosing Liver Disease and Method for Predicting or Diagnosing Liver Disease Using the Same)

[0015]

[0016] The technical task of the present invention is to provide a learning system for liver disease diagnosis by extracting intestinal microorganisms correlated with alcoholic liver disease from the stool of a patient and enabling an artificial intelligence model to make a rapid and accurate diagnosis.

[0017] Furthermore, another object of the present invention is to provide a learning method for liver disease diagnosis, which selects intestinal microorganisms having a high correlation with disease diagnosis, artificially generates data to resolve data imbalance, constructs liver disease diagnosis learning data for an artificial intelligence model to automatically diagnose alcoholic liver disease, and learns the constructed learning data.

[0018] In order to solve such technical problems, the learning system for diagnosing liver disease using data of the entire level of intestinal microorganisms according to the present invention comprises: a data receiving unit for receiving biological classification data of intestinal microorganisms contained in the stool of a patient; a data selection unit for selecting intestinal microorganisms having a high correlation with disease diagnosis for each level using data of the entire level among the biological classification data of intestinal microorganisms received from the data receiving unit to form a data sample in order to output a uniform data result value; a data correction unit for correcting the data sample by artificially generating data of intestinal microorganisms similar to the number of patients for each disease and the disease of the corresponding patient in order to resolve an imbalance in the data sample; and a data learning unit for constructing learning data capable of inferring a disease diagnosis result using the data sample corrected by the data correction unit for each level, and learning the learning data.

[0019] The above biological classification data is characterized by including the types and expression levels of the intestinal microorganisms contained in the stool of the patient.

[0020] The above data selection unit selects the intestinal microorganisms of the entire level at each level using a first artificial intelligence model, and repeats this process at least twice to select the intestinal microorganisms having a high correlation to disease diagnosis, and the data learning unit learns learning data that can infer the disease diagnosis result using a second artificial intelligence model.

[0021] The above disease is characterized by including at least one of normal, alcoholic fatty liver, alcoholic hepatitis, and alcoholic cirrhosis.

[0022] The above data correction unit is characterized in that it corrects the data sample by artificially generating the number of patients for each disease and intestinal microorganism data similar to the disease of the corresponding patient using upsampling.

[0023] Meanwhile, a learning method for diagnosing liver disease using full-level data according to another embodiment of the present invention is as follows:

[0024] The method is characterized by including the steps of: receiving bio-classification data of intestinal microorganisms contained in the stool of at least one patient; selecting intestinal microorganisms having a high correlation with disease diagnosis at each level of the data from the entire level of the bio-classification data of the received intestinal microorganisms to form a data sample using a first artificial intelligence model in order to output a uniform data result value; correcting the data sample by artificially generating the number of patients for each disease and intestinal microorganism data similar to the disease of the corresponding patient in order to resolve the imbalance of the data sample; and constructing learning data capable of inferring a disease diagnosis result using the corrected data sample at each level, and learning the learning data using a second artificial intelligence model.

[0025] The step of selecting the intestinal microorganisms is characterized in that the intestinal microorganism selection process is repeated at least twice for each level using the first artificial intelligence model to select the intestinal microorganisms having a high correlation with the diagnosis of the disease.

[0026] As a result of the above disease, it is characterized by outputting at least one of normal, alcoholic fatty liver, alcoholic hepatitis, and alcoholic cirrhosis.

[0027] Specific details of other embodiments are included in the detailed description and drawings.

[0028] According to the technical problem solving means described above, the learning system and method for diagnosing liver disease using data at the entire level of intestinal microorganisms according to the embodiment of the present invention has the following advantages because intestinal microorganisms extracted from a patient's stool can be used as a data sample for an artificial intelligence model.

[0029] First, a learning system and method for diagnosing liver disease using data at the overall level of intestinal microorganisms can build a system that can automatically diagnose alcoholic fatty liver disease, alcoholic hepatitis, and alcoholic cirrhosis simply by collecting stool from a patient requiring diagnosis.

[0030] Second, the learning system and method for diagnosing liver disease using data at the entire level of intestinal microorganisms can prevent the risk of overfitting and excessive memory usage during artificial intelligence model learning by selecting intestinal microorganisms with a high correlation to disease diagnosis, and can facilitate model interpretation by users by optimizing the number of data to the level necessary for model interpretation.

[0031] Third, a learning system and method for diagnosing liver disease using data at the overall level of intestinal microorganisms can resolve data imbalance by artificially generating data for disease samples, since the number of samples from patients with each disease is insufficient compared to normal patients.

[0032] Fourth, the learning system and method for diagnosing liver disease using data at the entire level of intestinal microorganisms can predict uniform results without being biased toward a specific level because it combines the result values ​​of each level obtained by analyzing the entire level (species, genus, family, order, class, and phylum) data extracted from the patient's stool and outputs the final result value.

[0033] Figure 1 illustrates a learning system for diagnosing liver disease using data at the overall level of intestinal microorganisms according to an embodiment of the present invention.

[0034] Figure 2 illustrates a learning process for diagnosing liver disease using data at the overall level of intestinal microorganisms.

[0035] Figure 3 illustrates the basic process of a learning system for diagnosing liver disease.

[0036] Figure 4 illustrates a full level data process using the full level of biological classification data.

[0037] The following detailed description of the present invention refers to the accompanying drawings, which illustrate specific embodiments in which the present invention may be practiced, in order to clarify the objects, techniques, solutions, and advantages of the present invention. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present invention. Furthermore, throughout the detailed description and claims, the word "comprise" and variations thereof are not intended to exclude other technical features, additions, components, or steps. Other objects, advantages, and features of the present invention will become apparent to those skilled in the art, in part from this description, and in part from practice of the present invention. The examples and drawings below are provided by way of illustration and are not intended to limit the present invention. Moreover, the present invention encompasses all possible combinations of the embodiments set forth herein. It should be understood that the various embodiments of the present invention, while different from one another, are not necessarily mutually exclusive. It should also be understood that the positions or arrangements of individual components within each disclosed embodiment may be varied without departing from the spirit and scope of the present invention. Accordingly, the detailed description set forth below is not intended to be limiting, and the scope of the present invention is defined solely by the appended claims, along with the full scope equivalent to which such claims are entitled, if properly described. Furthermore, unless otherwise indicated herein or clearly contradicted by context, items referred to in the singular encompass the plural unless the context otherwise requires. Furthermore, in describing the present invention, if a detailed description of a related known structure or function is determined to obscure the gist of the present invention, the detailed description will be omitted.

[0038] FIG. 1 illustrates a learning system for diagnosing liver disease using data at the entire level of intestinal microorganisms according to an embodiment of the present invention, and FIG. 2 illustrates a learning process for diagnosing liver disease using data at the entire level of intestinal microorganisms.

[0039] Before describing FIGS. 1 and 2, a learning system and method for diagnosing liver disease using intestinal microbiome data at the entire level according to an embodiment of the present invention can diagnose alcoholic liver disease using intestinal microbiome data extracted from a patient's stool through Next Generation Sequencing (NGS) testing. For reference, NGS testing extracts data on the biological classification of intestinal microbiomes by genome decoding and analyzing the patient's stool.

[0040] Here, biological classification means organizing organisms into species, genus, family, order, class, and phylum in order based on their similarities and biological forms. Hereinafter, data extracted from patient stool through NGS testing is referred to as biological classification data.

[0041] In addition, although the embodiment of the present invention describes a learning system and method for automatically diagnosing alcoholic liver disease, it is not limited thereto, and a learning system for diagnosing liver disease using data at the overall level of intestinal microorganisms can also perform learning for automatically diagnosing liver disease caused by various factors such as alcohol, drugs, viruses, and eating habits.

[0042] Referring to FIGS. 1 and 2, a learning system (100) for diagnosing liver disease using data at the entire level of intestinal microorganisms includes a data receiving unit (110), a data selection unit (120), a data correction unit (130), a data learning unit (140), a data storage unit (150), a data output unit (160), and a control unit (170).

[0043] The data receiving unit (110) can receive biological classification data of intestinal microorganisms contained in the patient's stool.

[0044] Biological taxonomy data may include data (Data1, Data2, Data3, ... Datan) obtained by extracting the types and expression levels of intestinal microorganisms from the patient's stool through NGS testing.

[0045] According to an embodiment of the present invention, intestinal microorganisms contained in the stool of a patient include, but are not limited to, at least one biological classification among species, genus, family, order, class, and phylum, and may include at least one biological classification among species, genus, family, order, class, phylum, kingdom, and domain.

[0046] The data selection unit (120) uses data of all levels (species, genus, family, order, class, phylum) among the biological classification data of intestinal microorganisms received from the data receiving unit (110) to output uniform data results, and the first artificial intelligence model can select intestinal microorganisms of all levels of data that have a high correlation with disease diagnosis at each level among the all levels of data to form data samples (Data6, Data9, Data10, Data13, ... Datan).

[0047] According to an embodiment of the present invention, the first artificial intelligence model may be a machine learning model, and according to an embodiment of the present invention, the first artificial intelligence model may use various machine learning models such as Logistic Regression, Support Vector Machine, Naive Bayes, Decision Tree, k Nearest Neighbors, and Random Forest.

[0048] The data selection unit (120) can select intestinal microorganisms with a high correlation to disease diagnosis by repeating the selection process of intestinal microorganisms with a high correlation to disease diagnosis at least twice among all level data using the first artificial intelligence model.

[0049] According to an embodiment of the present invention, as can be seen in FIG. 2, the data selection unit (120) can configure a data sample by using the results of the first selection (Data1, Data6, Data8, Data9,... Datan) from the biological classification data (Data1, Data2, Data3, Data4,... Datan) and the results of the second selection (Data6, Data9, Data10, Data13,... Datan).

[0050] In this way, if the step of selecting gut microbes is repeated at least twice, the risk of overfitting and excessive memory usage can be prevented during AI model training, and the amount of data can be reduced to a level optimized for model interpretation, making it easier for users to interpret the model.

[0051] The data correction unit (130) can correct the data sample by artificially generating the number of patients with each disease and intestinal microorganism data similar to the disease of the corresponding patient to resolve the imbalance of the data sample.

[0052] The data correction unit (130) can correct unbalanced data by using an upsampling technique when artificially generating data on the number of patients with each disease and intestinal microorganisms similar to the disease of the corresponding patient.

[0053] According to an embodiment of the present invention, the upsampling technique is a method of increasing the number of data by expressing that more data is collected than actually collected, but not actually measured.

[0054] For example, when the number of patients by disease is 100 normal, 80 fatty liver, 50 hepatitis, and 70 cirrhosis, the data correction unit (130) artificially creates the number of patients by disease as 100 normal, 100 fatty liver, 100 hepatitis, and 100 cirrhosis to maintain balance.

[0055] For reference, data imbalances can occur because it is more difficult to collect samples from patients with diseases than from normal patients. To address this, the data correction unit (130) artificially generates similar data for each disease being diagnosed.

[0056] In addition, the data correction unit (130) may generate data for insufficient diseases to be diagnosed based on the disease with the largest number of patients, or may artificially reduce or generate data for diseases to be diagnosed so that the data is the same as the average by averaging the total number of patients.

[0057] The data learning unit (140) configures learning data that can infer disease diagnosis results by using data samples corrected by the data correction unit (130) for each level, and the second artificial intelligence model can learn the learning data.

[0058] Here, the disease includes at least one of normal, alcoholic fatty liver, alcoholic hepatitis, and alcoholic cirrhosis.

[0059] According to an embodiment of the present invention, the second artificial intelligence model may be a Boosting-based machine learning model specialized for tabular data, but is not limited thereto, and the second artificial intelligence model may use various boosting-based algorithms such as AdaBoost, GBM (Gradient Boost Machine), XGBoost (Extreme Gradient Boosting), and Light GBM (Light Gradient Boost Machine).

[0060] Additionally, the second artificial intelligence model using a boosting-based algorithm uses a learning method that weights the error data of the iteration to learn the next model and a learning method that trains multiple models on the data and then aggregates the results.

[0061] According to an embodiment of the present invention, the data selection unit (120) uses a first artificial intelligence model, and the data learning unit (140) uses a second artificial intelligence model, but the first artificial intelligence model and the second artificial intelligence model use different models.

[0062] The data storage unit (150) stores learning data configured based on the corrected data sample, and the second artificial intelligence can use the learning data stored in the data storage unit (150).

[0063] The data output unit (160) can provide the user with the results learned using the second artificial intelligence model in the data learning unit (140).

[0064] According to an embodiment of the present invention, the data output unit (160) provides the user with the results learned using the second artificial intelligence model, so that the user can check whether the results learned have a high correlation with disease diagnosis.

[0065] The control unit (170) generally corresponds to a server of a computer system.

[0066] The control unit (170) performs overall control of the data receiving unit (110), data selection unit (120), data correction unit (130), data learning unit (140), data storage unit (150), and data output unit (160).

[0067] Figure 3 illustrates the basic process of a learning system for diagnosing liver disease.

[0068] Referring to FIG. 3, a learning system for diagnosing liver disease can analyze a patient's biological classification data as follows to configure and learn learning data capable of diagnosing the patient's disease.

[0069] A step is performed in which the data receiving unit (110) receives the biological classification data of intestinal microorganisms contained in the patient's stool (S100).

[0070] Using the biological classification data of intestinal microorganisms received from the data receiving unit (110), the first artificial intelligence model of the data selection unit (120) selects intestinal microorganisms with a high correlation to disease diagnosis and proceeds with the step of composing a data sample (S110).

[0071] At this time, the data selection unit (120) repeats the process of selecting intestinal microorganisms with a high correlation to disease diagnosis at least twice using the first artificial intelligence model to select the intestinal microorganisms with a high correlation to disease diagnosis and proceeds with the step of composing a data sample (S120).

[0072] To resolve the imbalance of the data sample, the data correction unit (130) performs a step of correcting the data sample by artificially generating the number of patients for each disease and intestinal microorganism data similar to the disease of the corresponding patient (S130).

[0073] The data learning unit (140) constructs learning data that can infer disease diagnosis results using the corrected data sample, and proceeds with a step in which the second artificial intelligence model of the data learning unit (140) learns the learning data (S140).

[0074] Figure 4 illustrates a full level data process using the full level of biological classification data.

[0075] Referring to FIG. 4, the entire level data process analyzes each level of species, genus, family, order, class, and phylum in the biological classification data in parallel to configure learning data for each level, and can learn the learning data configured for each level.

[0076] The overall level data process proceeds with a step of receiving bio-classification data of intestinal microorganisms contained in the stool of at least one patient from a data receiving unit (110) (S200).

[0077] The basic process is carried out as follows for each level using the intestinal microorganism classification data received from the data receiving unit (110) (S210).

[0078] The step of performing the basic process by level is, first, the data receiving unit (110) performs the step of receiving biological classification data of intestinal microorganisms contained in the stool of at least one patient.

[0079] The data selection unit (120) selects intestinal microorganisms with a high correlation to disease diagnosis by level among the received intestinal microorganism classification data at all levels (species, genus, family, order, class, phylum) to output uniform data results, and proceeds with a step of configuring a data sample using the first artificial intelligence model.

[0080] At this time, the step of selecting intestinal microorganisms can be performed by repeating the intestinal microorganism selection process at least twice for each level using the first artificial intelligence model to select intestinal microorganisms having a high correlation with the diagnosis of the disease.

[0081] Next, in order to correct the imbalance of the data sample, the data correction unit (130) can correct the data sample by artificially generating the number of patients with each disease and intestinal microorganism data similar to the disease of the corresponding patient by level.

[0082] For each level, the data learning unit (140) uses the data sample corrected by the data correction unit (130) to construct learning data that can infer the disease diagnosis results, and the second artificial intelligence model proceeds with the step of learning the learning data, completing the basic process that is performed in parallel for each level.

[0083] The second artificial intelligence can diagnose diseases from intestinal microorganisms in the patient's stool by learning learning data, and the results of the disease can include at least one of normal, alcoholic fatty liver, alcoholic hepatitis, and alcoholic cirrhosis.

[0084] For reference, the learning process using the entire level of biological classification data can predict unbiased and uniform results.

[0085] Although the present invention has been described with reference to the embodiments shown in the drawings, these are merely exemplary, and those skilled in the art will understand that various modifications and equivalent other embodiments are possible from the embodiments of the present invention. For example, the embodiments of the present invention have been described assuming a learning system for diagnosing liver disease using data on the entire level of intestinal microorganisms, but the present invention can be equally applied to all cases related to a learning system for disease diagnosis by analyzing biological classification data of intestinal microorganisms without any particular modification. Therefore, the true technical protection scope of the present invention should be defined solely by the appended claims.

[0086] [Explanation of symbols]

[0087] 100: A learning system for diagnosing liver disease using data from the entire gut microbiome.

[0088] 110: Data receiving unit

[0089] 120: Data Selection Department

[0090] 130: Data Correction Department

[0091] 140: Data Learning Department

[0092] 150: Data storage

[0093] 160: Data output section

[0094] 170: Control unit

Claims

1. A data receiving unit for receiving data on the biological classification of intestinal microorganisms contained in the patient's stool; A data selection unit that selects intestinal microorganisms having a high correlation to disease diagnosis at each level by using data of the entire level of the intestinal microorganism classification data received from the data receiving unit to form a data sample in order to output uniform data results; A data correction unit that corrects the data sample by artificially generating the number of patients by disease and intestinal microorganism data similar to the disease of the corresponding patient to resolve the imbalance of the data sample; and A learning system for diagnosing liver disease using data of the entire level of intestinal microorganisms, characterized in that it includes a data learning unit that learns the learning data and configures learning data capable of inferring disease diagnosis results using the corrected data sample from the data correction unit at each level.

2. In paragraph 1, A learning system for diagnosing liver disease using data at the overall level of intestinal microorganisms, characterized in that the above biological classification data includes the types and expression levels of intestinal microorganisms contained in the stool of the patient.

3. In paragraph 1, The above data selection unit selects the intestinal microorganisms of each level using the first artificial intelligence model, and repeats this process at least twice to select the intestinal microorganisms that have a high correlation with disease diagnosis. A learning system for diagnosing liver disease using data at the entire level of intestinal microorganisms, characterized in that the above data learning unit learns learning data capable of inferring the results of the disease diagnosis using a second artificial intelligence model.

4. In paragraph 1, A learning system for diagnosing liver disease using data on the entire level of intestinal microbiota, wherein the disease comprises at least one of normal, alcoholic fatty liver, alcoholic hepatitis, and alcoholic cirrhosis.

5. In paragraph 1, A learning system for diagnosing liver disease using data at the entire level of intestinal microorganisms, characterized in that the above data correction unit uses upsampling to artificially generate data on the number of patients by disease and intestinal microorganisms similar to the disease of the corresponding patient to correct the data sample.

6. A step of receiving biotaxonomy data of intestinal microorganisms contained in the stool of at least one patient; A step of selecting intestinal microorganisms having a high correlation to disease diagnosis at each level among the received intestinal microorganism classification data for the entire level of data to output uniform data results, and configuring a data sample using a first artificial intelligence model; A step of correcting the data sample by artificially generating the number of patients by disease and intestinal microorganism data similar to the disease of the corresponding patient to resolve the imbalance of the data sample; and A learning method for diagnosing liver disease using data of the entire level of intestinal microorganisms, characterized by including the steps of: configuring learning data capable of inferring disease diagnosis results using the above-mentioned corrected data samples for each level, and learning the learning data using a second artificial intelligence model; 7. In paragraph 6, The step of selecting the above intestinal microorganisms is: A learning method for diagnosing liver disease using data at all levels of intestinal microorganisms, characterized in that the intestinal microorganism selection process is repeated at least twice by level using the first artificial intelligence model to select intestinal microorganisms having a high correlation with the diagnosis of the disease.

8. In paragraph 6, A learning method for diagnosing liver disease using data of the entire level of intestinal microorganisms, characterized in that at least one of normal, alcoholic fatty liver, alcoholic hepatitis, and alcoholic cirrhosis is output as a result of the above disease.

Images